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 1098, Wed Feb 27 17:29:47 2002 UTC revision 1211, Tue May 21 14:54:32 2002 UTC
# Line 14  Line 14 
14    
15  ----------------------------------------------------------------------  ----------------------------------------------------------------------
16  Name: Matthias Blume  Name: Matthias Blume
17    Date: 2002/05/21 10:55:00 EDT
18    Tag: blume-20020521-misc
19    Description:
20    
21    - John tweaked runtime to be silent on heap export (except when
22      GC messages are on).
23    - I added a few more things (cross-compiling versions of CMB) to
24      config/preloads (as suggestions).
25    
26    ----------------------------------------------------------------------
27    Name: Matthias Blume
28    Date: 2002/05/20 22:25:00 EDT
29    Tag: blume-20020520-controls
30    Description:
31    
32    - Added ControlUtil structure to control-lib.cm.
33    - Use it throughout.
34    - Used Controls facility to define MLRISC controls (as opposed to
35      registering MLRISC control ref cells with Controls after the fact)
36    - Fixed messed-up controls priorities.
37    
38    * Removed again all the stuff from config/preloads that one wouldn't
39      be able to preload at the time the initial heap image is built.
40      (Many libraries, e.g., CML, do not exist yet at this time.  The
41       only libraries that can be preloaded via config/preloads are those
42       that come bundled with the bootfiles.)
43    
44    ----------------------------------------------------------------------
45    Name: Matthias Blume
46    Date: 2002/05/20 10:59:00 EDT
47    Tag: blume-20020520-preloads
48    Description:
49    
50    Added a lot of commented-out suggestions for things to be included
51    in config/preloads.
52    
53    ----------------------------------------------------------------------
54    Name: Allen Leung
55    Date: 2002/05/18 14:20:00 EDT
56    Tag: leunga-20020518-mdl
57    Description:
58    
59    o Made the mdl tool stuff compile and run again.
60    o I've disabled all the stuff that depends on RTL specifications; they
61      are all badly broken anyway.
62    
63    ----------------------------------------------------------------------
64    Name: Matthias Blume
65    Date: 2002/05/17 16:49:00 EDT
66    Tag: blume-20020517-controls
67    Description:
68    
69    0. John Reppy made several modifications to the SML/NJ library.
70       In particular, there is a shiny new controls-lib.cm.
71    
72    1. Pushed new controls interface through compiler so that everything
73       compiles again.
74    
75    2. Added FormatComb and FORMAT_COMB to the CML version of the
76       SML/NJ library (so that CML compiles again).
77    
78    3. Modified init scripts because XXX_DEFAULT environment variables
79       are no longer with us.  (Boot-time initialization is now done
80       using the same environment variables that are also used for
81       startup-time initialization of controls.)
82    
83    ----------------------------------------------------------------------
84    Name: Lal George
85    Date: 2002/05/15 09:20:10 EDT
86    Tag: george-20020515-pseudo-op-decls
87    Description:
88    
89            All pseudo-ops emitted before the first segment declaration
90    such as TEXT, DATA, and BSS directives are assumed to be global
91    declarations and are emitted first in the assembly file. This is
92    useful in a number of situations where one has pseudo-ops that are not
93    specific to any segment, and also works around the constraint that one
94    cannot have client pseudo-ops in the TEXT segment.
95    
96            Because no segment is associated with these declarations it is
97    an error to allocate any space or objects before the first segment
98    directive and an exception will be raised. However, we cannot make
99    this check for client pseudo-ops.
100    
101            These top level declarations are a field in the CFG graph_info.
102    In theory you can continue to add to this field after the CFG has been
103    built -- provided you know what you are doing;-)
104    
105    ----------------------------------------------------------------------
106    Name: Matthias Blume
107    Date: 2002/05/13 16:40:00 EDT
108    Tag: blume-20020513-pp-etc
109    Description:
110    
111    A few minor bugfixes:
112    
113      - Stopgap measure for bug recently reported by Elsa Gunter (ppDec).
114        (Bogus printouts for redefined bindings still occur.  Compiler
115        bug should no longer occur now.  We need to redo the prettyprinter
116        from scratch.)
117    
118      - CM pathname printer now also adds escape sequences for ( and )
119    
120      - commend and docu fixes for ml-nlffi
121    
122    ----------------------------------------------------------------------
123    Name: Matthias Blume
124    Date: 2002/05/10 16:40:00 EDT
125    Tag: blume-20020510-erg-textio
126    Description:
127    
128    Applied the following bugfix provided by Emden Gansner:
129    
130        Output is corrupted when outputSubstr is used rather than output.
131    
132        The problem occurs when a substring
133    
134            ss = (s, dataStart, dataLen)
135    
136        where dataStart > 0, fills a stream buffer with avail bytes left.
137        avail bytes of s, starting at index dataStart, are copied into the
138        buffer, the buffer is flushed, and then the remaining dataLen-avail
139        bytes of ss are copied into the beginning of the buffer. Instead of
140        starting this copy at index dataStart+avail in s, the current code
141        starts the copy at index avail.
142    
143        Fix:
144        In text-io-fn.sml, change line 695 from
145             val needsFlush = copyVec(v, avail, dataLen-avail, buf, 0)
146        to
147             val needsFlush = copyVec(v, dataStart+avail, dataLen-avail, buf, 0)
148    
149    ----------------------------------------------------------------------
150    Name: Matthias Blume
151    Date: 2002/04/12 13:55:00 EDT
152    Tag: blume-20020412-assyntax
153    Description:
154    
155    1. Grabbed newer assyntax.h from the XFree86 project.
156    2. Fiddled with how to compile X86.prim.asm without warnings.
157    3. (Very) Minor cleanup in CM.
158    
159    ----------------------------------------------------------------------
160    Name: Matthias Blume
161    Date: 2002/04/01 (no joke!) 17:07:00 EST
162    Tag: blume-20020401-x86div
163    Description:
164    
165    Added full support for div/mod/rem/quot on the x86, using the machine
166    instruction's two results (without clumsily recomputing the remainder)
167    directly where appropriate.
168    
169    Some more extensive power-of-two support was added to the x86 instruction
170    selector (avoiding expensive divs, mods, and muls where they can be
171    replaced with cheaper shifts and masks).  However, this sort of thing
172    ought to be done earlier, e.g., within the CPS optimizer so that
173    all architectures benefit from it.
174    
175    The compiler compiles to a fixed point, but changes might be somewhat
176    fragile nevertheless.  Please, report any strange things that you might
177    see wrt. div/mod/quot/rem...
178    
179    ----------------------------------------------------------------------
180    Name: Matthias Blume
181    Date: 2002/03/29 17:22:00
182    Tag: blume-20020329-div
183    Description:
184    
185    Fixed my broken div/mod logic.  Unfortunately, this means that the
186    inline code for div/mod now has one more comparison than before.
187    Fast paths (quotient > 0 or remainder = 0) are not affected, though.
188    The problem was with quotient = 0, because that alone does not tell
189    us which way the rounding went.  One then has to look at whether
190    remainder and divisor have the same sign...  :(
191    
192    Anyway, I replaced the bootfiles with fresh ones...
193    
194    ----------------------------------------------------------------------
195    Name: Matthias Blume
196    Date: 2002/03/29 14:10:00 EST
197    Tag: blume-20020329-inlprims
198    Description:
199    
200    NEW BOOTFILES!!!    Version number bumped to 110.39.3.
201    
202    Primops have changed. This means that the bin/boot-file formats have
203    changed as well.
204    
205    To make sure that there is no confusion, I made a new version.
206    
207    
208    CHANGES:
209    
210    * removed REMT from mltree (remainder should never overflow).
211    
212    * added primops to deal with divisions of all flavors to the frontend
213    
214    * handled these primops all the way through so they map to their respective
215      MLRISC support
216    
217    * used these primops in the implementation of Int, Int32, Word, Word32
218    
219    * removed INLDIV, INLMOD, and INLREM as they are no longer necessary
220    
221    * parameterized INLMIN, INLMAX, and INLABS by a numkind
222    
223    * translate.sml now deals with all flavors of INL{MIN,MAX,ABS}, including
224      floating point
225    
226    * used INL{MIN,MAX,ABS} in the implementation of Int, Int32, Word, Word32,
227      and Real (but Real.abs maps to a separate floating-point-only primop)
228    
229    
230    TODO items:
231    
232    * Hacked Alpha32 instruction selection, disabling the selection of REMx
233      instructions because the machine instruction encoder cannot handle
234      them.  (Hppa, PPC, and Sparc instruction selection did not handle
235      REM in the first place, and REM is supported by the x86 machine coder.)
236    
237    * Handle DIV and MOD with DIV_TO_NEGINF directly in the x86 instruction
238      selection phase.  (The two can be streamlined because the hardware
239      delivers both quotient and remainder at the same time anyway.)
240    
241    * Think about what to do with "valOf(Int32.minInt) div ~1" and friends.
242      (Currently the behavior is inconsistent both across architectures and
243      wrt. the draft Basis spec.)
244    
245    * Word8 should eventually be handled natively, too.
246    
247    * There seems to be one serious bug in mltree-gen.sml.  It appears, though,
248      as if there currently is no execution path that could trigger it in
249      SML/NJ.  (The assumptions underlying functions arith and promotable do not
250      hold for things like multiplication and division.)
251    
252    ----------------------------------------------------------------------
253    Name: Matthias Blume
254    Date: 2002/03/27 16:27:00 EST
255    Tag: blume-20020327-mlrisc-divisions
256    Description:
257    
258    Added support for all four division operations (ML's div, mod, quot,
259    and rem) to MLRISC.  In the course of doing so, I also rationalized
260    the naming (no more annoying switch-around of DIV and QUOT), by
261    parameterizing the operation by div_rounding_mode (which can be either
262    DIV_TO_ZERO or DIV_TO_NEGINF).
263    
264    The generic MLTreeGen functor takes care of compiling all four
265    operations down to only round-to-zero div.
266    
267    Missing pieces:
268    
269      * Doing something smarter than relying on MLTreeGen on architectures
270        like, e.g., the x86 where hardware division delivers both quotient and
271        remainder at the same time.  With this, the implementation of the
272        round-to-neginf operations could be further streamlined.
273    
274      * Remove inlining support for div/mod/rem from the frontend and replace it
275        with primops that get carried through to the backend.  Do this for all
276        int and word types.
277    
278    ----------------------------------------------------------------------
279    Name: Matthias Blume
280    Date: 2002/03/25 17:25:00 EST
281    Tag: blume-20020325-divmod
282    Description:
283    
284    I improved (hopefully without breaking them) the implementation of Int.div,
285    Int.mod, and Int.rem.   For this, the code in translate.sml now takes
286    advantage of the following observations:
287    
288      Let  q = x quot y      r = x rem y
289           d = x div  y      m = x mod y
290    
291    where "quot" is the round-to-zero version of integer division that
292    hardware usually provides.  Then we have:
293    
294         r = x - q * y        where neither the * nor the - will overflow
295         d = if q >= 0 orelse x = q * y then q else q - 1
296                              where neither the * nor the - will overflow
297         m = if q >= 0 orelse r = 0 then r else r + y
298                              where the + will not overflow
299    
300    This results in substantial simplification of the generated code.
301    The following table shows the number of CFG nodes and edges generated
302    for
303            fun f (x, y) = x OPER y
304            (* with OPER \in div, mod, quot, rem *)
305    
306    
307        OPER | nodes(old) | edges(old) | nodes(new) | edges(new)
308        --------------------------------------------------------
309         div |         24 |         39 |         12 |         16
310         mod |         41 |         71 |         12 |         16
311        quot |          8 |         10 |          8 |         10
312         rem |         10 |         14 |          8 |         10
313    
314    
315    ----------------------------------------------------------------------
316    Name: Matthias Blume
317    Date: 2002/03/25 22:06:00 EST
318    Tag: blume-20020325-cprotobug
319    Description:
320    
321    Fixed a bug in cproto (c prototype decoder).
322    
323    ----------------------------------------------------------------------
324    Name: Matthias Blume
325    Date: 2002/03/25 16:00:00 EST
326    Tag: blume-20020325-raw-primops
327    Description:
328    
329    I did some cleanup to Allen's new primop code and
330    replaced yesterday's bootfiles with new ones.
331    (But they are stored in the same place.)
332    
333    ----------------------------------------------------------------------
334    Name: Matthias Blume
335    Date: 2002/03/24 22:40:00 EST
336    Tag: blume-20020324-bootfiles
337    Description:
338    
339    Made the bootfiles that Allen asked for.
340    
341    ----------------------------------------------------------------------
342    Name: Allen Leung
343    Date: 2002/03/23 15:50:00 EST
344    Tag: leunga-20020323-flint-cps-rcc-primops
345    Description:
346    
347      1. Changes to FLINT primops:
348    
349        (* make a call to a C-function;
350         * The primop carries C function prototype information and specifies
351         * which of its (ML-) arguments are floating point. C prototype
352         * information is for use by the backend, ML information is for
353         * use by the CPS converter. *)
354      | RAW_CCALL of { c_proto: CTypes.c_proto,
355                       ml_args: ccall_type list,
356                       ml_res_opt: ccall_type option,
357                       reentrant : bool
358                     } option
359       (* Allocate uninitialized storage on the heap.
360        * The record is meant to hold short-lived C objects, i.e., they
361        * are not ML pointers.  With the tag, the representation is
362        * the same as RECORD with tag tag_raw32 (sz=4), or tag_fblock (sz=8)
363        *)
364      | RAW_RECORD of {tag:bool,sz:int}
365      and ccall_type = CCALL_INT32 | CCALL_REAL64 | CCALL_ML_PTR
366    
367      2.  These CPS primops are now overloaded:
368    
369           rawload of {kind:numkind}
370           rawstore of {kind:numkind}
371    
372          The one argument form is:
373    
374             rawload {kind} address
375    
376          The two argument form is:
377    
378             rawload {kind} [ml object, byte-offset]
379    
380      3. RAW_CCALL/RCC now takes two extra arguments:
381    
382         a. The first is whether the C call is reentrant, i.e., whether
383            ML state should be saved and restored.
384         b. The second argument is a string argument specifying the name of
385            library and the C function.
386    
387         These things are currently not handled in the code generator, yet.
388    
389      4. In CProto,
390    
391         An encoding type of "bool" means "ml object" and is mapped into
392         C prototype of PTR.  Note that "bool" is different than "string",
393         even though "string" is also mapped into PTR, because "bool"
394         is assigned an CPS type of BOGt, while "string" is assigned INT32t.
395    
396      5. Pickler/unpicker
397    
398         Changed to handle RAW_RECORD and newest RAW_CCALL
399    
400      6. MLRiscGen,
401    
402         1. Changed to handle the new rawload/rawstore/rawrecord operators.
403         2. Code for handling C Calls has been moved to a new module CPSCCalls,
404            in the file CodeGen/cpscompile/cps-c-calls.sml
405    
406      7. Added the conditional move operator
407    
408             condmove of branch
409    
410         to cps.  Generation of this is still buggy so it is currently
411         disabled.
412    
413    ----------------------------------------------------------------------
414    Name: Lal George
415    Date: 2002/03/22 14:18:25 EST
416    Tag: george-20020322-cps-branch-prob
417    Description:
418    
419    Implemented the Ball-Larus branch prediction-heuristics, and
420    incorporated graphical viewers for control flow graphs.
421    
422    Ball-Larus Heuristics:
423    ---------------------
424    See the file compiler/CodeGen/cpscompile/cpsBranchProb.sml.
425    
426    By design it uses the Dempster-Shafer theory for combining
427    probabilities.  For example, in the function:
428    
429        fun f(n,acc) = if n = 0 then acc else f(n-1, n*acc)
430    
431    the ball-larus heuristics predicts that the n=0 is unlikely
432    (OH-heuristic), and the 'then' branch is unlikely because of the
433    RH-heuristic -- giving the 'then' branch an even lower combined
434    probability using the Dempster-Shafer theory.
435    
436    Finally, John Reppy's loop analysis in MLRISC, further lowers the
437    probability of the 'then' branch because of the loop in the else
438    branch.
439    
440    
441    Graphical Viewing:
442    ------------------
443    I merely plugged in Allen's graphical viewers into the compiler. The
444    additional code is not much. At the top level, saying:
445    
446            Control.MLRISC.getFlag "cfg-graphical-view" := true;
447    
448    will display the graphical view of the control flow graph just before
449    back-patching.  daVinci must be in your path for this to work. If
450    daVinci is not available, then the default viewer can be changed
451    using:
452    
453            Control.MLRISC.getString "viewer"
454    
455    which can be set to "dot" or "vcg" for the corresponding viewers. Of
456    course, these viewers must be in your path.
457    
458    The above will display the compilation unit at the level of clusters,
459    many of which are small, boring, and un-interesting. Also setting:
460    
461            Control.MLRISC.getInt "cfg-graphical-view_size"
462    
463    will display clusters that are larger than the value set by the above.
464    
465    
466    ----------------------------------------------------------------------
467    Name: Matthias Blume
468    Date: 2002/03/21 22:20:00 EST
469    Tag: blume-20020321-kmp-bugfix
470    Description:
471    
472    Changed the interface to the KMP routine in PreString and fixed
473    a minor bug in one place where it was used.
474    
475    ----------------------------------------------------------------------
476    Name: Allen Leung
477    Date: 2002/03/21 20:30:00 EST
478    Tag: leunga-20020321-cfg
479    Description:
480    
481      Fixed a potential problem in cfg edge splitting.
482    
483    ----------------------------------------------------------------------
484    Name: Allen Leung
485    Date: 2002/03/21 17:15:00 EST
486    Tag: leunga-20020321-x86-fp-cfg
487    Description:
488    
489      1. Recoded the buggy parts of x86-fp.
490    
491         a. All the block reordering code has been removed.
492            We now depend on the block placement phases to do this work.
493    
494         b. Critical edge splitting code has been simplified and moved into the
495            CFG modules, as where they belong.
496    
497         Both of these were quite buggy and complex.  The code is now much, much
498         simpler.
499    
500      2. X86 backend.
501    
502         a. Added instructions for 64-bit support.  Instruction selection for
503            64-bit has not been committed, however, since that
504            requires changes to MLTREE which haven't been approved by
505            Lal and John.
506    
507         b. Added support for FUCOMI and FUCOMIP when generating code for
508            PentiumPro and above.  We only generate these instructions in
509            the fast-fp mode.
510    
511         c. Added cases for JP and JNP in X86FreqProps.
512    
513      3. CFG
514    
515         CFG now has a bunch of methods for edge splitting and merging.
516    
517      4. Machine description.
518    
519         John's simplification of MLTREE_BASIS.fcond broke a few machine
520         description things:
521    
522         rtl-build.{sig,sml} and hppa.mdl fixed.
523    
524         NOTE: the machine description stuff in the repository is still broken.
525               Again, I can't put my fixes in because that involves
526               changes to MLTREE.
527    
528    ----------------------------------------------------------------------
529    Name: Matthias Blume
530    Date: 2002/03/20 15:55:00 EST
531    Tag: blume-20020320-kmp
532    Description:
533    
534    Implemented Knuth-Morris-Pratt string matching in PreString and used
535    it for String.isSubstring, Substring.isSubstring, and
536    Substring.position.
537    
538    (Might need some stress-testing.  Simple examples worked fine.)
539    
540    ----------------------------------------------------------------------
541    Name: Matthias Blume
542    Date: 2002/03/19 16:37:00 EST
543    Tag: blume-20020319-witnesses
544    Description:
545    
546    Added a structure C.W and functions convert/Ptr.convert to ml-nlffi-lib.
547    
548    This implements a generic mechanism for changing constness qualifiers
549    anywhere within big C types without resorting to outright "casts".
550    (So far, functions such as C.rw/C.ro or C.Ptr.rw/C.Ptr.ro only let you
551    modify the constness at the outermost level.)
552    The implementation of "convert" is based on the idea of "witness"
553    values -- values that are not used by the operation but whose types
554    "testify" to their applicability.  On the implementation side, "convert"
555    is simply a projection (returning its second curried argument).  With
556    cross-module inlining, it should not result in any machine code being
557    generated.
558    
559    ----------------------------------------------------------------------
560    Name: Matthias Blume
561    Date: 2002/03/15 16:40:00 EST
562    Tag: blume-20020315-basis
563    Description:
564    
565    Provided (preliminary?) implementations for
566    
567      {String,Substring}.{concatWith,isSuffix,isSubstring}
568    
569    and
570    
571      Substring.full
572    
573    Those are in the Basis spec but they were missing in SML/NJ.
574    
575    ----------------------------------------------------------------------
576    Name: Matthias Blume
577    Date: 2002/03/14 21:30:00 EST
578    Tag: blume-20020314-controls
579    Description:
580    
581    Controls:
582    ---------
583    
584    1. Factored out the recently-added Controls : CONTROLS stuff and put
585       it into its own library $/controls-lib.cm.  The source tree for
586       this is under src/smlnj-lib/Controls.
587    
588    2. Changed the names of types and functions in this interface, so they
589       make a bit more "sense":
590    
591          module -> registry
592          'a registry -> 'a group
593    
594    3. The interface now deals in ref cells only.  The getter/setter interface
595       is (mostly) gone.
596    
597    4. Added a function that lets one register an already-existing ref cell.
598    
599    5. Made the corresponding modifications to the rest of the code so that
600       everything compiles again.
601    
602    6. Changed the implementation of Controls.MLRISC back to something closer
603       to the original.  In particular, this module (and therefore MLRISC)
604       does not depend on Controls.  There now is some link-time code in
605       int-sys.sml that registers the MLRISC controls with the Controls
606       module.
607    
608    CM:
609    ---
610    
611      * One can now specify the lambda-split aggressiveness in init.cmi.
612    
613    ----------------------------------------------------------------------
614    Name: Allen Leung
615    Date: 2002/03/13 17:30:00 EST
616    Tag: leunga-20020313-x86-fp-unary
617    Description:
618    
619    Bug fix for:
620    
621    > leunga@weaselbane:~/Yale/tmp/sml-dist{21} bin/sml
622    > Standard ML of New Jersey v110.39.1 [FLINT v1.5], March 08, 2002
623    > - fun f(x,(y,z)) = Real.~ y;
624    > [autoloading]
625    > [autoloading done]
626    >       fchsl   (%eax), 184(%esp)
627    > Error: MLRisc bug: X86MCEmitter.emitInstr
628    >
629    > uncaught exception Error
630    >   raised at: ../MLRISC/control/mlriscErrormsg.sml:16.14-16.19
631    
632    The problem was that the code generator did not generate any fp registers
633    in this case, and the ra didn't know that it needed to run the X86FP phase to
634    translate the pseudo fp instruction.   This only happened with unary fp
635    operators in certain situations.
636    
637    ----------------------------------------------------------------------
638    Name: Matthias Blume
639    Date: 2002/03/13 14:00:00 EST
640    Tag: blume-20020313-overload-etc
641    Description:
642    
643    1. Added _overload as a synonym for overload for backward compatibility.
644       (Control.overloadKW must be true for either version to be accepted.)
645    
646    2. Fixed bug in install script that caused more things to be installed
647       than what was requested in config/targets.
648    
649    3. Made CM aware of the (_)overload construct so that autoloading
650       works.
651    
652    ----------------------------------------------------------------------
653    Name: Matthias Blume
654    Date: 2002/03/12 22:03:00 EST
655    Tag: blume-20020312-url
656    Description:
657    
658    Forgot to update BOOT and srcarchiveurl.
659    
660    ----------------------------------------------------------------------
661    Name: Matthias Blume
662    Date: 2002/03/12 17:30:00 EST
663    Tag: blume-20020312-version110392
664    Description:
665    
666    Yet another version number bump (because of small changes to the
667    binfile format).  Version number is now 110.39.2.  NEW BOOTFILES!
668    
669    Changes:
670    
671      The new pid generation scheme described a few weeks ago was overly
672      complicated.  I implemented a new mechanism that is simpler and
673      provides a bit more "stability":  Once CM has seen a compilation
674      unit, it keeps its identity constant (as long as you do not delete
675      those crucial CM/GUID/* files).  This means that when you change
676      an interface, compile, then go back to the old interface, and
677      compile again, you arrive at the original pid.
678    
679      There now also is a mechanism that instructs CM to use the plain
680      environment hash as a module's pid (effectively making its GUID
681      the empty string).  For this, "noguid" must be specified as an
682      option to the .sml file in question within its .cm file.
683      This is most useful for code that is being generated by tools such
684      as ml-nlffigen (because during development programmers tend to
685      erase the tool's entire output directory tree including CM's cached
686      GUIDs).  "noguid" is somewhat dangerous (since it can be used to locally
687      revert to the old, broken behavior of SML/NJ, but in specific cases
688      where there is no danger of interface confusion, its use is ok
689      (I think).
690    
691      ml-nlffigen by default generates "noguid" annotations.  They can be
692      turned off by specifying -guid in its command line.
693    
694    ----------------------------------------------------------------------
695    Name: Lal George
696    Date: 2002/03/12 12 14:42:36 EST
697    Tag: george-20020312-frequency-computation
698    Description:
699    
700    Integrated jump chaining and static block frequency into the
701    compiler. More details and numbers later.
702    
703    ----------------------------------------------------------------------
704    Name: Lal George
705    Date: 2002/03/11 11 22:38:53 EST
706    Tag: george-20020311-jump-chain-elim
707    Description:
708    
709    Tested the jump chain elimination on all architectures (except the
710    hppa).  This is on by default right now and is profitable for the
711    alpha and x86, however, it may not be profitable for the sparc and ppc
712    when compiling the compiler.
713    
714    The gc test will typically jump to a label at the end of the cluster,
715    where there is another jump to an external cluster containing the actual
716    code to invoke gc. This is to allow factoring of common gc invocation
717    sequences. That is to say, we generate:
718    
719            f:
720               testgc
721               ja   L1      % jump if above to L1
722    
723            L1:
724               jmp L2
725    
726    
727    After jump chain elimination the 'ja L1' instructions is converted to
728    'ja L2'. On the sparc and ppc, many of the 'ja L2' instructions may end
729    up being implemented in their long form (if L2 is far away) using:
730    
731            jbe     L3      % jump if below or equal to L3
732            jmp     L2
733         L3:
734            ...
735    
736    
737    For large compilation units L2  may be far away.
738    
739    
740    ----------------------------------------------------------------------
741    Name: Matthias Blume
742    Date: 2002/03/11 13:30:00 EST
743    Tag: blume-20020311-mltreeeval
744    Description:
745    
746    A functor parameter was missing.
747    
748    ----------------------------------------------------------------------
749    Name: Allen Leung
750    Date: 2002/03/11 10:30:00 EST
751    Tag: leunga-20020311-runtime-string0
752    Description:
753    
754       The representation of the empty string now points to a
755    legal null terminated C string instead of unit.  It is now possible
756    to convert an ML string into C string with InlineT.CharVector.getData.
757    This compiles into one single machine instruction.
758    
759    ----------------------------------------------------------------------
760    Name: Allen Leung
761    Date: 2002/03/10 23:55:00 EST
762    Tag: leunga-20020310-x86-call
763    Description:
764    
765       Added machine generation for CALL instruction (relative displacement mode)
766    
767    ----------------------------------------------------------------------
768    Name: Matthias Blume
769    Date: 2002/03/08 16:05:00
770    Tag: blume-20020308-entrypoints
771    Description:
772    
773    Version number bumped to 110.39.1.  NEW BOOTFILES!
774    
775    Entrypoints: non-zero offset into a code object where execution should begin.
776    
777    - Added the notion of an entrypoint to CodeObj.
778    - Added reading/writing of entrypoint info to Binfile.
779    - Made runtime system bootloader aware of entrypoints.
780    - Use the address of the label of the first function given to mlriscGen
781      as the entrypoint.  This address is currently always 0, but it will
782      not be 0 once we turn on block placement.
783    - Removed the linkage cluster code (which was The Other Way(tm) of dealing
784      with entry points) from mlriscGen.
785    
786    ----------------------------------------------------------------------
787    Name: Allen Leung
788    Date: 2002/03/07 20:45:00 EST
789    Tag: leunga-20020307-x86-cmov
790    Description:
791    
792       Bug fixes for CMOVcc on x86.
793    
794       1. Added machine code generation for CMOVcc
795       2. CMOVcc is now generated in preference over SETcc on PentiumPro or above.
796       3. CMOVcc cannot have an immediate operand as argument.
797    
798    ----------------------------------------------------------------------
799    Name: Matthias Blume
800    Date: 2002/03/07 16:15:00 EST
801    Tag: blume-20020307-controls
802    Description:
803    
804    This is a very large but mostly boring patch which makes (almost)
805    every tuneable compiler knob (i.e., pretty much everything under
806    Control.* plus a few other things) configurable via both the command
807    line and environment variables in the style CM did its configuration
808    until now.
809    
810    Try starting sml with '-h' (or, if you are brave, '-H')
811    
812    To this end, I added a structure Controls : CONTROLS to smlnj-lib.cm which
813    implements the underlying generic mechanism.
814    
815    The interface to some of the existing such facilities has changed somewhat.
816    For example, the MLRiscControl module now provides mkFoo instead of getFoo.
817    (The getFoo interface is still there for backward-compatibility, but its
818    use is deprecated.)
819    
820    The ml-build script passes -Cxxx=yyy command-line arguments through so
821    that one can now twiddle the compiler settings when using this "batch"
822    compiler.
823    
824    TODO items:
825    
826    We should go through and throw out all controls that are no longer
827    connected to anything.  Moreover, we should go through and provide
828    meaningful (and correct!) documentation strings for those controls
829    that still are connected.
830    
831    Currently, multiple calls to Controls.new are accepted (only the first
832    has any effect).  Eventually we should make sure that every control
833    is being made (via Controls.new) exactly once.  Future access can then
834    be done using Controls.acc.
835    
836    Finally, it would probably be a good idea to use the getter-setter
837    interface to controls rather than ref cells.  For the time being, both
838    styles are provided by the Controls module, but getter-setter pairs are
839    better if thread-safety is of any concern because they can be wrapped.
840    
841    *****************************************
842    
843    One bug fix: The function blockPlacement in three of the MLRISC
844    backpatch files used to be hard-wired to one of two possibilities at
845    link time (according to the value of the placementFlag).  But (I
846    think) it should rather sense the flag every time.
847    
848    *****************************************
849    
850    Other assorted changes (by other people who did not supply a HISTORY entry):
851    
852    1. the cross-module inliner now works much better (Monnier)
853    2. representation of weights, frequencies, and probabilities in MLRISC
854       changed in preparation of using those for weighted block placement
855       (Reppy, George)
856    
857    ----------------------------------------------------------------------
858    Name: Lal George
859    Date: 2002/03/07 14:44:24 EST 2002
860    Tag: george-20020307-weighted-block-placement
861    
862    Tested the weighted block placement optimization on all architectures
863    (except the hppa) using AMPL to generate the block and edge frequencies.
864    Changes were required in the machine properties to correctly
865    categorize trap instructions. There is an MLRISC flag
866    "weighted-block-placement" that can be used to enable weighted block
867    placement, but this will be ineffective without block/edge
868    frequencies (coming soon).
869    
870    
871    ----------------------------------------------------------------------
872    Name: Lal George
873    Date: 2002/03/05 17:24:48 EST
874    Tag: george-20020305-linkage-cluster
875    
876    In order to support the block placement optimization, a new cluster
877    is generated as the very first cluster (called the linkage cluster).
878    It contains a single jump to the 'real' entry point for the compilation
879    unit. Block placement has no effect on the linkage cluster itself, but
880    all the other clusters  have full freedom in the manner in which they
881    reorder blocks or functions.
882    
883    On the x86 the typical linkage code that is generated is:
884       ----------------------
885            .align 2
886       L0:
887            addl    $L1-L0, 72(%esp)
888            jmp     L1
889    
890    
891            .align  2
892       L1:
893       ----------------------
894    
895    72(%esp) is the memory location for the stdlink register. This
896    must contain the address of the CPS function being called. In the
897    above example, it contains the address of  L0; before
898    calling L1 (the real entry point for the compilation unit), it
899    must contain the address for L1, and hence
900    
901            addl $L1-L0, 72(%esp)
902    
903    I have tested this on all architectures except the hppa.The increase
904    in code size is of course negligible
905    
906    ----------------------------------------------------------------------
907    Name: Allen Leung
908    Date: 2002/03/03 13:20:00 EST
909    Tag: leunga-20020303-mlrisc-tools
910    
911      Added #[ ... ] expressions to mlrisc tools
912    
913    ----------------------------------------------------------------------
914    Name: Matthias Blume
915  Date: 2002/02/27 12:29:00 EST  Date: 2002/02/27 12:29:00 EST
916  Tag: blume-20020227-cdebug  Tag: blume-20020227-cdebug
917  Description:  Description:
# Line 134  Line 1032 
1032         TOTAL                                   2375.26u  57.21s  48.00g         TOTAL                                   2375.26u  57.21s  48.00g
1033    
1034  As you can see the old cps spill module suffers from some serious  As you can see the old cps spill module suffers from some serious
1035  performance problem but since I cannot decipher the old code fully,  performance problem.  But since I cannot decipher the old code fully,
1036  innstead of patching the problems up, I'm reimplementing it  instead of patching the problems up, I'm reimplementing it
1037  with a different algorithm.  The new code is more modular,  with a different algorithm.  The new code is more modular,
1038  smaller when compiled, and substantially faster  smaller when compiled, and substantially faster
1039  (O(n log n) time and O(n) space).  Timing of the new spill module:  (O(n log n) time and O(n) space).  Timing of the new spill module:

Legend:
Removed from v.1098  
changed lines
  Added in v.1211

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