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 773, Mon Jan 8 16:18:37 2001 UTC revision 1178, Mon Mar 25 20:51:48 2002 UTC
# Line 8  Line 8 
8  The form of an entry should be:  The form of an entry should be:
9    
10  Name:  Name:
11  Date:  Date: yyyy/mm/dd
12  Tag: <post-commit CVS tag>  Tag: <post-commit CVS tag>
13  Description:  Description:
14    
15    ----------------------------------------------------------------------
16    Name: Matthias Blume
17    Date: 2002/03/25 16:00:00 EST
18    Tag: blume-20020325-raw-primops
19    Description:
20    
21    I did some cleanup to Allen's new primop code and
22    replaced yesterday's bootfiles with new ones.
23    (But they are stored in the same place.)
24    
25    ----------------------------------------------------------------------
26    Name: Matthias Blume
27    Date: 2002/03/24 22:40:00 EST
28    Tag: blume-20020324-bootfiles
29    Description:
30    
31    Made the bootfiles that Allen asked for.
32    
33    ----------------------------------------------------------------------
34    Name: Allen Leung
35    Date: 2002/03/23 15:50:00 EST
36    Tag: leunga-20020323-flint-cps-rcc-primops
37    Description:
38    
39      1. Changes to FLINT primops:
40    
41        (* make a call to a C-function;
42         * The primop carries C function prototype information and specifies
43         * which of its (ML-) arguments are floating point. C prototype
44         * information is for use by the backend, ML information is for
45         * use by the CPS converter. *)
46      | RAW_CCALL of { c_proto: CTypes.c_proto,
47                       ml_args: ccall_type list,
48                       ml_res_opt: ccall_type option,
49                       reentrant : bool
50                     } option
51       (* Allocate uninitialized storage on the heap.
52        * The record is meant to hold short-lived C objects, i.e., they
53        * are not ML pointers.  With the tag, the representation is
54        * the same as RECORD with tag tag_raw32 (sz=4), or tag_fblock (sz=8)
55        *)
56      | RAW_RECORD of {tag:bool,sz:int}
57      and ccall_type = CCALL_INT32 | CCALL_REAL64 | CCALL_ML_PTR
58    
59      2.  These CPS primops are now overloaded:
60    
61           rawload of {kind:numkind}
62           rawstore of {kind:numkind}
63    
64          The one argument form is:
65    
66             rawload {kind} address
67    
68          The two argument form is:
69    
70             rawload {kind} [ml object, byte-offset]
71    
72      3. RAW_CCALL/RCC now takes two extra arguments:
73    
74         a. The first is whether the C call is reentrant, i.e., whether
75            ML state should be saved and restored.
76         b. The second argument is a string argument specifying the name of
77            library and the C function.
78    
79         These things are currently not handled in the code generator, yet.
80    
81      4. In CProto,
82    
83         An encoding type of "bool" means "ml object" and is mapped into
84         C prototype of PTR.  Note that "bool" is different than "string",
85         even though "string" is also mapped into PTR, because "bool"
86         is assigned an CPS type of BOGt, while "string" is assigned INT32t.
87    
88      5. Pickler/unpicker
89    
90         Changed to handle RAW_RECORD and newest RAW_CCALL
91    
92      6. MLRiscGen,
93    
94         1. Changed to handle the new rawload/rawstore/rawrecord operators.
95         2. Code for handling C Calls has been moved to a new module CPSCCalls,
96            in the file CodeGen/cpscompile/cps-c-calls.sml
97    
98      7. Added the conditional move operator
99    
100             condmove of branch
101    
102         to cps.  Generation of this is still buggy so it is currently
103         disabled.
104    
105    ----------------------------------------------------------------------
106    Name: Lal George
107    Date: 2002/03/22 14:18:25 EST
108    Tag: george-20020322-cps-branch-prob
109    Description:
110    
111    Implemented the Ball-Larus branch prediction-heuristics, and
112    incorporated graphical viewers for control flow graphs.
113    
114    Ball-Larus Heuristics:
115    ---------------------
116    See the file compiler/CodeGen/cpscompile/cpsBranchProb.sml.
117    
118    By design it uses the Dempster-Shafer theory for combining
119    probabilities.  For example, in the function:
120    
121        fun f(n,acc) = if n = 0 then acc else f(n-1, n*acc)
122    
123    the ball-larus heuristics predicts that the n=0 is unlikely
124    (OH-heuristic), and the 'then' branch is unlikely because of the
125    RH-heuristic -- giving the 'then' branch an even lower combined
126    probability using the Dempster-Shafer theory.
127    
128    Finally, John Reppy's loop analysis in MLRISC, further lowers the
129    probability of the 'then' branch because of the loop in the else
130    branch.
131    
132    
133    Graphical Viewing:
134    ------------------
135    I merely plugged in Allen's graphical viewers into the compiler. The
136    additional code is not much. At the top level, saying:
137    
138            Control.MLRISC.getFlag "cfg-graphical-view" := true;
139    
140    will display the graphical view of the control flow graph just before
141    back-patching.  daVinci must be in your path for this to work. If
142    daVinci is not available, then the default viewer can be changed
143    using:
144    
145            Control.MLRISC.getString "viewer"
146    
147    which can be set to "dot" or "vcg" for the corresponding viewers. Of
148    course, these viewers must be in your path.
149    
150    The above will display the compilation unit at the level of clusters,
151    many of which are small, boring, and un-interesting. Also setting:
152    
153            Control.MLRISC.getInt "cfg-graphical-view_size"
154    
155    will display clusters that are larger than the value set by the above.
156    
157    
158    ----------------------------------------------------------------------
159    Name: Matthias Blume
160    Date: 2002/03/21 22:20:00 EST
161    Tag: blume-20020321-kmp-bugfix
162    Description:
163    
164    Changed the interface to the KMP routine in PreString and fixed
165    a minor bug in one place where it was used.
166    
167    ----------------------------------------------------------------------
168    Name: Allen Leung
169    Date: 2002/03/21 20:30:00 EST
170    Tag: leunga-20020321-cfg
171    Description:
172    
173      Fixed a potential problem in cfg edge splitting.
174    
175    ----------------------------------------------------------------------
176    Name: Allen Leung
177    Date: 2002/03/21 17:15:00 EST
178    Tag: leunga-20020321-x86-fp-cfg
179    Description:
180    
181      1. Recoded the buggy parts of x86-fp.
182    
183         a. All the block reordering code has been removed.
184            We now depend on the block placement phases to do this work.
185    
186         b. Critical edge splitting code has been simplified and moved into the
187            CFG modules, as where they belong.
188    
189         Both of these were quite buggy and complex.  The code is now much, much
190         simpler.
191    
192      2. X86 backend.
193    
194         a. Added instructions for 64-bit support.  Instruction selection for
195            64-bit has not been committed, however, since that
196            requires changes to MLTREE which haven't been approved by
197            Lal and John.
198    
199         b. Added support for FUCOMI and FUCOMIP when generating code for
200            PentiumPro and above.  We only generate these instructions in
201            the fast-fp mode.
202    
203         c. Added cases for JP and JNP in X86FreqProps.
204    
205      3. CFG
206    
207         CFG now has a bunch of methods for edge splitting and merging.
208    
209      4. Machine description.
210    
211         John's simplification of MLTREE_BASIS.fcond broke a few machine
212         description things:
213    
214         rtl-build.{sig,sml} and hppa.mdl fixed.
215    
216         NOTE: the machine description stuff in the repository is still broken.
217               Again, I can't put my fixes in because that involves
218               changes to MLTREE.
219    
220    ----------------------------------------------------------------------
221    Name: Matthias Blume
222    Date: 2002/03/20 15:55:00 EST
223    Tag: blume-20020320-kmp
224    Description:
225    
226    Implemented Knuth-Morris-Pratt string matching in PreString and used
227    it for String.isSubstring, Substring.isSubstring, and
228    Substring.position.
229    
230    (Might need some stress-testing.  Simple examples worked fine.)
231    
232    ----------------------------------------------------------------------
233    Name: Matthias Blume
234    Date: 2002/03/19 16:37:00 EST
235    Tag: blume-20020319-witnesses
236    Description:
237    
238    Added a structure C.W and functions convert/Ptr.convert to ml-nlffi-lib.
239    
240    This implements a generic mechanism for changing constness qualifiers
241    anywhere within big C types without resorting to outright "casts".
242    (So far, functions such as C.rw/C.ro or C.Ptr.rw/C.Ptr.ro only let you
243    modify the constness at the outermost level.)
244    The implementation of "convert" is based on the idea of "witness"
245    values -- values that are not used by the operation but whose types
246    "testify" to their applicability.  On the implementation side, "convert"
247    is simply a projection (returning its second curried argument).  With
248    cross-module inlining, it should not result in any machine code being
249    generated.
250    
251    ----------------------------------------------------------------------
252    Name: Matthias Blume
253    Date: 2002/03/15 16:40:00 EST
254    Tag: blume-20020315-basis
255    Description:
256    
257    Provided (preliminary?) implementations for
258    
259      {String,Substring}.{concatWith,isSuffix,isSubstring}
260    
261    and
262    
263      Substring.full
264    
265    Those are in the Basis spec but they were missing in SML/NJ.
266    
267    ----------------------------------------------------------------------
268    Name: Matthias Blume
269    Date: 2002/03/14 21:30:00 EST
270    Tag: blume-20020314-controls
271    Description:
272    
273    Controls:
274    ---------
275    
276    1. Factored out the recently-added Controls : CONTROLS stuff and put
277       it into its own library $/controls-lib.cm.  The source tree for
278       this is under src/smlnj-lib/Controls.
279    
280    2. Changed the names of types and functions in this interface, so they
281       make a bit more "sense":
282    
283          module -> registry
284          'a registry -> 'a group
285    
286    3. The interface now deals in ref cells only.  The getter/setter interface
287       is (mostly) gone.
288    
289    4. Added a function that lets one register an already-existing ref cell.
290    
291    5. Made the corresponding modifications to the rest of the code so that
292       everything compiles again.
293    
294    6. Changed the implementation of Controls.MLRISC back to something closer
295       to the original.  In particular, this module (and therefore MLRISC)
296       does not depend on Controls.  There now is some link-time code in
297       int-sys.sml that registers the MLRISC controls with the Controls
298       module.
299    
300    CM:
301    ---
302    
303      * One can now specify the lambda-split aggressiveness in init.cmi.
304    
305    ----------------------------------------------------------------------
306    Name: Allen Leung
307    Date: 2002/03/13 17:30:00 EST
308    Tag: leunga-20020313-x86-fp-unary
309    Description:
310    
311    Bug fix for:
312    
313    > leunga@weaselbane:~/Yale/tmp/sml-dist{21} bin/sml
314    > Standard ML of New Jersey v110.39.1 [FLINT v1.5], March 08, 2002
315    > - fun f(x,(y,z)) = Real.~ y;
316    > [autoloading]
317    > [autoloading done]
318    >       fchsl   (%eax), 184(%esp)
319    > Error: MLRisc bug: X86MCEmitter.emitInstr
320    >
321    > uncaught exception Error
322    >   raised at: ../MLRISC/control/mlriscErrormsg.sml:16.14-16.19
323    
324    The problem was that the code generator did not generate any fp registers
325    in this case, and the ra didn't know that it needed to run the X86FP phase to
326    translate the pseudo fp instruction.   This only happened with unary fp
327    operators in certain situations.
328    
329    ----------------------------------------------------------------------
330    Name: Matthias Blume
331    Date: 2002/03/13 14:00:00 EST
332    Tag: blume-20020313-overload-etc
333    Description:
334    
335    1. Added _overload as a synonym for overload for backward compatibility.
336       (Control.overloadKW must be true for either version to be accepted.)
337    
338    2. Fixed bug in install script that caused more things to be installed
339       than what was requested in config/targets.
340    
341    3. Made CM aware of the (_)overload construct so that autoloading
342       works.
343    
344    ----------------------------------------------------------------------
345    Name: Matthias Blume
346    Date: 2002/03/12 22:03:00 EST
347    Tag: blume-20020312-url
348    Description:
349    
350    Forgot to update BOOT and srcarchiveurl.
351    
352    ----------------------------------------------------------------------
353    Name: Matthias Blume
354    Date: 2002/03/12 17:30:00 EST
355    Tag: blume-20020312-version110392
356    Description:
357    
358    Yet another version number bump (because of small changes to the
359    binfile format).  Version number is now 110.39.2.  NEW BOOTFILES!
360    
361    Changes:
362    
363      The new pid generation scheme described a few weeks ago was overly
364      complicated.  I implemented a new mechanism that is simpler and
365      provides a bit more "stability":  Once CM has seen a compilation
366      unit, it keeps its identity constant (as long as you do not delete
367      those crucial CM/GUID/* files).  This means that when you change
368      an interface, compile, then go back to the old interface, and
369      compile again, you arrive at the original pid.
370    
371      There now also is a mechanism that instructs CM to use the plain
372      environment hash as a module's pid (effectively making its GUID
373      the empty string).  For this, "noguid" must be specified as an
374      option to the .sml file in question within its .cm file.
375      This is most useful for code that is being generated by tools such
376      as ml-nlffigen (because during development programmers tend to
377      erase the tool's entire output directory tree including CM's cached
378      GUIDs).  "noguid" is somewhat dangerous (since it can be used to locally
379      revert to the old, broken behavior of SML/NJ, but in specific cases
380      where there is no danger of interface confusion, its use is ok
381      (I think).
382    
383      ml-nlffigen by default generates "noguid" annotations.  They can be
384      turned off by specifying -guid in its command line.
385    
386    ----------------------------------------------------------------------
387    Name: Lal George
388    Date: 2002/03/12 12 14:42:36 EST
389    Tag: george-20020312-frequency-computation
390    Description:
391    
392    Integrated jump chaining and static block frequency into the
393    compiler. More details and numbers later.
394    
395    ----------------------------------------------------------------------
396    Name: Lal George
397    Date: 2002/03/11 11 22:38:53 EST
398    Tag: george-20020311-jump-chain-elim
399    Description:
400    
401    Tested the jump chain elimination on all architectures (except the
402    hppa).  This is on by default right now and is profitable for the
403    alpha and x86, however, it may not be profitable for the sparc and ppc
404    when compiling the compiler.
405    
406    The gc test will typically jump to a label at the end of the cluster,
407    where there is another jump to an external cluster containing the actual
408    code to invoke gc. This is to allow factoring of common gc invocation
409    sequences. That is to say, we generate:
410    
411            f:
412               testgc
413               ja   L1      % jump if above to L1
414    
415            L1:
416               jmp L2
417    
418    
419    After jump chain elimination the 'ja L1' instructions is converted to
420    'ja L2'. On the sparc and ppc, many of the 'ja L2' instructions may end
421    up being implemented in their long form (if L2 is far away) using:
422    
423            jbe     L3      % jump if below or equal to L3
424            jmp     L2
425         L3:
426            ...
427    
428    
429    For large compilation units L2  may be far away.
430    
431    
432    ----------------------------------------------------------------------
433    Name: Matthias Blume
434    Date: 2002/03/11 13:30:00 EST
435    Tag: blume-20020311-mltreeeval
436    Description:
437    
438    A functor parameter was missing.
439    
440    ----------------------------------------------------------------------
441    Name: Allen Leung
442    Date: 2002/03/11 10:30:00 EST
443    Tag: leunga-20020311-runtime-string0
444    Description:
445    
446       The representation of the empty string now points to a
447    legal null terminated C string instead of unit.  It is now possible
448    to convert an ML string into C string with InlineT.CharVector.getData.
449    This compiles into one single machine instruction.
450    
451    ----------------------------------------------------------------------
452    Name: Allen Leung
453    Date: 2002/03/10 23:55:00 EST
454    Tag: leunga-20020310-x86-call
455    Description:
456    
457       Added machine generation for CALL instruction (relative displacement mode)
458    
459    ----------------------------------------------------------------------
460    Name: Matthias Blume
461    Date: 2002/03/08 16:05:00
462    Tag: blume-20020308-entrypoints
463    Description:
464    
465    Version number bumped to 110.39.1.  NEW BOOTFILES!
466    
467    Entrypoints: non-zero offset into a code object where execution should begin.
468    
469    - Added the notion of an entrypoint to CodeObj.
470    - Added reading/writing of entrypoint info to Binfile.
471    - Made runtime system bootloader aware of entrypoints.
472    - Use the address of the label of the first function given to mlriscGen
473      as the entrypoint.  This address is currently always 0, but it will
474      not be 0 once we turn on block placement.
475    - Removed the linkage cluster code (which was The Other Way(tm) of dealing
476      with entry points) from mlriscGen.
477    
478    ----------------------------------------------------------------------
479    Name: Allen Leung
480    Date: 2002/03/07 20:45:00 EST
481    Tag: leunga-20020307-x86-cmov
482    Description:
483    
484       Bug fixes for CMOVcc on x86.
485    
486       1. Added machine code generation for CMOVcc
487       2. CMOVcc is now generated in preference over SETcc on PentiumPro or above.
488       3. CMOVcc cannot have an immediate operand as argument.
489    
490    ----------------------------------------------------------------------
491    Name: Matthias Blume
492    Date: 2002/03/07 16:15:00 EST
493    Tag: blume-20020307-controls
494    Description:
495    
496    This is a very large but mostly boring patch which makes (almost)
497    every tuneable compiler knob (i.e., pretty much everything under
498    Control.* plus a few other things) configurable via both the command
499    line and environment variables in the style CM did its configuration
500    until now.
501    
502    Try starting sml with '-h' (or, if you are brave, '-H')
503    
504    To this end, I added a structure Controls : CONTROLS to smlnj-lib.cm which
505    implements the underlying generic mechanism.
506    
507    The interface to some of the existing such facilities has changed somewhat.
508    For example, the MLRiscControl module now provides mkFoo instead of getFoo.
509    (The getFoo interface is still there for backward-compatibility, but its
510    use is deprecated.)
511    
512    The ml-build script passes -Cxxx=yyy command-line arguments through so
513    that one can now twiddle the compiler settings when using this "batch"
514    compiler.
515    
516    TODO items:
517    
518    We should go through and throw out all controls that are no longer
519    connected to anything.  Moreover, we should go through and provide
520    meaningful (and correct!) documentation strings for those controls
521    that still are connected.
522    
523    Currently, multiple calls to Controls.new are accepted (only the first
524    has any effect).  Eventually we should make sure that every control
525    is being made (via Controls.new) exactly once.  Future access can then
526    be done using Controls.acc.
527    
528    Finally, it would probably be a good idea to use the getter-setter
529    interface to controls rather than ref cells.  For the time being, both
530    styles are provided by the Controls module, but getter-setter pairs are
531    better if thread-safety is of any concern because they can be wrapped.
532    
533    *****************************************
534    
535    One bug fix: The function blockPlacement in three of the MLRISC
536    backpatch files used to be hard-wired to one of two possibilities at
537    link time (according to the value of the placementFlag).  But (I
538    think) it should rather sense the flag every time.
539    
540    *****************************************
541    
542    Other assorted changes (by other people who did not supply a HISTORY entry):
543    
544    1. the cross-module inliner now works much better (Monnier)
545    2. representation of weights, frequencies, and probabilities in MLRISC
546       changed in preparation of using those for weighted block placement
547       (Reppy, George)
548    
549    ----------------------------------------------------------------------
550    Name: Lal George
551    Date: 2002/03/07 14:44:24 EST 2002
552    Tag: george-20020307-weighted-block-placement
553    
554    Tested the weighted block placement optimization on all architectures
555    (except the hppa) using AMPL to generate the block and edge frequencies.
556    Changes were required in the machine properties to correctly
557    categorize trap instructions. There is an MLRISC flag
558    "weighted-block-placement" that can be used to enable weighted block
559    placement, but this will be ineffective without block/edge
560    frequencies (coming soon).
561    
562    
563    ----------------------------------------------------------------------
564    Name: Lal George
565    Date: 2002/03/05 17:24:48 EST
566    Tag: george-20020305-linkage-cluster
567    
568    In order to support the block placement optimization, a new cluster
569    is generated as the very first cluster (called the linkage cluster).
570    It contains a single jump to the 'real' entry point for the compilation
571    unit. Block placement has no effect on the linkage cluster itself, but
572    all the other clusters  have full freedom in the manner in which they
573    reorder blocks or functions.
574    
575    On the x86 the typical linkage code that is generated is:
576       ----------------------
577            .align 2
578       L0:
579            addl    $L1-L0, 72(%esp)
580            jmp     L1
581    
582    
583            .align  2
584       L1:
585       ----------------------
586    
587    72(%esp) is the memory location for the stdlink register. This
588    must contain the address of the CPS function being called. In the
589    above example, it contains the address of  L0; before
590    calling L1 (the real entry point for the compilation unit), it
591    must contain the address for L1, and hence
592    
593            addl $L1-L0, 72(%esp)
594    
595    I have tested this on all architectures except the hppa.The increase
596    in code size is of course negligible
597    
598    ----------------------------------------------------------------------
599    Name: Allen Leung
600    Date: 2002/03/03 13:20:00 EST
601    Tag: leunga-20020303-mlrisc-tools
602    
603      Added #[ ... ] expressions to mlrisc tools
604    
605    ----------------------------------------------------------------------
606    Name: Matthias Blume
607    Date: 2002/02/27 12:29:00 EST
608    Tag: blume-20020227-cdebug
609    Description:
610    
611    - made types in structure C and C_Debug to be equal
612    - got rid of code duplication (c-int.sml vs. c-int-debug.sml)
613    - there no longer is a C_Int_Debug (C_Debug is directly derived from C)
614    
615    ----------------------------------------------------------------------
616    Name: Matthias Blume
617    Date: 2002/02/26 12:00:00 EST
618    Tag: blume-20020226-ffi
619    Description:
620    
621    1. Fixed a minor bug in CM's "noweb" tool:
622       If numbering is turned off, then truly don't number (i.e., do not
623       supply the -L option to noweb).  The previous behavior was to supply
624       -L'' -- which caused noweb to use the "default" line numbering scheme.
625       Thanks to Chris Richards for pointing this out (and supplying the fix).
626    
627    2. Once again, I reworked some aspects of the FFI:
628    
629       A. The incomplete/complete type business:
630    
631       - Signatures POINTER_TO_INCOMPLETE_TYPE and accompanying functors are
632         gone!
633       - ML types representing an incomplete type are now *equal* to
634         ML types representing their corresponding complete types (just like
635         in C).  This is still safe because ml-nlffigen will not generate
636         RTTI for incomplete types, nor will it generate functions that
637         require access to such RTTI.   But when ML code generated from both
638         incomplete and complete versions of the C type meet, the ML types
639         are trivially interoperable.
640    
641         NOTE:  These changes restore the full generality of the translation
642         (which was previously lost when I eliminated functorization)!
643    
644       B. Enum types:
645    
646       - Structure C now has a type constructor "enum" that is similar to
647         how the "su" constructor works.  However, "enum" is not a phantom
648         type because each "T enum" has values (and is isomorphic to
649         MLRep.Signed.int).
650       - There are generic access operations for enum objects (using
651         MLRep.Signed.int).
652       - ml-nlffigen will generate a structure E_foo for each "enum foo".
653         * The structure contains the definition of type "mlrep" (the ML-side
654         representation type of the enum).  Normally, mlrep is the same
655         as "MLRep.Signed.int", but if ml-nlffigen was invoked with "-ec",
656         then mlrep will be defined as a datatype -- thus facilitating
657         pattern matching on mlrep values.
658         ("-ec" will be suppressed if there are duplicate values in an
659          enumeration.)
660         * Constructors ("-ec") or values (no "-ec") e_xxx of type mlrep
661         will be generated for each C enum constant xxx.
662         * Conversion functions m2i and i2m convert between mlrep and
663         MLRep.Signed.int.  (Without "-ec", these functions are identities.)
664         * Coversion functions c and ml convert between mlrep and "tag enum".
665         * Access functions (get/set) fetch and store mlrep values.
666       - By default (unless ml-nlffigen was invoked with "-nocollect"), unnamed
667         enumerations are merged into one single enumeration represented by
668         structure E_'.
669    
670    ----------------------------------------------------------------------
671    Name: Allen Leung
672    Date: 2002/02/25 04:45:00 EST
673    Tag: leunga-20020225-cps-spill
674    
675    This is a new implementation of the CPS spill phase.
676    The new phase is in the new file compiler/CodeGen/cpscompile/spill-new.sml
677    In case of problems, replace it with the old file spill.sml
678    
679    The current compiler runs into some serious performance problems when
680    constructing a large record.  This can happen when we try to compile a
681    structure with many items.  Even a very simple structure like the following
682    makes the compiler slow down.
683    
684        structure Foo = struct
685           val x_1 = 0w1 : Word32.int
686           val x_2 = 0w2 : Word32.int
687           val x_3 = 0w3 : Word32.int
688           ...
689           val x_N = 0wN : Word32.int
690        end
691    
692    The following table shows the compile time, from N=1000 to N=4000,
693    with the old compiler:
694    
695    N
696    1000   CPS 100 spill                           0.04u  0.00s  0.00g
697           MLRISC ra                               0.06u  0.00s  0.05g
698              (spills = 0 reloads = 0)
699           TOTAL                                   0.63u  0.07s  0.21g
700    
701    1100   CPS 100 spill                           8.25u  0.32s  0.64g
702           MLRISC ra                               5.68u  0.59s  3.93g
703              (spills = 0 reloads = 0)
704           TOTAL                                   14.71u  0.99s  4.81g
705    
706    1500   CPS 100 spill                           58.55u  2.34s  1.74g
707           MLRISC ra                               5.54u  0.65s  3.91g
708              (spills = 543 reloads = 1082)
709           TOTAL                                   65.40u  3.13s  6.00g
710    
711    2000   CPS 100 spill                           126.69u  4.84s  3.08g
712           MLRISC ra                               0.80u  0.10s  0.55g
713              (spills = 42 reloads = 84)
714           TOTAL                                   129.42u  5.10s  4.13g
715    
716    3000   CPS 100 spill                           675.59u  19.03s  11.64g
717           MLRISC ra                               2.69u  0.27s  1.38g
718              (spills = 62 reloads = 124)
719           TOTAL                                   682.48u  19.61s  13.99g
720    
721    4000   CPS 100 spill                           2362.82u  56.28s  43.60g
722           MLRISC ra                               4.96u  0.27s  2.72g
723              (spills = 85 reloads = 170)
724           TOTAL                                   2375.26u  57.21s  48.00g
725    
726    As you can see the old cps spill module suffers from some serious
727    performance problem.  But since I cannot decipher the old code fully,
728    instead of patching the problems up, I'm reimplementing it
729    with a different algorithm.  The new code is more modular,
730    smaller when compiled, and substantially faster
731    (O(n log n) time and O(n) space).  Timing of the new spill module:
732    
733    4000  CPS 100 spill                           0.02u  0.00s  0.00g
734          MLRISC ra                               0.25u  0.02s  0.15g
735             (spills=1 reloads=3)
736          TOTAL                                   7.74u  0.34s  1.62g
737    
738    Implementation details:
739    
740    As far as I can tell, the purpose of the CPS spill module is to make sure the
741    number of live variables at any program point (the bandwidth)
742    does not exceed a certain limit, which is determined by the
743    size of the spill area.
744    
745    When the bandwidth is too large, we decrease the register pressure by
746    packing live variables into spill records.  How we achieve this is
747    completely different than what we did in the old code.
748    
749    First, there is something about the MLRiscGen code generator
750    that we should be aware of:
751    
752    o MLRiscGen performs code motion!
753    
754       In particular, it will move floating point computations and
755       address computations involving only the heap pointer to
756       their use sites (if there is only a single use).
757       What this means is that if we have a CPS record construction
758       statement
759    
760           RECORD(k,vl,w,e)
761    
762       we should never count the new record address w as live if w
763       has only one use (which is often the case).
764    
765       We should do something similar to floating point, but the transformation
766       there is much more complex, so I won't deal with that.
767    
768    Secondly, there are now two new cps primops at our disposal:
769    
770     1. rawrecord of record_kind option
771        This pure operator allocates some uninitialized storage from the heap.
772        There are two forms:
773    
774         rawrecord NONE [INT n]  allocates a tagless record of length n
775         rawrecord (SOME rk) [INT n] allocates a tagged record of length n
776                                     and initializes the tag.
777    
778     2. rawupdate of cty
779          rawupdate cty (v,i,x)
780          Assigns to x to the ith component of record v.
781          The storelist is not updated.
782    
783    We use these new primops for both spilling and increment record construction.
784    
785     1. Spilling.
786    
787        This is implemented with a linear scan algorithm (but generalized
788        to trees).  The algorithm will create a single spill record at the
789        beginning of the cps function and use rawupdate to spill to it,
790        and SELECT or SELp to reload from it.  So both spills and reloads
791        are fine-grain operations.  In contrast, in the old algorithm
792        "spills" have to be bundled together in records.
793    
794        Ideally, we should sink the spill record construction to where
795        it is needed.  We can even split the spill record into multiple ones
796        at the places where they are needed.  But CPS is not a good
797        representation for global code motion, so I'll keep it simple and
798        am not attempting this.
799    
800     2. Incremental record construction (aka record splitting).
801    
802        Long records with many component values which are simulatenously live
803        (recall that single use record addresses are not considered to
804         be live) are constructed with rawrecord and rawupdate.
805        We allocate space on the heap with rawrecord first, then gradually
806        fill it in with rawupdate.  This is the technique suggested to me
807        by Matthias.
808    
809        Some restrictions on when this is applicable:
810        1. It is not a VECTOR record.  The code generator currently does not handle
811           this case. VECTOR record uses double indirection like arrays.
812        2. All the record component values are defined in the same "basic block"
813           as the record constructor.  This is to prevent speculative
814           record construction.
815    
816    ----------------------------------------------------------------------
817    Name: Allen Leung
818    Date: 2002/02/22 01:02:00 EST
819    Tag: leunga-20020222-mlrisc-tools
820    
821    Minor bug fixes in the parser and rewriter
822    
823    ----------------------------------------------------------------------
824    Name: Allen Leung
825    Date: 2002/02/21 20:20:00 EST
826    Tag: leunga-20020221-peephole
827    
828    Regenerated the peephole files.  Some contained typos in the specification
829    and some didn't compile because of pretty printing bugs in the old version
830    of 'nowhere'.
831    
832    ----------------------------------------------------------------------
833    Name: Allen Leung
834    Date: 2002/02/19 20:20:00 EST
835    Tag: leunga-20020219-mlrisc-tools
836    Description:
837    
838       Minor bug fixes to the mlrisc-tools library:
839    
840       1.  Fixed up parsing colon suffixed keywords
841       2.  Added the ability to shut the error messages up
842       3.  Reimplemented the pretty printer and fixed up/improved
843           the pretty printing of handle and -> types.
844       4.  Fixed up generation of literal symbols in the nowhere tool.
845       5.  Added some SML keywords to to sml.sty
846    
847    ----------------------------------------------------------------------
848    Name: Matthias Blume
849    Date: 2002/02/19 16:20:00 EST
850    Tag: blume-20020219-cmffi
851    Description:
852    
853    A wild mix of changes, some minor, some major:
854    
855    * All C FFI-related libraries are now anchored under $c:
856        $/c.cm      --> $c/c.cm
857        $/c-int.cm  --> $c/internals/c-int.cm
858        $/memory.cm --> $c/memory/memory.cm
859    
860    * "make" tool (in CM) now treats its argument pathname slightly
861      differently:
862        1. If the native expansion is an absolute name, then before invoking
863           the "make" command on it, CM will apply OS.Path.mkRelative
864           (with relativeTo = OS.FileSys.getDir()) to it.
865        2. The argument will be passed through to subsequent phases of CM
866           processing without "going native".  In particular, if the argument
867           was an anchored path, then "make" will not lose track of that anchor.
868    
869    * Compiler backends now "know" their respective C calling conventions
870      instead of having to be told about it by ml-nlffigen.  This relieves
871      ml-nlffigen from one of its burdens.
872    
873    * The X86Backend has been split into X86CCallBackend and X86StdCallBackend.
874    
875    * Export C_DEBUG and C_Debug from $c/c.cm.
876    
877    * C type encoding in ml-nlffi-lib has been improved to model the conceptual
878      subtyping relationship between incomplete pointers and their complete
879      counterparts.  For this, ('t, 'c) ptr has been changed to 'o ptr --
880      with the convention of instantiating 'o with ('t, 'c) obj whenever
881      the pointer target type is complete.  In the incomplete case, 'o
882      will be instantiated with some "'c iobj" -- a type obtained by
883      using one of the functors PointerToIncompleteType or PointerToCompleteType.
884    
885      Operations that work on both incomplete and complete pointer types are
886      typed as taking an 'o ptr while operations that require the target to
887      be known are typed as taking some ('t, 'c) obj ptr.
888    
889      voidptr is now a bit "more concrete", namely "type voidptr = void ptr'"
890      where void is an eqtype without any values.  This makes it possible
891      to work on voidptr values using functions meant to operate on light
892      incomplete pointers.
893    
894    * As a result of the above, signature POINTER_TO_INCOMPLETE_TYPE has
895      been vastly simplified.
896    
897    ----------------------------------------------------------------------
898    Name: Matthias Blume
899    Date: 2002/02/19 10:48:00 EST
900    Tag: blume-20020219-pqfix
901    Description:
902    
903    Applied Chris Okasaki's bug fix for priority queues.
904    
905    ----------------------------------------------------------------------
906    Name: Matthias Blume
907    Date: 2002/02/15 17:05:00
908    Tag: Release_110_39
909    Description:
910    
911    Last-minute retagging is becoming a tradition... :-(
912    
913    This is the working release 110.39.
914    
915    ----------------------------------------------------------------------
916    Name: Matthias Blume
917    Date: 2002/02/15 16:00:00 EST
918    Tag: Release_110_39-orig
919    Description:
920    
921    Working release 110.39.  New bootfiles.
922    
923    (Update: There was a small bug in the installer so it wouldn't work
924    with all shells.  So I retagged. -Matthias)
925    
926    ----------------------------------------------------------------------
927    Name: Matthias Blume
928    Date: 2002/02/15 14:17:00 EST
929    Tag: blume-20020215-showbindings
930    Description:
931    
932    Added EnvRef.listBoundSymbols and CM.State.showBindings.  Especially
933    the latter can be useful for exploring what bindings are available at
934    the interactive prompt.  (The first function returns only the list
935    of symbols that are really bound, the second prints those but also the
936    ones that CM's autoloading mechanism knows about.)
937    
938    ----------------------------------------------------------------------
939    Name: Matthias Blume
940    Date: 2002/02/15 12:08:00 EST
941    Tag: blume-20020215-iptrs
942    Description:
943    
944    Two improvements to ml-nlffigen:
945    
946      1. Write files only if they do not exist or if their current contents
947         do not coincide with what's being written.  (That is, avoid messing
948         with the time stamps unless absolutely necessary.)
949    
950      2. Implement a "repository" mechanism for generated files related
951         to "incomplete pointer types".   See the README file for details.
952    
953    ----------------------------------------------------------------------
954    Name: Matthias Blume
955    Date: 2002/02/14 11:50:00 EST
956    Tag: blume-20020214-quote
957    Description:
958    
959    Added a type 't t_' to tag.sml (in ml-nlffi-lib.cm).  This is required
960    because of the new and improved tag generation scheme.  (Thanks to Allen
961    Leung for pointing it out.)
962    
963    ----------------------------------------------------------------------
964    Name: Lal George
965    Date: 2002/02/14 09:55:27 EST 2002
966    Tag: george-20020214-isabelle-bug
967    Description:
968    
969    Fixed the MLRISC bug sent by Markus Wenzel regarding the compilation
970    of Isabelle on the x86.
971    
972    From Allen:
973    -----------
974     I've found the problem:
975    
976         in ra-core.sml, I use the counter "blocked" to keep track of the
977         true number of elements in the freeze queue.  When the counter goes
978         to zero, I skip examining the queue.  But I've messed up the
979         bookkeeping in combine():
980    
981             else ();
982             case !ucol of
983               PSEUDO => (if !cntv > 0 then
984                     (if !cntu > 0 then blocked := !blocked - 1 else ();
985                                        ^^^^^^^^^^^^^^^^^^^^^^^
986                      moveu := mergeMoveList(!movev, !moveu)
987                     )
988                  else ();
989    
990         combine() is called to coalesce two nodes u and v.
991         I think I was thinking that if the move counts of u and v are both
992         greater than zero then after they are coalesced then one node is
993         removed from the freeze queue.  Apparently I was thinking that
994         both u and v are of low degree, but that's clearly not necessarily true.
995    
996    
997    02/12/2002:
998        Here's the patch.  HOL now compiles.
999    
1000        I don't know how this impact on performance (compile
1001        time or runtime).  This bug caused the RA (especially on the x86)
1002        to go thru the potential spill phase when there are still nodes on the
1003        freeze queue.
1004    
1005    
1006    
1007    
1008    ----------------------------------------------------------------------
1009    Name: Matthias Blume
1010    Date: 2002/02/13 22:40:00 EST
1011    Tag: blume-20020213-fptr-rtti
1012    Description:
1013    
1014    Fixed a bug in ml-nlffigen that was introduced with one of the previous
1015    updates.
1016    
1017    ----------------------------------------------------------------------
1018    Name: Matthias Blume
1019    Date: 2002/02/13 16:41:00 EST
1020    Tag: blume-20020213-cmlpq
1021    Description:
1022    
1023    Added new priority queue export symbols (which have just been added to
1024    smlnj-lib.cm) to CML's version of smlnj-lib.cm.  (Otherwise CML would
1025    not compile and the installer would choke.)
1026    
1027    ----------------------------------------------------------------------
1028    Name: Matthias Blume
1029    Date: 2002/02/13 16:15:00 EST
1030    Tag: blume-20020213-various
1031    Description:
1032    
1033    1. More tweaks to ml-nlffigen:
1034    
1035       - better internal datastructures (resulting in slight speedup)
1036       - "-match" option requires exact match
1037       - "localized" gensym counters (untagged structs/unions nested within
1038         other structs/unions or within typedefs get a fresh counter; their
1039         tag will be prefixed by a concatenation of their parents' tags)
1040       - bug fixes (related to calculation of transitive closure of types
1041         to be included in the output)
1042    
1043    2. Minor Basis updates:
1044    
1045       - added implementations for List.collate and Option.app
1046    
1047    ----------------------------------------------------------------------
1048    Name: Matthias Blume
1049    Date: 2002/02/11 15:55:00 EST
1050    Tag: blume-20020211-gensym
1051    Description:
1052    
1053    Added a "-gensym" option to command line of ml-nlffigen.  This can be
1054    used to specify a "stem" -- a string that is inserted in all "gensym'd"
1055    names (ML structure names that correspond to unnamed C structs, unions,
1056    and enums), so that separate runs of ml-nlffigen do not clash.
1057    
1058    ----------------------------------------------------------------------
1059    Name: Matthias Blume
1060    Date: 2002/02/11 12:05:00 EST
1061    Tag: blume-20020211-gensml
1062    Description:
1063    
1064    A quick fix for a problem with GenSML (in the pgraph-util library):
1065    Make generation of toplevel "local" optional.  (Strictly speaking,
1066    signature definitions within "local" are not legal SML.)
1067    
1068    Other than that: updates to INSTALL and cm/TODO.
1069    
1070    ----------------------------------------------------------------------
1071    Name: Matthias Blume
1072    Date: 2002/02/08 15:00:00 EST
1073    Tag: blume-20020208-uniquepid
1074    Description:
1075    
1076    0. Version number has been bumped to 110.38.1.  NEW BOOTFILES!!!
1077    
1078    1. The installer (config/install.sh) has gotten smarter:
1079    
1080         - Configuration options are a bit easier to specify now
1081           (in config/targets).
1082         - Bug in recognizing .tar.bz2 files fixed.
1083         - Installer automatically resolves dependencies between
1084           configuration options (e.g., if you ask for eXene, you will
1085           also get cml -- regardless whether you asked for it or not).
1086         - Installer can run in "quieter mode" by setting the environment
1087           variable INSTALL_QUIETLY to "true".  "Quieter" does not mean
1088           "completely silent", though.
1089         - Build HashCons library as part of smlnj-lib.
1090    
1091    2. A new scheme for assigning persistent identifiers to compilation
1092       units (and, by extension, to types etc.) has been put into place.
1093       This fixes a long-standing bug where types and even dynamic values
1094       can get internally confused, thereby compromising type safety
1095       (abstraction) and dynamic correctness.  See
1096    
1097         http://cm.bell-labs.com/cm/cs/who/blume/pid-confusion.tgz
1098    
1099       for an example of how things could go wrong until now.
1100    
1101       The downside of the new scheme is that pids are not quite as
1102       persistent as they used to be: CM will generate a fresh pid
1103       for every compilation unit that it thinks it sees for the first
1104       time.  That means that if you compile starting from a clean, fresh
1105       source tree at two different times, you end up with different
1106       binaries.
1107    
1108       Cutoff recompilation, however, has not been compromised because
1109       CM keeps pid information in special caches between runs.
1110    
1111    ----------------------------------------------------------------------
1112    Name: Lal George
1113    Date: 2002/02/07 15:34:13 EST 2002
1114    Tag: <none>
1115    Description:
1116    
1117    Compilers that generate assembly code may produce  global labels
1118    whose value is resolved at link time. The various peephole optimization
1119    modules did not take this in account.
1120    
1121    TODO. The Labels.addrOf function should really return an option
1122    type so that clients are forced to deal with this issue, rather
1123    than an exception being raised.
1124    
1125    ----------------------------------------------------------------------
1126    Name: Lal George
1127    Date: 2002/02/06 13:55:02 EST
1128    Tag: george-20020206-ra-breakup
1129    Description:
1130    
1131    1. A bug fix from Allen.
1132    
1133        A typo causes extra fstp %st(0)'s to be generated at compensation
1134        edges, which might cause stack underflow traps at runtime.  This
1135        occurs in fft where there are extraneous fstps right before the 'into'
1136        trap instruction (in this case they are harmless since none of the
1137        integers overflow.)
1138    
1139    2. Pulled out various utility modules that were embedded in the modules
1140       of the register allocator. I need these modules for other purposes, but
1141       they are not complete enough to put into a library (just yet).
1142    ----------------------------------------------------------------------
1143    Name: Matthias Blume
1144    Date: 2002/01/31 16:05:00 EST
1145    Tag: blume-20020131-sparc-ccalls
1146    Description:
1147    
1148    1. C-calls on Sparc needlessly allocated a huge chunk (96 bytes)
1149       of extra stack space by mistake.  Fixed.
1150    
1151    2. Bug in logic of handling of command-line options in ml-nlffigen fixed.
1152    
1153    ----------------------------------------------------------------------
1154    Name: Allen Leung
1155    Date: 2002/01/30
1156    Tag: leunga-20020130-nowhere-bug-fix
1157    Description:
1158    
1159       MLRISC bug fixes:
1160       1. Fixed a bindings computation bug in the 'nowhere' program generator tool.
1161       2. MachineInt.fromString was negating its value.
1162    
1163    ----------------------------------------------------------------------
1164    Name: Matthias Blume
1165    Date: 2002/01/29
1166    Tag: blume-20020129-INSTALL
1167    Description:
1168    
1169    - Added somewhat detailed installation instructions (file INSTALL).
1170    - Fixed curl-detection bug in config/install.sh.
1171    - It is now possible to select the URL getter using the URLGETTER
1172      environment variable:
1173    
1174          not set / "unknown"      --> automatic detection (script tries wget,
1175                                       curl, and lynx)
1176          "wget" / "curl" / "lynx" --> use the specified program (script "knows"
1177                                       how to properly invoke them)
1178          other                    --> use $URLGETTER directly, it must take
1179                                       precisely two command-line arguments
1180                                       (source URL and destination file name)
1181    
1182    ----------------------------------------------------------------------
1183    Name: Matthias Blume
1184    Date: 2002/01/28
1185    Tag: blume-20020128-sparc-ccalls
1186    Description:
1187    
1188    - Fixed problem with calculation of "used" registers in sparc-c-calls.
1189    - Make use of the allocParam argument in sparc-c-calls.
1190    
1191    ----------------------------------------------------------------------
1192    Name: Matthias Blume
1193    Date: 2002/01/28
1194    Tag: blume-20020128-allocParam
1195    Description:
1196    
1197    John Reppy:  Changes c-calls API to accept client-callback for
1198    allocating extra stack space.
1199    me: Corresponding changes to mlriscGen (using a dummy argument that
1200        does not change the current behavior).
1201    
1202    ----------------------------------------------------------------------
1203    Name: Matthias Blume
1204    Date: 2002/01/28 12:00:00
1205    Tag: Release_110_38
1206    Description:
1207    
1208    This time for real!!!
1209    
1210    ----------------------------------------------------------------------
1211    Name: Matthias Blume
1212    Date: 2002/01/28 10:56:00 EST
1213    Tag: blume-20020128-retraction
1214    Description:
1215    
1216    0. Retracted earlier 110.38.  (The Release_110_38 tag has been replaced
1217       with blume-Release_110_38-retracted.)
1218    
1219    1. Fixed a problem with incorrect rounding modes in real64.sml.
1220       (Thanks to Andrew Mccreight <andrew.mccreight@yale.edu>.)
1221    
1222    2. A bug in ml-nlffigen related to the handling of unnamed structs, unions,
1223       and enums fixed.  The naming of corresponding ML identifiers should
1224       now be consistent again.
1225    
1226    ----------------------------------------------------------------------
1227    Name: Allen Leung
1228    Date: 2002/01/27
1229    Tag: leunga-20020127-nowhere
1230    Description:
1231    
1232       Added a target called nowhere in the configuration scripts.
1233       Enabling this will build the MLRISC 'nowhere' tool (for translating
1234       programs with where-clauses into legal SML code) during installation.
1235    
1236    ----------------------------------------------------------------------
1237    Name: Matthias Blume
1238    Date: 2002/01/25 21:27:00 EST
1239    Tag: blume-Release_110_38-retracted
1240    Description:
1241    
1242    Call it a (working) release!  Version is 110.38. Bootfiles are ready.
1243    
1244    README will be added later.
1245    
1246    !!! NOTE:  Re-tagged as blume-Release_110_38-retracted. Original tag
1247    (Release_110_38) removed.  Reason: Last-minute bug fixes.
1248    
1249    ----------------------------------------------------------------------
1250    Name: Matthias Blume
1251    Date: 2002/01/25
1252    Tag: blume-20020125-ffi
1253    Description:
1254    
1255    A large number of tweaks and improvements to ml-nlffi-lib and
1256    ml-nlffigen:
1257    
1258       - ML represenation types have been streamlined
1259       - getter and setter functions work with concrete values, not abstract
1260         ones where possible
1261       - ml-nlffigen command line more flexible (see README file there)
1262       - some bugs have been fixed (hopefully)
1263    
1264    ----------------------------------------------------------------------
1265    Name: Lal George
1266    Date: 2002/01/24
1267    Tag: george-20020124-risc-ra-interface
1268    Description:
1269    
1270       There is a dramatic simplification in the interface to the
1271       register allocator for RISC architectures as a result of making
1272       parallel copy instructions explicit.
1273    
1274    ----------------------------------------------------------------------
1275    Name: Matthias Blume
1276    Date: 2002/01/22
1277    Tag: blume-20020122-x86-ccalls
1278    Description:
1279    
1280    Bug fix for c-calls on x86 (having to do with how char- and
1281    short-arguments are being handled).
1282    
1283    ----------------------------------------------------------------------
1284    Name: Matthias Blume
1285    Date: 2002/01/21
1286    Tag: blume-20020121-ff
1287    Description:
1288    
1289    Another day of fiddling with the FFI...
1290    
1291    1. Bug fix/workaround:  CKIT does not complain about negative array
1292       dimensions, so ml-nlffigen has to guard itself against this possibility.
1293       (Otherwise a negative dimension would send it into an infinite loop.)
1294    
1295    2. Some of the abstract types (light objects, light pointers, most "base"
1296       types) in structure C are now eqtypes.
1297    
1298    3. Added constructors and test functions for NULL function pointers.
1299    
1300    ----------------------------------------------------------------------
1301    Name: Matthias Blume
1302    Date: 2002/01/18
1303    Tag: blume-20020118-ready-for-new-release
1304    Description:
1305    
1306    Made config/srcarchiveurl point to a new place.  (Will provide boot
1307    files shortly.)
1308    
1309    Maybe we christen this to be 110.38?
1310    
1311    ----------------------------------------------------------------------
1312    Name: Matthias Blume
1313    Date: 2002/01/18
1314    Tag: blume-20020118-more-ffifiddle
1315    Description:
1316    
1317    Today's FFI fiddling:
1318    
1319      - Provided a structure CGetSet with "convenient" versions of C.Get.* and
1320        C.Set.* that use concrete (MLRep.*) arguments and results instead
1321        of abstract ones.
1322    
1323      - Provided word-style bit operations etc. for "int" representation
1324        types in MLRep.S<Foo>Bitops where <Foo> ranges over Char, Int, Short,
1325        and Long.
1326    
1327    ----------------------------------------------------------------------
1328    Name: Matthias Blume
1329    Date: 2002/01/18
1330    Tag: blume-20020118-use-x86-fp
1331    Description:
1332    
1333    Now that x86-fast-fp seems to be working, I turned it back on again
1334    by default.  (Seems to work fine now, even with the FFI.)
1335    
1336    Other than that, I added some documentation about the FFI to
1337    src/ml-nlffigen/README and updated the FFI test examples in
1338    src/ml-nlffi-lib/Tests/*.
1339    
1340    ----------------------------------------------------------------------
1341    Name: Allen Leung
1342    Date: 2002/01/17
1343    Tag: leunga-20020117-x86-fast-fp-call
1344    Description:
1345    
1346       1. Fixed a problem with handling return fp values when x86's fast fp
1347          mode is turned on.
1348    
1349       2. Minor pretty printing fix for cellset.  Print %st(0) as %st(0) instead
1350          of %f32.
1351    
1352       3. Added a constructor INT32lit to the ast of MLRISC tools.
1353    
1354    ----------------------------------------------------------------------
1355    Name: Matthias Blume
1356    Date: 2002/01/16
1357    Tag: blume-20020116-ffifiddle
1358    Description:
1359    
1360    More fiddling with the FFI interface:
1361    
1362     - Make constness 'c instead of rw wherever possible.  This eliminates
1363       the need for certain explicit coercions.  (However, due to ML's
1364       value polymorphism, there will still be many cases where explicit
1365       coercions are necessary.  Phantom types are not the whole answer
1366       to modeling a subtyping relationship in ML.)
1367    
1368     - ro/rw coersions for pointers added.  (Avoids the detour through */&.)
1369    
1370     - "printf" test example added to src/ml-nlffi-lib/Tests.  (Demonstrates
1371       clumsy workaround for varargs problem.)
1372    
1373    ----------------------------------------------------------------------
1374    Name: Lal George
1375    Date: 2002/01/15
1376    Tag: <none>
1377    Description:
1378    
1379    1. Since COPY instructions are no longer native to the architecture,
1380       a generic functor can be used to implement the expandCopies function.
1381    
1382    2. Allowed EXPORT and IMPORT pseudo-op declarations to appear inside a
1383       TEXT segment.
1384    
1385    ----------------------------------------------------------------------
1386    Name: Matthias Blume
1387    Date: 2002/01/15
1388    Tag: blume-20020115-ffiupdates
1389    Description:
1390    
1391    1. Fix for bug resulting in single-precision float values being returned
1392       incorrectly from FFI calls.
1393    
1394    2. Small modifications to C FFI API:
1395    
1396        - memory-allocation routines return straight objects (no options)
1397          and raise an exception in out-of-memory situations
1398        - unsafe extensions to cast between function pointers and pointers
1399          from/to ints
1400        - added structure C_Debug as an alternative to structure C where
1401          pointer-dereferencing (|*| and |*!) always check for null-pointers
1402        - added open_lib' to DynLinkage;  open_lib' works like open_lib
1403          but also takes a (possibly empty) list of existing library handles
1404          that the current library depends on
1405    
1406    ----------------------------------------------------------------------
1407    Name: Matthias Blume
1408    Date: 2002/01/10
1409    Tag: blume-20020110-newffigen
1410    Description:
1411    
1412    1. Updates to portable graph code.
1413    
1414    2. Major update to ml-nlffigen and ml-nlffi-lib.  Things are much
1415       more scalable now so that even huge interfaces such as the one
1416       for GTK compile in finite time and space. :-)
1417       See src/ml-nlffigen/README for details on what's new.
1418    
1419    ----------------------------------------------------------------------
1420    Name: Lal George
1421    Date: 2001/01/09 14:31:35 EST 2002
1422    Tag: george-20011206-rm-native-copy
1423    Description:
1424    
1425            Removed the native COPY and FCOPY instructions
1426            from all the architectures and replaced it with the
1427            explicit COPY instruction from the previous commit.
1428    
1429            It is now possible to simplify many of the optimizations
1430            modules that manipulate copies. This has not been
1431            done in this change.
1432    
1433    ----------------------------------------------------------------------
1434    Name: Lal George
1435    Date: 2001/12/06 16:50:13 EST 2001
1436    Tag: george-20011206-mlrisc-instruction
1437    Description:
1438    
1439    Changed the representation of instructions from being fully abstract
1440    to being partially concrete. That is to say:
1441    
1442      from
1443            type instruction
1444    
1445      to
1446            type instr                              (* machine instruction *)
1447    
1448            datatype instruction =
1449                LIVE of {regs: C.cellset, spilled: C.cellset}
1450              | KILL of {regs: C.cellset, spilled: C.cellset}
1451              | COPYXXX of {k: CB.cellkind, dst: CB.cell list, src: CB.cell list}
1452              | ANNOTATION of {i: instruction, a: Annotations.annotation}
1453              | INSTR of instr
1454    
1455    This makes the handling of certain special instructions that appear on
1456    all architectures easier and uniform.
1457    
1458    LIVE and KILL say that a list of registers are live or killed at the
1459    program point where they appear. No spill code is generated when an
1460    element of the 'regs' field is spilled, but the register is moved to
1461    the 'spilled' (which is present, more for debugging than anything else).
1462    
1463    LIVE replaces the (now deprecated) DEFFREG instruction on the alpha.
1464    We used to generate:
1465    
1466            DEFFREG f1
1467            f1 := f2 + f3
1468            trapb
1469    
1470    but now generate:
1471    
1472            f1 := f2 + f3
1473            trapb
1474            LIVE {regs=[f1,f2,f3], spilled=[]}
1475    
1476    Furthermore, the DEFFREG (hack) required that all floating point instruction
1477    use all registers mentioned in the instruction. Therefore f1 := f2 + f3,
1478    defines f1 and uses [f1,f2,f3]! This hack is no longer required resulting
1479    in a cleaner alpha implementation. (Hopefully, intel will not get rid of
1480    this architecture).
1481    
1482    COPYXXX is intended to replace the parallel COPY and FCOPY  available on
1483    all the architectures. This will result in further simplification of the
1484    register allocator that must be aware of them for coalescing purposes, and
1485    will also simplify certain aspects of the machine description that provides
1486    callbacks related to parallel copies.
1487    
1488    ANNOTATION should be obvious, and now INSTR represents the honest to God
1489    machine instruction set!
1490    
1491    The <arch>/instructions/<arch>Instr.sml files define certain utility
1492    functions for making porting easier -- essentially converting upper case
1493    to lower case. All machine instructions (of type instr) are in upper case,
1494    and the lower case form generates an MLRISC instruction. For example on
1495    the alpha we have:
1496    
1497      datatype instr =
1498         LDA of {r:cell, b:cell, d:operand}
1499       | ...
1500    
1501      val lda : {r:cell, b:cell, d:operand} -> instruction
1502        ...
1503    
1504    where lda is just (INSTR o LDA), etc.
1505    
1506    ----------------------------------------------------------------------
1507    Name: Matthias Blume
1508    Date: 2001/11/22 21:40:00 EST
1509    Tag: Release_110_37
1510    Description:
1511    
1512    Release 110.37.  This time for real.
1513    
1514    ----------------------------------------------------------------------
1515    Name: Matthias Blume
1516    Date: 2001/11/21 16:35:00 EST
1517    Tag: blume-20011121-foot-in-mouth
1518    Description:
1519    
1520    Removed the "Release_110_37" tag because of a serious bug.
1521    This will be re-tagged once the bug is fixed.
1522    
1523    ----------------------------------------------------------------------
1524    Name: Matthias Blume
1525    Date: 2001/11/21 16:14:00 EST
1526    Tag: blume-20011121-forgottenfile
1527    Description:
1528    
1529    Forgot to add a file.  (Just a .tex-file -- part of
1530    the CM manual source.)
1531    
1532    ----------------------------------------------------------------------
1533    Name: Matthias Blume
1534    Date: 2001/11/21 16:10:00 EST
1535    Tag: blume-20011121-invalid_110_37
1536    Description:
1537    
1538    Note: I removed the original tag "Release_110_37" from this commit
1539          because we found a serious bug in all non-x86 backends.
1540          - Matthias
1541    
1542    1. Modifications to the SML/NJ code generator and to the runtime system
1543       so that code object name strings are directly inserted into code
1544       objects at code generation time.  The only business the runtime system
1545       has with this is now to read the name strings on occasions.
1546       (The encoding of the name string has also changed somewhat.)
1547    
1548    2. CM now implements a simple "set calculus" for specifying export lists.
1549       In particular, it is now possible to refer to the export lists of
1550       other libraries/groups/sources and form unions as well as differences.
1551       See the latest CM manual for details.
1552    
1553    3. An separate notion of "proxy" libraries has again be eliminated from
1554       CM's model.  (Proxy libraries are now simply a special case of using
1555       the export list calculus.)
1556    
1557    4. Some of the existing libraries now take advantage of the new set
1558       calculus.
1559       (Notice that not all libraries have been converted because some
1560       of the existing .cm-files are supposed to be backward compatible
1561       with 110.0.x.)
1562    
1563    5. Some cleanup in stand-alone programs.  (Don't use "exnMessage" -- use
1564       "General.exnMessage"!  The former relies on a certain hook to be
1565       initialized, and that often does not happen in the stand-alone case.)
1566    
1567    ----------------------------------------------------------------------
1568    Name: Lal George
1569    Date: 2001/11/21  13:56:18 EST
1570    Tag: george-2001121-pseudo-ops
1571    Description:
1572    
1573      Implemented a complete redesign of MLRISC pseudo-ops. Now there
1574      ought to never be any question of incompatabilities with
1575      pseudo-op syntax expected by host assemblers.
1576    
1577      For now, only modules supporting GAS syntax are implemented
1578      but more should follow, such as MASM, and vendor assembler
1579      syntax, e.g. IBM as, Sun as, etc.
1580    
1581    ----------------------------------------------------------------------
1582    Name: Matthias Blume
1583    Date: 2001/11/14 11:52:00 EST
1584    Tag: blume-20011114-srcname
1585    Description:
1586    
1587    1. Routed the name of the current source file to mlriscgen where it
1588       should be directly emitted into the code object.  (This last part
1589       is yet to be done.)
1590    
1591    2. Some cleanup of the pgraph code to make it match the proposal that
1592       I put out the other day.  (The proposal notwithstanding, things are
1593       still in flux here.)
1594    
1595    ----------------------------------------------------------------------
1596    Name: Lal George
1597    Date: 2001/11/14 09:44:04 EST
1598    Tag:
1599    Description:
1600    
1601      Fix for a backpatching bug reported by Allen.
1602    
1603      Because the boundary between short and long span-dependent
1604      instructions is +/- 128, there are an astounding number of
1605      span-dependent instructions whose size is over estimated.
1606    
1607      Allen came up with the idea of letting the size of span
1608      dependent instructions be non-monotonic, for a maxIter
1609      number of times, after which the size must be monotonically
1610      increasing.
1611    
1612      This table shows the number of span-dependent instructions
1613      whose size was over-estimated as a function of maxIter, for the
1614      file Parse/parse/ml.grm.sml:
1615    
1616         maxIter            # of instructions:
1617            10                      687
1618            20                      438
1619            30                      198
1620            40                        0
1621    
1622      In compiling the compiler, there is no significant difference in
1623      compilation speed between maxIter=10 and maxIter=40. Actually,
1624      my measurements showed that maxIter=40 was a tad faster than
1625      maxIter=10! Also 96% of the  files in the compiler reach a fix
1626      point within 13 iterations, so fixing maxIter at 40, while high,
1627      is okay.
1628    
1629    ----------------------------------------------------------------------
1630    Name: Matthias Blume
1631    Date: 2001/10/31 15:25:00 EST
1632    Tag: blume-20011031-pgraph
1633    Description:
1634    
1635    CKIT:
1636    * Changed the "Function" constructor of type Ast.ctype to carry optional
1637      argument identifiers.
1638    * Changed the return type of TypeUtil.getFunction accordingly.
1639    * Type equality ignores the argument names.
1640    * TypeUtil.composite tries to preserve argument names but gives up quickly
1641      if there is a mismatch.
1642    
1643    installation script:
1644    * attempts to use "curl" if available (unless "wget" is available as well)
1645    
1646    CM:
1647    * has an experimental implementation of "portable graphs" which I will
1648      soon propose as an implementation-independent library format
1649    * there are also new libraries $/pgraph.cm and $/pgraph-util.cm
1650    
1651    NLFFI-LIB:
1652    * some cleanup (all cosmetic)
1653    
1654    NLFFIGEN:
1655    * temporarily disabled the mechanism that suppresses ML output for
1656      C definitions whose identifiers start with an underscore character
1657    * generate val bindings for enum constants
1658    * user can request that only one style (light or heavy) is being used;
1659      default is to use both (command-line arguments: -heavy and -light)
1660    * fixed bug in handling of function types involving incomplete pointers
1661    * generate ML entry points that take record arguments (i.e., using
1662      named arguments) for C functions that have a prototype with named
1663      arguments
1664      (see changes to CKIT)
1665    
1666    ----------------------------------------------------------------------
1667    Name: Allen Leung
1668    Date: 2001/10/27 20:34:00 EDT
1669    Tag: leunga-20011027-x86-fast-fp-call
1670    Description:
1671    
1672       Fixed the bug described in blume-20010920-slowfp.
1673    
1674       The fix involves
1675          1. generating FCOPYs in FSTP in ia32-svid
1676          2. marking a CALL with the appropriate annotation
1677    
1678    ----------------------------------------------------------------------
1679    Name: Matthias Blume
1680    Date: 2001/10/16 11:32:00 EDT
1681    Tag: blume-20011016-netbsd
1682    Description:
1683    
1684    Underscore patch from Chris Richards (fixing problem with compiling
1685    runtime system under recent NetBSD).
1686    
1687    ----------------------------------------------------------------------
1688    Name: Allen Leung
1689    Date: 2001/10/12 17:18:32 EDT 2001
1690    Tag: leung-20011012-x86-printflowgraph
1691    Description:
1692    
1693    X86RA now uses a valid (instead of dummy) PrintFlowgraph module.
1694    
1695    ----------------------------------------------------------------------
1696    Name: Lal George
1697    Date: 2001/10/11 23:51:34 EDT
1698    Tag: george-20011011-too-many-instrs
1699    Description:
1700    
1701    The representation of a program point never expected to see more
1702    than 65536 instructions in a basic block!
1703    
1704    ----------------------------------------------------------------------
1705    Name: Lal George
1706    Date: 2001/10/09 09:41:37 EDT
1707    Tag: george-20011008-mlrisc-labels
1708    Description:
1709    
1710    Changed the machine description files to support printing of
1711    local and global labels in assembly code, based on host assembler
1712    conventions.
1713    
1714    ----------------------------------------------------------------------
1715    Name: Matthias Blume
1716    Date: 2001/09/25 15:25:00 EDT
1717    Tag: blume-20010925-exninfo
1718    Description:
1719    
1720    I provided a non-hook implementation of exnName (at the toplevel) and
1721    made the "dummy" implementation of exnMessage (at the toplevel) more
1722    useful: if nothing gets "hooked in", then at least you are going to
1723    see the exception name and a message indicating why you don't see more.
1724    
1725    [For the time being, programs that need exnMessage and want to use
1726    ml-build should either use General.exnMessage (strongly recommended) or
1727    refer to structure General at some other point so that CM sees a
1728    static dependency.]
1729    
1730    [Similar remarks go for "print" and "use":  If you want to use their
1731    functionality in stand-alone programs generated by ml-build, then use
1732    TextIO.output and Backend.Interact.useFile (from $smlnj/compiler.cm).]
1733    
1734    ----------------------------------------------------------------------
1735    Name: Matthias Blume
1736    Date: 2001/09/20 17:28:00 EDT
1737    Tag: blume-20010920-slowfp
1738    Description:
1739    
1740    Allen says that x86-fast-fp is not safe yet, so I turned it off again...
1741    
1742    ----------------------------------------------------------------------
1743    Name: Matthias Blume
1744    Date: 2001/09/20 17:20:00 EDT
1745    Tag: blume-20010920-canonicalpaths
1746    Description:
1747    
1748    0. Updated the BOOT file (something that I forgot to do earlier).
1749    
1750    1. Small internal change to CM so that it avoids "/../" in filenames
1751       as much as possible (but only where it is safe).
1752    
1753    2. Changed config/_run-sml (resulting in a changed bin/.run-sml) so
1754       that arguments that contain delimiters are passed through correctly.
1755       This change also means that all "special" arguments of the form
1756       @SMLxxx... must come first.
1757    
1758    3. Changed install script to put relative anchor names for tool commands
1759       into pathconfig.
1760    
1761    ----------------------------------------------------------------------
1762    Name: Matthias Blume
1763    Date: 2001/09/18 15:35:00 EDT
1764    Tag: blume-20010918-readme11036
1765    Description:
1766    
1767    Added README files.
1768    
1769    ----------------------------------------------------------------------
1770    Name: Matthias Blume
1771    Date: 2001/09/18 11:45:00 EDT
1772    Tag: Release_110_36 (retag)
1773    Description:
1774    
1775    Fixed mistake in config/preloads. Retagged as 110.36.
1776    
1777    ----------------------------------------------------------------------
1778    Name: Matthias Blume
1779    Date: 2001/09/18 09:40:00 EDT
1780    Tag: Release_110_36_orig (tag changed)
1781    Description:
1782    
1783    New version (110.36).  New bootfiles.
1784    
1785    ----------------------------------------------------------------------
1786    Name: Matthias Blume
1787    Date: 2001/09/14 16:15:00 EDT
1788    Tag: blume-20010914-x86fastfp
1789    Description:
1790    
1791    John committed some changes that Allen made, in particular a (hopefully)
1792    correctly working version of the x86-fp module.
1793    
1794    I changed the default setting of the Control.MLRISC.getFlag "x86-fast-fp"
1795    flag to "true".  Everything seems to compile to a fixpoint ok, and
1796    "mandelbrot" speeds up by about 15%.
1797    
1798    ----------------------------------------------------------------------
1799    Name: Matthias Blume
1800    Date: 2001/09/13 11:20:00 EDT
1801    Tag: blume-20010913-minimal
1802    Description:
1803    
1804    1. Stefan Monnier's patch to fix a miscompilation problem that
1805       was brought to light by John Reppy's work on Moby.
1806    
1807    2. Implemented a minimal "structure Compiler" that contains just
1808       "version" and "architecture".  The minimal version will be
1809       available when the full version is not.  This is for backward-
1810       compatibility with code that wants to test Compiler.version.
1811    
1812    ----------------------------------------------------------------------
1813    Name: Matthias Blume
1814    Date: 2001/08/28 14:03:00 EDT
1815    Tag: blume-20010828-ml-lex
1816    Description:
1817    
1818    Fix for bug 1581, received from Neophytos Michael.
1819    
1820    ----------------------------------------------------------------------
1821    Name: Matthias Blume
1822    Date: 2001/08/27 11:20:00 EDT
1823    Tag: blume-20010827-readme11035
1824    Description:
1825    
1826    Fleshed out the README file for 110.35.
1827    
1828    ----------------------------------------------------------------------
1829    Name: Matthias Blume
1830    Date: 2001/08/24 17:10:00 EDT
1831    Tag: Release_110_35
1832    Description:
1833    
1834    New version number (110.35).  New bootfiles.
1835    
1836    ----------------------------------------------------------------------
1837    Name: Lal George
1838    Date: 2001/08/24 13:47:18 EDT 2001
1839    Tag: george-20010824-MLRISC-graphs
1840    Description:
1841    
1842     removed clusters from MLRISC completely and replaced with graphs.
1843    
1844    ----------------------------------------------------------------------
1845    Name: Matthias Blume
1846    Date: 2001/08/23 17:50:00 EDT
1847    Tag: blume-20010823-toplevel
1848    Description:
1849    
1850    - some reorganization of the code that implements various kinds of
1851      environments in the compiler (static, dynamic, symbolic, combined)
1852    - re-implemented the EnvRef module so that evalStream works properly
1853      (if the stream contains references to "use", "CM.make", etc.)
1854    - cleaned up evalloop.sml and interact.sml (but they need more cleaning)
1855    
1856    ----------------------------------------------------------------------
1857    Name: Matthias Blume
1858    Date: 2001/08/20 15:50 EDT
1859    Tag: blume20010820-slipup
1860    Description:
1861    
1862    I forgot to commit a few files.  Here they are...
1863    
1864    ----------------------------------------------------------------------
1865    Name: Matthias Blume
1866    Date: 2001/08/20 15:35:00 EDT
1867    Tag: blume-20010820-debugprof
1868    Description:
1869    
1870    !!!! NEW BOOTFILES !!!!
1871    
1872    This is another round of reorganizing the compiler sources.  This
1873    time the main goal was to factor out all the "instrumentation"
1874    passes (for profiling and backtracing) into their own library.
1875    The difficulty was to do it in such a way that it does not depend
1876    on elaborate.cm but only on elabdata.cm.
1877    
1878    Therefore there have been further changes to both elaborate.cm and
1879    elabdata.cm -- more "generic" things have been moved from the former
1880    to the latter.  As a result, I was forced to split the assignment
1881    of numbers indicating "primtyc"s into two portions: SML-generic and
1882    SML/NJ-specific.  Since it would have been awkward to maintain,
1883    I bit the bullet and actually _changed_ the mapping between these
1884    numbers and primtycs.  The bottom line of this is that you need
1885    a new set of bin- and bootfiles.
1886    
1887    I have built new bootfiles for all architectures, so doing a fresh
1888    checkout and config/install.sh should be all you need.
1889    
1890    The newly created library's name is
1891    
1892        $smlnj/viscomp/debugprof.cm
1893    
1894    and its sources live under
1895    
1896        src/compiler/DebugProf
1897    
1898    ----------------------------------------------------------------------
1899    Name: Matthias Blume
1900    Date: 2001/08/15 17:15:00 EDT
1901    Tag: blume-20010815-compreorg
1902    Description:
1903    
1904    This is a first cut at reorganizing the CM libraries that make up the
1905    core of the compiler.  The idea is to separate out pieces that could
1906    be used independently by tools, e.g., the parser, the typechecker, etc.
1907    
1908    The current status is a step in this direction, but it is not quite
1909    satisfactory yet.  Expect more changes in the future.
1910    
1911    Here is the current (new) organization...
1912    
1913        What used to be $smlnj/viscomp/core.cm is now divided into
1914        six CM libraries:
1915    
1916             $smlnj/viscomp/basics.cm
1917                           /parser.cm
1918                           /elabdata.cm
1919                           /elaborate.cm
1920                           /execute.cm
1921                           /core.cm
1922    
1923        The CM files for these libraries live under src/system/smlnj/viscomp.
1924        All these libraries are proxy libraries that contain precisely
1925        one CM library component.  Here are the locations of the components
1926        (all within the src/compiler tree):
1927    
1928             Basics/basics.cm
1929             Parse/parser.cm
1930             ElabData/elabdata.cm
1931             Elaborator/elaborate.cm
1932             Execution/execute.cm
1933             core.cm
1934    
1935         [This organization is the same that has been used already
1936         for a while for the architecture-specific parts of the visible
1937         compiler and for the old version of core.cm.]
1938    
1939         As you will notice, many source files have been moved from their
1940         respective original locations to a new home in one of the above
1941         subtrees.
1942    
1943         The division of labor between the new libraries is the following:
1944    
1945             basics.cm:
1946                - Simple, basic definitions that pertain to many (or all) of
1947                  the other libraries.
1948             parser.cm:
1949                - The SML parser, producing output of type Ast.dec.
1950                - The type family for Ast is also defined and exported here.
1951             elabdata.cm:
1952                - The datatypes that describe input and output of the elaborator.
1953                  This includes types, absyn, and static environments.
1954             elaborator.cm:
1955                - The SML/NJ type checker and elaborator.
1956                  This maps an Ast.dec (with a given static environment) to
1957                  an Absyn.dec (with a new static environment).
1958                - This libraries implements certain modules that used to be
1959                  structures as functors (to remove dependencies on FLINT).
1960             execute.cm:
1961                - Everything having to do with executing binary code objects.
1962                - Dynamic environments.
1963             core.cm:
1964                - SML/NJ-specific instantiations of the elaborator and MLRISC.
1965                - Top-level modules.
1966                - FLINT (this should eventually become its own library)
1967    
1968    Notes:
1969    
1970    I am not 100% happy with the way I separated the elaborator (and its
1971    data structures) from FLINT.  Two instances of the same problem:
1972    
1973        1. Data structures contain certain fields that carry FLINT-specific
1974           information.  I hacked around this using exn and the property list
1975           module from smlnj-lib.  But the fact that there are middle-end
1976           specific fields around at all is a bit annoying.
1977    
1978        2. The elaborator calculates certain FLINT-related information.  I tried
1979           to make this as abstract as I could using functorization, but, again,
1980           the fact that the elaborator has to perform calculations on behalf
1981           of the middle-end at all is not nice.
1982    
1983        3. Having to used exn and property lists is unfortunate because it
1984           weakens type checking.  The other alternative (parameterizing
1985           nearly *everything*) is not appealing, though.
1986    
1987    I removed the "rebinding =" warning hack because due to the new organization
1988    it was awkward to maintain it.  As a result, the compiler now issues some of
1989    these warnings when compiling init.cmi during bootstrap compilation. On
1990    the plus side, you also get a warning when you do, for example:
1991       val op = = Int32.+
1992    which was not the case up to now.
1993    
1994    I placed "assign" and "deref" into the _Core structure so that the
1995    code that deals with the "lazy" keyword can find them there.  This
1996    removes the need for having access to the primitive environment
1997    during elaboration.
1998    
1999    ----------------------------------------------------------------------
2000    Name: Matthias Blume
2001    Date: 2001/08/13
2002    Tag: blume-20010813-closures
2003    Description:
2004    
2005    This fix was sent to us by Zhong Shao.  It is supposed to improve the
2006    performance of certain loops by avoiding needless closure allocation.
2007    
2008    ----------------------------------------------------------------------
2009    Name: Lal George
2010    Date: 2001/07/31 10:03:23 EDT 2001
2011    Tag: george-20010731-x86-fmalloc
2012    Description: Fixed bug in x86 calls
2013    
2014        There was a bug where call instructions would mysteriously
2015        vanish. The call instruction had to be one that returned
2016        a floating point value.
2017    
2018    ----------------------------------------------------------------------
2019    Name: Lal George
2020    Date: 2001/07/19 16:36:29 EDT 2001
2021    Tag: george-20010719-simple-cells
2022    Description:
2023    
2024    I have dramatically simplified the interface for CELLS in MLRISC.
2025    
2026    In summary, the cells interface is broken up into three parts:
2027    
2028      1. CellsBasis : CELLS_BASIS
2029    
2030            CellsBasis is a top level structure and common for all
2031            architectures.  it contains the definitions of basic datatypes
2032            and utility  functions over these types.
2033    
2034      2. functor Cells() : CELLS
2035    
2036            Cells generates an interface for CELLS that incorporates the
2037            specific resources on the target architecture, such as the
2038            presence of special register classes, their number and size,
2039            and various useful substructures.
2040    
2041      3. <ARCH>CELLS
2042    
2043            e.g. SparcCells: SPARCCELLS
2044    
2045            <ARCH>CELLS usually contains additional bindings for special
2046            registers  on the architecture, such as:
2047    
2048                    val r0 : cell           (* register zero *)
2049                    val y : cell            (* Y register *)
2050                    val psr : cell          (* processor status register *)
2051                    ...
2052    
2053            The structure returned by applying the Cells functor is opened
2054            in this interface.
2055    
2056    The main implication of all this is that the datatypes for cells is
2057    split between CellsBasis and CELLS -- a fairly simple change for user
2058    code.
2059    
2060    In the old scheme the CELLS interface had a definitional binding of
2061    the form:
2062    
2063            signature CELLS = sig
2064    
2065               structure CellsBasis = CellsBasis
2066    
2067               ...
2068    
2069            end
2070    
2071    With all the sharing constraints that goes on in MLRISC, this old
2072    design  quickly leads to errors such as:
2073    
2074            "structure definition spec inside of sharing ... "
2075    
2076    
2077    and appears to require an unacceptable amount of sharing and where
2078    constraint hackery.
2079    
2080    I think this error message (the interaction of definitional specs and
2081    sharing) requires more explanation on our web page.
2082    
2083    ----------------------------------------------------------------------
2084    Name: Matthias Blume
2085    Date: 2001/07/19 15:00:00 EDT
2086    Tag: blume-20010719-libreorg
2087    Description:
2088    
2089    This update puts together a fairly extensive but straightforward change
2090    to the way the libraries that implement the interactive system are
2091    organized:
2092    
2093       The biggest change is the elimination of structure Compiler.  As a
2094       replacement for this structure, there is now a CM library
2095       (known as $smlnj/compiler.cm or $smlnj/compiler/current.cm)
2096       that exports all the substructures of the original structure Compiler
2097       directly.  So instead of saying Compiler.Foo.bar one now simply
2098       says Foo.bar.  (The CM libraries actually export a collection of
2099       structures that is richer than the collection of substructures of
2100       structure Compiler.)
2101    
2102       To make the transition smooth, there is a separate library called
2103       $smlnj/compiler/compiler.cm which puts together and exports the
2104       original structure Compiler (or at least something very close to it).
2105    
2106       There are five members of the original structure Compiler
2107       that are not exported directly but which instead became members
2108       of a new structure Backend (described by signature BACKEND).  These are:
2109       structure Profile (: PROFILE), structure Compile (: COMPILE), structure
2110       Interact (: INTERACT), structure Machine (: MACHINE), and val
2111       architecture (: string).
2112    
2113       Structure Compiler.Version has become structure CompilerVersion.
2114    
2115       Cross-compilers for alpha32, hppa, ppc, sparc, and x86 are provided
2116       by $smlnj/compiler/<arch>.cm where <arch> is alpha32, hppa, ppc, sparc,
2117       or x86, respectively.
2118       Each of these exports the same frontend structures that
2119       $smlnj/compiler.cm exports.  But they do not have a structure Backend
2120       and instead export some structure <Arch>Backend where <Arch> is Alpha32,
2121       Hppa, PPC, Sparc, or X86, respectively.
2122    
2123       Library $smlnj/compiler/all.cm exports the union of the exports of
2124       $smlnj/compiler/<arch>.cm
2125    
2126       There are no structures <Arch>Compiler anymore, use
2127       $smlnj/compiler/<arch>.cm instead.
2128    
2129       Library host-compiler-0.cm is gone.  Instead, the internal library
2130       that instantiates CM is now called cm0.cm.  Selection of the host
2131       compiler (backend) is no longer done here but. (Responsibility for it
2132       now lies with $smlnj/compiler/current.cm.  This seems to be more
2133       logical.)
2134    
2135       Many individual files have been moved or renamed.  Some files have
2136       been split into multiple files, and some "dead" files have been deleted.
2137    
2138    Aside from these changes to library organization, there are also changes
2139    to the way the code itself is organized:
2140    
2141       Structure Binfile has been re-implemented in such a way that it no
2142       longer needs any knowledge of the compiler.  It exclusively deals
2143       with the details of binfile layout.  It no longer invokes the
2144       compiler (for the purpose of creating new prospective binfile
2145       content), and it no longer has any knowledge of how to interpret
2146       pickles.
2147    
2148       Structure Compile (: COMPILE) has been stripped down to the bare
2149       essentials of compilation.  It no longer deals with linking/execution.
2150       The interface has been cleaned up considerably.
2151    
2152       Utility routines for dealing with linking and execution have been
2153       moved into their own substructures.
2154    
2155       (The ultimate goal of these changes is to provide a light-weight
2156       binfile loader/linker (at least for, e.g., stable libraries) that
2157       does not require CM or the compiler to be present.)
2158    
2159    CM documentation has been updated to reflect the changes to library
2160    organization.
2161    
2162    ----------------------------------------------------------------------
2163    Name: Matthias Blume
2164    Date: 2001/07/10 17:30:00 EDT
2165    Tag: Release_110_34
2166    Description:
2167    
2168    Minor tweak to 110.34 (re-tagged):
2169    
2170      - README.html file added to CVS repository
2171      - runtime compiles properly under FreeBSD 3.X and 4.X
2172    
2173    ----------------------------------------------------------------------
2174    Name: Matthias Blume
2175    Date: 2001/07/10 17:30:00 EDT
2176    Tag: Release_110_34
2177    Description:
2178    
2179    New version number (110.34). New bootfiles.
2180    
2181    ----------------------------------------------------------------------
2182    Name: Matthias Blume
2183    Date: 2001/07/09 16:00:00 EDT
2184    Tag: blume-20010709-more-varargs
2185    Description:
2186    
2187    I changed the handling of varargs in ml-nlffigen again:
2188    The ellipsis ... will now simply be ignored (with an accompanying warning).
2189    
2190    The immediate effect is that you can actually call a varargs function
2191    from ML -- but you can't actually supply any arguments beyond the ones
2192    specified explicitly.  (For example, you can call printf with its format
2193    string, but you cannot pass additional arguments.)
2194    
2195    This behavior is only marginally more useful than the one before, but
2196    it has the advantage that a function or, more importantly, a function
2197    type never gets dropped on the floor, thus avoiding follow-up problems with
2198    other types that refer to the offending one.
2199    
2200    ----------------------------------------------------------------------
2201    Name: Matthias Blume
2202    Date: 2001/07/09 11:25:00 EDT
2203    Tag: blume-20010709-varargs
2204    Description:
2205    
2206    1. ckit-lib.cm now exports structure Error
2207    2. ml-nlffigen reports occurences of "..." (i.e., varargs function types)
2208       with a warning accompanied by a source location.  Moreover, it
2209       merely skips the offending function or type and proceeds with the
2210       rest of its work.u  As a result, one can safely feed C code containing
2211       "..." to ml-nlffigen.
2212    3. There are some internal improvements to CM, providing slightly
2213       more general string substitutions in the tools subsystem.
2214    
2215    ----------------------------------------------------------------------
2216    Name: Matthias Blume
2217    Date: 2001/06/27 15:10:00 EDT
2218    Tag: blume-20010627-concur
2219    Description:
2220    
2221    Fixed a small bug in CM's handling of parallel compilation.
2222    (You could observe the bug by Control-C-interrupting an ordinary
2223    CMB.make or CM.stabilize and then attaching some compile servers.
2224    The result was that all of a sudden the previously interrupted
2225    compilation would continue on its own.  This was because of
2226    an over-optimization: CM did not bother to clean out certain queues
2227    when no servers were attached "anyway", resulting in the contents
2228    of these queues to grab control when new servers did get attached.)
2229    
2230    There is also another minor update to the CM manual.
2231    
2232    ----------------------------------------------------------------------
2233    Name: Matthias Blume
2234    Date: 2001/06/26 16:15:00 EDT
2235    Tag: blume-20010626-cmdoc
2236    Description:
2237    
2238    Minor typo fixed in CM manual (syntax diagram for libraries).
2239    
2240    ----------------------------------------------------------------------
2241    Name: Matthias Blume
2242    Date: 2001/06/25 22:55:00 EDT
2243    Tag: blume-20010625-x86pc
2244    Description:
2245    
2246    Fixed a nasty bug in the X86 assembly code that caused signal
2247    handlers to fail (crash) randomly.
2248    
2249    ----------------------------------------------------------------------
2250    Name: Matthias Blume
2251    Date: 2001/06/25 12:05:00 EDT
2252    Tag: blume-20010625-nlffigen
2253    Description:
2254    
2255    This update fixes a number of minor bugs in ml-nlffigen as reported by
2256    Nick Carter <nbc@andrew.cmu.edu>.
2257    
2258      1. Silly but ok typedefs of the form "typedef void myvoid;" are now accepted.
2259      2. Default names for generated files are now derived from the name of
2260         the C file *without its directory*.  In particular, this causes generated
2261         files to be placed locally even if the C file is in some system directory.
2262      3. Default names for generated signatures and structures are also derived
2263         from the C file name without its directory.  This avoids silly things
2264         like "structure GL/GL".
2265         (Other silly names are still possible because ml-nlffigen does not do
2266          a thorough check of whether generated names are legal ML identifiers.
2267          When in doubt, use command line arguments to force particular names.)
2268    
2269    ----------------------------------------------------------------------
2270    Name: Matthias Blume
2271    Date: 2001/06/21 12:25:00 EDT
2272    Tag: blume-20010621-eXene
2273    Description:
2274    
2275    eXene now compiles and (sort of) works again.
2276    
2277    The library name (for version > 110.33) is $/eXene.cm.
2278    
2279    I also added an new example in src/eXene/examples/nbody.  See the
2280    README file there for details.
2281    
2282    ----------------------------------------------------------------------
2283    Name: Matthias Blume
2284    Date: 2001/06/20 16:40:00 EDT
2285    Tag: blume-20010620-cml
2286    Description:
2287    
2288    CML now compiles and works again.
2289    
2290    Libraries (for version > 110.33):
2291    
2292      $cml/cml.cm            Main CML library.
2293      $cml/basis.cm          CML's version of $/basis.cm.
2294      $cml/cml-internal.cm   Internal helper library.
2295      $cml/core-cml.cm       Internal helper library.
2296      $cml-lib/trace-cml.cm  Tracing facility.
2297      $cml-lib/smlnj-lib.cm  CML's version of $/smlnj-lib.cm
2298    
2299    The installer (config/install.sh) has been taught how to properly
2300    install this stuff.
2301    
2302    ----------------------------------------------------------------------
2303    Name: Matthias Blume
2304    Date: 2001/06/19 17:55:00 EDT
2305    Tag: blume-20010619-instantiate
2306    Description:
2307    
2308    This un-breaks the fix for bug 1432.
2309    (The bug was originally fixed in 110.9 but I broke it again some
2310    time after that.)
2311    
2312    ----------------------------------------------------------------------
2313    Name: Matthias Blume
2314    Date: 2001/06/19 17:25:00 EDT
2315    Tag: blume-20010619-signals
2316    Description:
2317    
2318    This should (hopefully) fix the long-standing signal handling bug.
2319    (The runtime system was constructing a continuation record with an
2320    incorrect descriptor which would cause the GC to drop data on the floor...)
2321    
2322    ----------------------------------------------------------------------
2323    Name: Matthias Blume
2324    Date: 2001/06/15 15:05:00 EDT
2325    Tag: blume-20010615-moresparc
2326    Description:
2327    
2328    Here is a short late-hour update related to Sparc c-calls:
2329    
2330     -- made handling of double-word arguments a bit smarter
2331    
2332     -- instruction selection phase tries to collapse certain clumsily
2333        constructed ML-Trees; typical example:
2334    
2335            ADD(ty,ADD(_,e,LI d1),LI d2)  ->  ADD(ty,e,LI(d1+d2))
2336    
2337        This currently has no further impact on SML/NJ since mlriscGen does
2338        not seem to generate such patterns in the first place, and c-calls
2339        (which did generate them in the beginning) has meanwhile been fixed
2340        so as to avoid them as well.
2341    
2342    ----------------------------------------------------------------------
2343    Name: Matthias Blume
2344    Date: 2001/06/15 15:05:00 EDT
2345    Tag: blume-20010615-sparc
2346    Description:
2347    
2348    The purpose of this update is to provide an implementation of NLFFI
2349    on Sparc machines.
2350    
2351    Here are the changes in detail:
2352    
2353       * src/MLRISC/sparc/c-calls/sparc-c-calls.sml is a new file containing
2354       the Sparc implementation of the c-calls API.
2355       * The Sparc backend of SML/NJ has been modified to uniformely use %fp
2356       for accessing the ML frame.  Thus, we have a real frame pointer and
2357       can freely modify %sp without need for an omit-frame-ptr phase.
2358       The vfp logic in src/compiler/CodeGen/* has been changed to accomodate
2359       this case.
2360       * ml-nlffigen has been taught to produce code for different architectures
2361       and calling conventions.
2362       * In a way similar to what was done in the x86 case, the Sparc
2363       backend uses its own specific extension to mltree.  (For example,
2364       it needs to be able to generate UNIMP instructions which are part
2365       of the calling convention.)
2366       * ml-nlffi-lib was reorganized to make it more modular (in particular,
2367       to make it easier to plug in new machine- and os-dependent parts).
2368    
2369    There are some other fairly unrelated bug fixes and cleanups as well:
2370    
2371       * I further hacked the .cm files for MLRISC tools (like MDLGen) so
2372       that they properly share their libraries with existing SML/NJ libraries.
2373       * I fixed a minor cosmetic bug in CM, supressing certain spurious
2374       follow-up error messages.
2375       * Updates to CM/CMB documentation.
2376    
2377    TODO items:
2378    
2379       * MLRISC should use a different register as its asmTemp on the Sparc.
2380         (The current %o2 is a really bad choice because it is part of the
2381         calling conventions, so things might interfere in unexpected ways.)
2382    
2383    ----------------------------------------------------------------------
2384    Name: Matthias Blume
2385    Date: 2001/06/07
2386    Tag: blume-20010607-calls
2387    Description:
2388    
2389    A number of internal changes related to C calls and calling conventions:
2390    
2391    1. ML-Tree CALL statements now carry a "pops" field.  It indicates the
2392       number of bytes popped implicitly (by the callee).  In most cases
2393       this field is 0 but on x86/win32 it is some non-zero value.  This
2394       is information provided for the benefit of the "omit-frameptr" pass.
2395    2. The CALL instruction on the x86 carries a similar "pops" field.
2396       The instruction selection phase copies its value from the ML-Tree
2397       CALL statement.
2398    3. On all other architectures, the instruction selection phase checks
2399       whether "pops=0" and complains if not.
2400    4. The c-calls implementation for x86 now accepts two calling conventions:
2401       "ccall" and "stdcall".  When "ccall" is selected, the caller cleans
2402       up after the call and pops is set to 0.  For "stdcall", the caller
2403       does nothing, leaving the cleanup to the callee; pops is set to
2404       the number of bytes that were pushed onto the stack.
2405    5. The cproto decoder (compiler/Semant/types/cproto.sml) now can
2406       distinguish between "ccall" and "stdcall".
2407    6. The UNIMP instruction has been added to the supported Sparc instruction
2408       set. (This is needed for implementing the official C calling convention
2409       on this architecture.)
2410    7. I fixed some of the .cm files under src/MLRISC/Tools to make them
2411       work with the latest CM.
2412    
2413    ----------------------------------------------------------------------
2414    Name: Matthias Blume
2415    Date: 2001/06/05 15:10:00 EDT
2416    Tag: blume-20010605-cm-index
2417    Description:
2418    
2419    0. The "lambdasplit" parameter for class "sml" in CM has been documented.
2420    
2421    1. CM can now generate "index files".  These are human-readable files
2422       that list on a per-.cm-file basis each toplevel symbol defined or
2423       imported.  The location of the index file for
2424       <p>/<d>.cm is <p>/CM/INDEX/<d>.cm.
2425       To enable index-file generation, set CM.Control.generate_index to true
2426       or export an environment-symbol: export CM_GENERATE_INDEX=true.
2427    
2428       The CM manual has been updated accordingly.
2429    
2430    2. I made some slight modifications to the c-calls API in MLRISC.
2431    
2432         a) There is now a callback to support saving/restoring of
2433            dedicated but caller-save registers around the actual call
2434            instruction.
2435         b) One can optionally specify a comment-annotation for the
2436            call instruction.
2437    
2438    3. SML/NJ (mlriscGen.sml) uses this new API for the rawccall primop.
2439       (For example, the comment annotation shows the C prototype of
2440        the function being called.)
2441    
2442    ----------------------------------------------------------------------
2443    Name: Matthias Blume
2444    Date: 2001/06/01 13:30:00 EDT
2445    Tag: blume-20010601-nlffi-cleanup
2446    Description:
2447    
2448    This is mostly a cleanup of MLFFI stuff:
2449    
2450       - some signature files have been put into a more exposed place
2451       - the ugly 'f type parameter is gone (simplifies types tremendously!)
2452       - ml-nlffigen changed accordingly
2453       - tutorial updated
2454    
2455    Other changes:
2456    
2457       - author's affiliation in CM manual(s) updated
2458       - some more recognized keywords added to Allen's sml.sty
2459    
2460    ----------------------------------------------------------------------
2461    Name: Matthias Blume
2462    Date: 2001/05/25 15:30:00 EDT
2463    Tag: blume-20010525-iptr
2464    Description:
2465    
2466      - put the official 110.33-README (as it appears on the ftp server) under
2467        CVS
2468      - fixed a small bug related to incomplete pointer types in
2469        ml-nlffigen
2470      - small cosmetic change to the ml-nlffi-lib's "arr" type constructor
2471        (it does not need the 'f type parameter)
2472    
2473    ----------------------------------------------------------------------
2474    Name: Matthias Blume
2475    Date: 2001/05/23 14:30:00 EDT
2476    Tag: Release_110_33
2477    Description:
2478    
2479    New version number (110.33).  New bootfiles.
2480    
2481    ----------------------------------------------------------------------
2482    Name: Matthias Blume
2483    Date: 2001/05/22 18:06:00 EDT
2484    Tag: blume-20010522-targets
2485    Description:
2486    
2487    Made install.sh use file config/targets.customized if it exists, falling
2488    back to config/targets if it doesn't.  This way one can have a customized
2489    version of the targets file without touching the "real thing", thus
2490    eliminating the constant fear of accidentally checking something bogus
2491    back into the CVS repository...  (File config/targets.customized must
2492    not be added to the repository!)
2493    
2494    ----------------------------------------------------------------------
2495    Name: Matthias Blume
2496    Date: 2001/05/22 16:30:00 EDT
2497    Tag: blume-20010522-minitut
2498    Description:
2499    
2500    1. Bug fix in ml-nlffigen; now (hopefully) correctly handling
2501       struct returns.
2502    2. Added src/ml-nlffi-lib/Doc/mini-tutorial.txt.  This is some very
2503       incomplete, preliminary documentation for NLFFI.
2504    
2505    ----------------------------------------------------------------------
2506    Name: Matthias Blume
2507    Date: 2001/05/14 11:30:00 EDT
2508    Tag: blume-20010514-script
2509    Description:
2510    
2511    Some bugs in install script fixed.
2512    
2513    In addition to that I also made a slight change to the NLFFI API:
2514    Functors generated by ml-nlffigen now take the dynamic library as a
2515    straight functor argument, not as a suspended one.  (The original
2516    functor code used to force the suspension right away anyway, so there
2517    was nothing gained by this complication of the interface.)
2518    
2519    ----------------------------------------------------------------------
2520    Name: Matthias Blume
2521    Date: 2001/05/11 14:35:00 EDT
2522    Tag: blume-20010511-ml-nlffi
2523    Description:
2524    
2525    I finally took the plunge and added my new FFI code to the main
2526    repository.  For x86-linux it is now ready for prime-time.
2527    
2528    There are two new subdirectories of "src":
2529    
2530      - ml-nlffi-lib:
2531           The utility library for programs using the FFI interface.
2532           Here is the implementation of $/c.cm and its associated low-level
2533           partners $/c-int.cm and $/memory.cm.
2534      - ml-nlffigen:
2535           A stand-alone program for generating ML glue code from C source
2536           code.
2537    
2538    Building ml-nlffigen requires $/ckit-lib.cm.
2539    
2540    The config/install.sh script has been updates to do the Right Thing
2541    (hopefully).
2542    
2543    Notice that the source tree for the C-Kit will not be put under "src"
2544    but directly under the installation root directory.  (This is the
2545    structure that currently exists on the CVS server when you check out
2546    module "sml".)  Fortunately, config/install.sh knows about this oddity.
2547    
2548    Bugs: No documentation yet.
2549    
2550    ----------------------------------------------------------------------
2551    Name: Matthias Blume
2552    Date: 2001/05/09 16:35:00 EDT
2553    Tag: blume-20010509-cpscontract
2554    Description:
2555    
2556    Fixed a bug in the accounting code in cpsopt/contract.sml.  (The
2557    wrapper/unwrapper elimination did not decrement usage counts and some
2558    dead variables got overlooked by the dead-up logic.)
2559    
2560    ----------------------------------------------------------------------
2561    Name: Lal George
2562    Date: 2001/05/08  17:26:09 EDT
2563    Tag: george-20010508-omit-frameptr
2564    Description:
2565    
2566    Changes to implement the omit-frame-pointer optimization to support
2567    raw C calls. For now, there is only support on the Intel x86, but
2568    other architectures will follow as more experience is gained with this.
2569    
2570    
2571    ----------------------------------------------------------------------
2572    Name: Matthias Blume
2573    Date: 2001/05/07 14:40:00 EDT
2574    Tag: blume-20010507-proxies
2575    Description:
2576    
2577    I made into "proxy libraries" all libraries that qualify for such a
2578    change.  (A qualifying library is a library that has another library or
2579    groups as its sole member and repeats that member's export list
2580    verbatim.  A proxy library avoids this repetition by omitting its export
2581    list, effectively inheriting the list that its (only) member exports.
2582    See the CM manual for more explanation.)
2583    The main effect is that explicit export lists for these libraries
2584    do not have to be kepts in sync, making maintenance a bit easier.
2585    
2586    I also added copyright notices to many .cm-files.
2587    
2588    Last but not least, I made a new set of bootfiles.
2589    
2590    ----------------------------------------------------------------------
2591    Name: Matthias Blume
2592    Date: 2001/05/04 17:00:00 EDT
2593    Tag: blume-20010504-cm-lsplit
2594    Description:
2595    
2596    0. John merged pending changes to $/smlnj-lib.cm
2597    
2598    1. Allen's previous change accidentally backed out of one of Lal's
2599       earlier changes.  I undid this mistake (re-introducing Lal's change).
2600    
2601    2. I used the new topOrder' function from graph-scc.sml (from $/smlnj-lib.cm)
2602       within the compiler where applicable.  There is some code simplification
2603       because of that.
2604    
2605    3. The "split" phase (in FLINT) is now part of the default list of phases.
2606       Compiler.Control.LambdaSplitting.* can be used to globally control the
2607       lambda-splitting (cross-module-inlining) engine.  In addition to that,
2608       it can now also be controlled on a per-source basis: CM has been taught
2609       a new tool parameter applicable to ML source files.
2610    
2611       - To turn lambda-splitting off completely:
2612            local open Compiler.Control.LambdaSplitting in
2613                val _ = set Off
2614            end
2615       - To make "no lambda-splitting" the global default (but allow per-source
2616         overriding); this is the initial setting:
2617            local open Compiler.Control.LambdaSplitting in
2618                val _ = set (Default NONE)
2619            end
2620       - To make "lambda-splitting with aggressiveness a" the global default
2621         (and allow per-source overriding):
2622            local open Compiler.Control.LambdaSplitting in
2623                val _ = set (Default (SOME a))
2624            end
2625    
2626       - To turn lambda-splitting off for a given ML souce file (say: a.sml)
2627         write (in the respective .cm-file):
2628            a.sml (lambdasplitting:off)
2629       - To turn lambda-splitting for a.sml on with minimal aggressiveness:
2630            a.sml (lambdasplitting:on)
2631       - To turn lambda-splitting for a.sml on with aggressiveness <a> (where
2632         <a> is a decimal non-negative integer):
2633            a.sml (lambdasplitting:<a>)
2634       - To turn lambda-splitting for a.sml on with maximal aggressiveness:
2635            a.sml (lambdasplitting:infinity)
2636       - To use the global default for a.sml:
2637            a.sml (lambdasplitting:default)
2638         or simply
2639            a.sml
2640    
2641    ----------------------------------------------------------------------
2642    Name: Allen Leung
2643    Date: 2001/05/04 01:57:00 EDT
2644    Tag: leunga-20010504-sync
2645    Description:
2646    
2647      MLRISC features.
2648    
2649      1. Fix to CMPXCHG instructions.
2650      2. Changed RA interface to allow annotations in callbacks.
2651      3. Added a new method to the stream interface to allow annotations updates.
2652    
2653    ----------------------------------------------------------------------
2654    Name: Matthias Blume
2655    Date: 2001/05/01 11:45:00 EDT
2656    Tag: blume-20010501-pcedittmp
2657    Description:
2658    
2659    Changed install.sh to use the current working directory instead of
2660    /usr/tmp for a temporary file (pcedittmp).  The previous choice
2661    of /usr/tmp caused trouble with MacOS X because of file premission
2662    problems.
2663    
2664    ----------------------------------------------------------------------
2665    Name: Matthias Blume
2666    Date: 2001/04/20 11:10:00 EDT
2667    Tag: blume-20010420-inMLflag
2668    Description:
2669    
2670     - added vp_limitPtrMask to vproc-state.h
2671       (for use by the raw-C-calls mechanism to implement proper interrupt
2672        handling)
2673     - made the ML compiler aware of various data-structure offsets so it
2674       can generate code for accessing the vp_inML flag and vp_limitPtrMask
2675     - tweaked mlriscGen.sml to have it emit interrupt-handling code for
2676       raw C-calls
2677    
2678    ----------------------------------------------------------------------
2679    Name: Lal George
2680    Date: 2001/04/20 09:15:28 EDT
2681    Tag: george-20010420-macosX
2682    Description:
2683    
2684     - Changes to port to Mac OS X; Darwin.
2685    
2686     - In the process I found that sqrt was broken on the PPC, because the
2687       fsqrt instruction is not implemented.
2688    
2689    ----------------------------------------------------------------------
2690    Name: Matthias Blume
2691    Date: 2001/04/18 12:45:00 EDT
2692    Tag: blume-20010418-ccalls
2693    Description:
2694    
2695     - fixed two off-by-4 errors in the x86-specific c-calls implementation
2696       (this bug prevented structure arguments containing pointers from being
2697        passed correctly)
2698     - changed the raw-C-call code in mlriscGen.sml in such a way that
2699       structure arguments are represented as a pointer to the beginning
2700       of the structure (instead of having a series of synthesized arguments,
2701       one for each structure member)
2702    
2703     - made makeml script's verbosity level configurable via environment
2704       variable (MAKEML_VERBOSITY)
2705    
2706     - eliminated placeholder implementations for f32l, w16s, i16s, and f32s
2707       in rawmem-x86.sml; we are now using the real thing
2708    
2709    ----------------------------------------------------------------------
2710    Name: Matthias Blume
2711    Date: 2001/03/22 16:25:00 EST
2712    Tag: blume-20010322-bootfiles
2713    Description:
2714    
2715    Created a new set of bootfiles (for your automatic installation convenience).
2716    
2717    ----------------------------------------------------------------------
2718    Name: Matthias Blume
2719    Date: 2001/03/22 15:10:00 EST
2720    Tag: blume-20010322-rawmem-parcm
2721    Description:
2722    
2723    1. All "raw memory access" primitives for the new FFI are implemented now
2724       (at least on the x86).
2725    2. Some further cleanup of CM's parallel make mechanism.
2726    
2727    ----------------------------------------------------------------------
2728    Name: Matthias Blume
2729    Date: 2001/03/19 17:53:00 EST
2730    Tag: blume-20010319-parallel
2731    Description:
2732    
2733    Parallel make (using compile servers) now works again.
2734    
2735    To this end, CM.stabilize and CMB.make have been modified to work in
2736    two passes when compile servers are attached:
2737       1. Compile everything, do not perform stabilization; this pass
2738          uses compile servers
2739       2. Stabilize everything; this pass does not use compile servers
2740    If there are no compile servers, the two passes are combined into one
2741    (as before).  Splitting the passes increases the inherent parallelism
2742    in the dependency graph because the entire graph including all
2743    libraries is available at the same time.  This, in turn, improves
2744    server utilization.  The downside is that the master process will
2745    have to do some extra work after compilation is done (because for
2746    technical reasons it must re-read all the binfiles during stabilization).
2747    
2748    ----------------------------------------------------------------------
2749    Name: Matthias Blume
2750    Date: 2001/03/16 12:22:00 EST
2751    Tag: blume-20010316-bootfiles
2752    Description:
2753    
2754    Created a new set of bootfiles (for your automatic installation convenience).
2755    
2756    ----------------------------------------------------------------------
2757    Name: Matthias Blume
2758    Date: 2001/03/16 11:00:00 EST
2759    Tag: blume-20010316-MLTREE-fixup
2760    Description:
2761    
2762    This is a minor fixup for an (untagged) earlier commit by Allen.
2763    (A file was missing).
2764    
2765    ----------------------------------------------------------------------
2766    Name: Allen Leung
2767    Date: Mon Mar  5 18:54:57 EST 2001
2768    Tag: leunga-20010305-cut-support
2769    
2770    1. New support for alternative control-flow in MLTREE.
2771       Currently we support
2772    
2773          FLOW_TO(CALL ...., [k1,...,kn])
2774    
2775       This is needed for 'cuts to' in C-- and try/handle-like constructs
2776       in Moby
2777    
2778       New assembler flag "asm-show-cutsto" to turn on control-flow debugging.
2779    
2780    2. Register Allocator
2781    
2782       Changes in interface [from Fermin, John]
2783    
2784    3. Alpha 8-bit SLL support [Fermin]
2785    
2786    4. All architectures
2787    
2788       A new module (ClusterExpandCopies) for expanding parallel copies.
2789    
2790    ----------------------------------------------------------------------
2791    Name: Allen Leung
2792    Date: 2001/02/27 23:07:00 EST
2793    Tag: leunga-20010227-minor-stuff
2794    
2795    1. Alpha bug fix for CMOVNE
2796    2. Handle mltree COND(..,FCMP ...,...)
2797    3. Bug fix in simplifier
2798    
2799    ----------------------------------------------------------------------
2800    Name: Matthias Blume
2801    Date: 2001/01/30 17:50:00 EST
2802    Tag: blume-20010130-sync
2803    Description:
2804    
2805    This is just a minor update to sync my devel branch with the main brach.
2806    The only visible change is the addition of some README files.
2807    
2808    ----------------------------------------------------------------------
2809    Name: Matthias Blume
2810    Date: 2001/01/12 23:30:00 JST
2811    Tag: blume-20010112-bootfiles
2812    Description:
2813    
2814    Made a new set of bootfiles that goes with the current state of the
2815    repository.
2816    
2817    ----------------------------------------------------------------------
2818    Name: Matthias Blume
2819    Date: 2001/01/12 21:20:00 JST
2820    Tag: blume-20010112-sync
2821    Description:
2822    
2823    I am just flushing out some minor changes that had accumulated in
2824    my private branch in order to sync with the main tree.  (This is
2825    mainly because I had CVS trouble when trying to merge _into_ my
2826    private branch.)
2827    
2828    Most people should be completely unaffected by this.
2829    
2830    ----------------------------------------------------------------------
2831    Name: Allen Leung
2832    Date: Thu Jan 11 21:03:00 EST 2001
2833    Tag: leunga-20010111-labexp=mltree
2834    Description:
2835    
2836    1.  Removed the type LabelExp and replace it by MLTree.
2837    2.  Rewritten mltree-simplify with the pattern matcher tool.
2838    3.  There were some bugs in alpha code generator which would break
2839        64-bit code generation.
2840    4.  Redo the tools to generate code with the
2841    5.  The CM files in MLRISC (and in src/system/smlnj/MLRISC)
2842        are now generated by perl scripts.
2843    
2844    ----------------------------------------------------------------------
2845    Name: Matthias Blume
2846    Date: 2001/01/10 21:55:00 JST
2847    Tag: blume-20010110-rcc
2848    Description:
2849    
2850    The RCC stuff now seems to work (but only on the x86).
2851    This required hacking of the c-calls interface (and -implementation) in
2852    MLRISC.
2853    
2854    Normal compiler users should be unaffected.
2855    
2856  ----------------------------------------------------------------------  ----------------------------------------------------------------------
2857  Name: Matthias Blume  Name: Matthias Blume
2858  Date: 2001/01/09 01:20:00 JST  Date: 2001/01/09 01:20:00 JST
# Line 1894  Line 4736 
4736       elaborator). There were a lot of changes during my "linkpath" trials       elaborator). There were a lot of changes during my "linkpath" trials
4737       that could have been reverted to their original state but weren't.       that could have been reverted to their original state but weren't.
4738       Please, don't be too harsh on me for messing with this code a bit more       Please, don't be too harsh on me for messing with this code a bit more
4739       than what was strictly necessary...  (I _did_ resist the tempation       than what was strictly necessary...  (I _did_ resist the temptation
4740       of doing any "global reformatting" to avoid an untimely death at       of doing any "global reformatting" to avoid an untimely death at
4741       Dave's hands. :)       Dave's hands. :)
4742    

Legend:
Removed from v.773  
changed lines
  Added in v.1178

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