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 972, Sun Oct 28 00:37:02 2001 UTC revision 1029, Fri Jan 18 21:42:35 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: 2002/01/18
18    Tag: blume-20020118-more-ffifiddle
19    Description:
20    
21    Today's FFI fiddling:
22    
23      - Provided a structure CGetSet with "convenient" versions of C.Get.* and
24        C.Set.* that use concrete (MLRep.*) arguments and results instead
25        of abstract ones.
26    
27      - Provided word-style bit operations etc. for "int" representation
28        types in MLRep.S<Foo>Bitops where <Foo> ranges over Char, Int, Short,
29        and Long.
30    
31    ----------------------------------------------------------------------
32    Name: Matthias Blume
33    Date: 2002/01/18
34    Tag: blume-20020118-use-x86-fp
35    Description:
36    
37    Now that x86-fast-fp seems to be working, I turned it back on again
38    by default.  (Seems to work fine now, even with the FFI.)
39    
40    Other than that, I added some documentation about the FFI to
41    src/ml-nlffigen/README and updated the FFI test examples in
42    src/ml-nlffi-lib/Tests/*.
43    
44    ----------------------------------------------------------------------
45    Name: Allen Leung
46    Date: 2002/01/17
47    Tag: leunga-20020117-x86-fast-fp-call
48    Description:
49    
50       1. Fixed a problem with handling return fp values when x86's fast fp
51          mode is turned on.
52    
53       2. Minor pretty printing fix for cellset.  Print %st(0) as %st(0) instead
54          of %f32.
55    
56       3. Added a constructor INT32lit to the ast of MLRISC tools.
57    
58    ----------------------------------------------------------------------
59    Name: Matthias Blume
60    Date: 2002/01/16
61    Tag: blume-20020116-ffifiddle
62    Description:
63    
64    More fiddling with the FFI interface:
65    
66     - Make constness 'c instead of rw wherever possible.  This eliminates
67       the need for certain explicit coercions.  (However, due to ML's
68       value polymorphism, there will still be many cases where explicit
69       coercions are necessary.  Phantom types are not the whole answer
70       to modeling a subtyping relationship in ML.)
71    
72     - ro/rw coersions for pointers added.  (Avoids the detour through */&.)
73    
74     - "printf" test example added to src/ml-nlffi-lib/Tests.  (Demonstrates
75       clumsy workaround for varargs problem.)
76    
77    ----------------------------------------------------------------------
78    Name: Lal George
79    Date: 2002/01/15
80    Tag: <none>
81    Description:
82    
83    1. Since COPY instructions are no longer native to the architecture,
84       a generic functor can be used to implement the expandCopies function.
85    
86    2. Allowed EXPORT and IMPORT pseudo-op declarations to appear inside a
87       TEXT segment.
88    
89    ----------------------------------------------------------------------
90    Name: Matthias Blume
91    Date: 2002/01/15
92    Tag: blume-20020115-ffiupdates
93    Description:
94    
95    1. Fix for bug resulting in single-precision float values being returned
96       incorrectly from FFI calls.
97    
98    2. Small modifications to C FFI API:
99    
100        - memory-allocation routines return straight objects (no options)
101          and raise an exception in out-of-memory situations
102        - unsafe extensions to cast between function pointers and pointers
103          from/to ints
104        - added structure C_Debug as an alternative to structure C where
105          pointer-dereferencing (|*| and |*!) always check for null-pointers
106        - added open_lib' to DynLinkage;  open_lib' works like open_lib
107          but also takes a (possibly empty) list of existing library handles
108          that the current library depends on
109    
110    ----------------------------------------------------------------------
111    Name: Matthias Blume
112    Date: 2002/01/10
113    Tag: blume-20020110-newffigen
114    Description:
115    
116    1. Updates to portable graph code.
117    
118    2. Major update to ml-nlffigen and ml-nlffi-lib.  Things are much
119       more scalable now so that even huge interfaces such as the one
120       for GTK compile in finite time and space. :-)
121       See src/ml-nlffigen/README for details on what's new.
122    
123    ----------------------------------------------------------------------
124    Name: Lal George
125    Date: 2001/01/09 14:31:35 EST 2002
126    Tag: george-20011206-rm-native-copy
127    Description:
128    
129            Removed the native COPY and FCOPY instructions
130            from all the architectures and replaced it with the
131            explicit COPY instruction from the previous commit.
132    
133            It is now possible to simplify many of the optimizations
134            modules that manipulate copies. This has not been
135            done in this change.
136    
137    ----------------------------------------------------------------------
138    Name: Lal George
139    Date: 2001/12/06 16:50:13 EST 2001
140    Tag: george-20011206-mlrisc-instruction
141    Description:
142    
143    Changed the representation of instructions from being fully abstract
144    to being partially concrete. That is to say:
145    
146      from
147            type instruction
148    
149      to
150            type instr                              (* machine instruction *)
151    
152            datatype instruction =
153                LIVE of {regs: C.cellset, spilled: C.cellset}
154              | KILL of {regs: C.cellset, spilled: C.cellset}
155              | COPYXXX of {k: CB.cellkind, dst: CB.cell list, src: CB.cell list}
156              | ANNOTATION of {i: instruction, a: Annotations.annotation}
157              | INSTR of instr
158    
159    This makes the handling of certain special instructions that appear on
160    all architectures easier and uniform.
161    
162    LIVE and KILL say that a list of registers are live or killed at the
163    program point where they appear. No spill code is generated when an
164    element of the 'regs' field is spilled, but the register is moved to
165    the 'spilled' (which is present, more for debugging than anything else).
166    
167    LIVE replaces the (now deprecated) DEFFREG instruction on the alpha.
168    We used to generate:
169    
170            DEFFREG f1
171            f1 := f2 + f3
172            trapb
173    
174    but now generate:
175    
176            f1 := f2 + f3
177            trapb
178            LIVE {regs=[f1,f2,f3], spilled=[]}
179    
180    Furthermore, the DEFFREG (hack) required that all floating point instruction
181    use all registers mentioned in the instruction. Therefore f1 := f2 + f3,
182    defines f1 and uses [f1,f2,f3]! This hack is no longer required resulting
183    in a cleaner alpha implementation. (Hopefully, intel will not get rid of
184    this architecture).
185    
186    COPYXXX is intended to replace the parallel COPY and FCOPY  available on
187    all the architectures. This will result in further simplification of the
188    register allocator that must be aware of them for coalescing purposes, and
189    will also simplify certain aspects of the machine description that provides
190    callbacks related to parallel copies.
191    
192    ANNOTATION should be obvious, and now INSTR represents the honest to God
193    machine instruction set!
194    
195    The <arch>/instructions/<arch>Instr.sml files define certain utility
196    functions for making porting easier -- essentially converting upper case
197    to lower case. All machine instructions (of type instr) are in upper case,
198    and the lower case form generates an MLRISC instruction. For example on
199    the alpha we have:
200    
201      datatype instr =
202         LDA of {r:cell, b:cell, d:operand}
203       | ...
204    
205      val lda : {r:cell, b:cell, d:operand} -> instruction
206        ...
207    
208    where lda is just (INSTR o LDA), etc.
209    
210    ----------------------------------------------------------------------
211    Name: Matthias Blume
212    Date: 2001/11/22 21:40:00 EST
213    Tag: Release_110_37
214    Description:
215    
216    Release 110.37.  This time for real.
217    
218    ----------------------------------------------------------------------
219    Name: Matthias Blume
220    Date: 2001/11/21 16:35:00 EST
221    Tag: blume-20011121-foot-in-mouth
222    Description:
223    
224    Removed the "Release_110_37" tag because of a serious bug.
225    This will be re-tagged once the bug is fixed.
226    
227    ----------------------------------------------------------------------
228    Name: Matthias Blume
229    Date: 2001/11/21 16:14:00 EST
230    Tag: blume-20011121-forgottenfile
231    Description:
232    
233    Forgot to add a file.  (Just a .tex-file -- part of
234    the CM manual source.)
235    
236    ----------------------------------------------------------------------
237    Name: Matthias Blume
238    Date: 2001/11/21 16:10:00 EST
239    Tag: blume-20011121-invalid_110_37
240    Description:
241    
242    Note: I removed the original tag "Release_110_37" from this commit
243          because we found a serious bug in all non-x86 backends.
244          - Matthias
245    
246    1. Modifications to the SML/NJ code generator and to the runtime system
247       so that code object name strings are directly inserted into code
248       objects at code generation time.  The only business the runtime system
249       has with this is now to read the name strings on occasions.
250       (The encoding of the name string has also changed somewhat.)
251    
252    2. CM now implements a simple "set calculus" for specifying export lists.
253       In particular, it is now possible to refer to the export lists of
254       other libraries/groups/sources and form unions as well as differences.
255       See the latest CM manual for details.
256    
257    3. An separate notion of "proxy" libraries has again be eliminated from
258       CM's model.  (Proxy libraries are now simply a special case of using
259       the export list calculus.)
260    
261    4. Some of the existing libraries now take advantage of the new set
262       calculus.
263       (Notice that not all libraries have been converted because some
264       of the existing .cm-files are supposed to be backward compatible
265       with 110.0.x.)
266    
267    5. Some cleanup in stand-alone programs.  (Don't use "exnMessage" -- use
268       "General.exnMessage"!  The former relies on a certain hook to be
269       initialized, and that often does not happen in the stand-alone case.)
270    
271    ----------------------------------------------------------------------
272    Name: Lal George
273    Date: 2001/11/21  13:56:18 EST
274    Tag: george-2001121-pseudo-ops
275    Description:
276    
277      Implemented a complete redesign of MLRISC pseudo-ops. Now there
278      ought to never be any question of incompatabilities with
279      pseudo-op syntax expected by host assemblers.
280    
281      For now, only modules supporting GAS syntax are implemented
282      but more should follow, such as MASM, and vendor assembler
283      syntax, e.g. IBM as, Sun as, etc.
284    
285    ----------------------------------------------------------------------
286    Name: Matthias Blume
287    Date: 2001/11/14 11:52:00 EST
288    Tag: blume-20011114-srcname
289    Description:
290    
291    1. Routed the name of the current source file to mlriscgen where it
292       should be directly emitted into the code object.  (This last part
293       is yet to be done.)
294    
295    2. Some cleanup of the pgraph code to make it match the proposal that
296       I put out the other day.  (The proposal notwithstanding, things are
297       still in flux here.)
298    
299    ----------------------------------------------------------------------
300    Name: Lal George
301    Date: 2001/11/14 09:44:04 EST
302    Tag:
303    Description:
304    
305      Fix for a backpatching bug reported by Allen.
306    
307      Because the boundary between short and long span-dependent
308      instructions is +/- 128, there are an astounding number of
309      span-dependent instructions whose size is over estimated.
310    
311      Allen came up with the idea of letting the size of span
312      dependent instructions be non-monotonic, for a maxIter
313      number of times, after which the size must be monotonically
314      increasing.
315    
316      This table shows the number of span-dependent instructions
317      whose size was over-estimated as a function of maxIter, for the
318      file Parse/parse/ml.grm.sml:
319    
320         maxIter            # of instructions:
321            10                      687
322            20                      438
323            30                      198
324            40                        0
325    
326      In compiling the compiler, there is no significant difference in
327      compilation speed between maxIter=10 and maxIter=40. Actually,
328      my measurements showed that maxIter=40 was a tad faster than
329      maxIter=10! Also 96% of the  files in the compiler reach a fix
330      point within 13 iterations, so fixing maxIter at 40, while high,
331      is okay.
332    
333    ----------------------------------------------------------------------
334    Name: Matthias Blume
335    Date: 2001/10/31 15:25:00 EST
336    Tag: blume-20011031-pgraph
337    Description:
338    
339    CKIT:
340    * Changed the "Function" constructor of type Ast.ctype to carry optional
341      argument identifiers.
342    * Changed the return type of TypeUtil.getFunction accordingly.
343    * Type equality ignores the argument names.
344    * TypeUtil.composite tries to preserve argument names but gives up quickly
345      if there is a mismatch.
346    
347    installation script:
348    * attempts to use "curl" if available (unless "wget" is available as well)
349    
350    CM:
351    * has an experimental implementation of "portable graphs" which I will
352      soon propose as an implementation-independent library format
353    * there are also new libraries $/pgraph.cm and $/pgraph-util.cm
354    
355    NLFFI-LIB:
356    * some cleanup (all cosmetic)
357    
358    NLFFIGEN:
359    * temporarily disabled the mechanism that suppresses ML output for
360      C definitions whose identifiers start with an underscore character
361    * generate val bindings for enum constants
362    * user can request that only one style (light or heavy) is being used;
363      default is to use both (command-line arguments: -heavy and -light)
364    * fixed bug in handling of function types involving incomplete pointers
365    * generate ML entry points that take record arguments (i.e., using
366      named arguments) for C functions that have a prototype with named
367      arguments
368      (see changes to CKIT)
369    
370    ----------------------------------------------------------------------
371  Name: Allen Leung  Name: Allen Leung
372  Date: 2001/10/27 20:34:00 EDT  Date: 2001/10/27 20:34:00 EDT
373  Tag: leunga-20011027-x86-fast-fp-call  Tag: leunga-20011027-x86-fast-fp-call

Legend:
Removed from v.972  
changed lines
  Added in v.1029

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