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 1021, Wed Jan 16 20:33:57 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    Name: Matthias Blume
16    Date: 2002/01/16
17    Tag: blume-20020116-ffifiddle
18    Description:
19    
20    More fiddling with the FFI interface:
21    
22     - Make constness 'c instead of rw wherever possible.  This eliminates
23       the need for certain explicit coercions.  (However, due to ML's
24       value polymorphism, there will still be many cases where explicit
25       coercions are necessary.  Phantom types are not the whole answer
26       to modeling a subtyping relationship in ML.)
27    
28     - ro/rw coersions for pointers added.  (Avoids the detour through */&.)
29    
30     - "printf" test example added to src/ml-nlffi-lib/Tests.  (Demonstrates
31       clumsy workaround for varargs problem.)
32    
33    ----------------------------------------------------------------------
34    Name: Lal George
35    Date: 2002/01/15
36    Tag: <none>
37    Description:
38    
39    1. Since COPY instructions are no longer native to the architecture,
40       a generic functor can be used to implement the expandCopies function.
41    
42    2. Allowed EXPORT and IMPORT pseudo-op declarations to appear inside a
43       TEXT segment.
44    
45    ----------------------------------------------------------------------
46    Name: Matthias Blume
47    Date: 2002/01/15
48    Tag: blume-20020115-ffiupdates
49    Description:
50    
51    1. Fix for bug resulting in single-precision float values being returned
52       incorrectly from FFI calls.
53    
54    2. Small modifications to C FFI API:
55    
56        - memory-allocation routines return straight objects (no options)
57          and raise an exception in out-of-memory situations
58        - unsafe extensions to cast between function pointers and pointers
59          from/to ints
60        - added structure C_Debug as an alternative to structure C where
61          pointer-dereferencing (|*| and |*!) always check for null-pointers
62        - added open_lib' to DynLinkage;  open_lib' works like open_lib
63          but also takes a (possibly empty) list of existing library handles
64          that the current library depends on
65    
66    ----------------------------------------------------------------------
67    Name: Matthias Blume
68    Date: 2002/01/10
69    Tag: blume-20020110-newffigen
70    Description:
71    
72    1. Updates to portable graph code.
73    
74    2. Major update to ml-nlffigen and ml-nlffi-lib.  Things are much
75       more scalable now so that even huge interfaces such as the one
76       for GTK compile in finite time and space. :-)
77       See src/ml-nlffigen/README for details on what's new.
78    
79    ----------------------------------------------------------------------
80    Name: Lal George
81    Date: 2001/01/09 14:31:35 EST 2002
82    Tag: george-20011206-rm-native-copy
83    Description:
84    
85            Removed the native COPY and FCOPY instructions
86            from all the architectures and replaced it with the
87            explicit COPY instruction from the previous commit.
88    
89            It is now possible to simplify many of the optimizations
90            modules that manipulate copies. This has not been
91            done in this change.
92    
93    ----------------------------------------------------------------------
94    Name: Lal George
95    Date: 2001/12/06 16:50:13 EST 2001
96    Tag: george-20011206-mlrisc-instruction
97    Description:
98    
99    Changed the representation of instructions from being fully abstract
100    to being partially concrete. That is to say:
101    
102      from
103            type instruction
104    
105      to
106            type instr                              (* machine instruction *)
107    
108            datatype instruction =
109                LIVE of {regs: C.cellset, spilled: C.cellset}
110              | KILL of {regs: C.cellset, spilled: C.cellset}
111              | COPYXXX of {k: CB.cellkind, dst: CB.cell list, src: CB.cell list}
112              | ANNOTATION of {i: instruction, a: Annotations.annotation}
113              | INSTR of instr
114    
115    This makes the handling of certain special instructions that appear on
116    all architectures easier and uniform.
117    
118    LIVE and KILL say that a list of registers are live or killed at the
119    program point where they appear. No spill code is generated when an
120    element of the 'regs' field is spilled, but the register is moved to
121    the 'spilled' (which is present, more for debugging than anything else).
122    
123    LIVE replaces the (now deprecated) DEFFREG instruction on the alpha.
124    We used to generate:
125    
126            DEFFREG f1
127            f1 := f2 + f3
128            trapb
129    
130    but now generate:
131    
132            f1 := f2 + f3
133            trapb
134            LIVE {regs=[f1,f2,f3], spilled=[]}
135    
136    Furthermore, the DEFFREG (hack) required that all floating point instruction
137    use all registers mentioned in the instruction. Therefore f1 := f2 + f3,
138    defines f1 and uses [f1,f2,f3]! This hack is no longer required resulting
139    in a cleaner alpha implementation. (Hopefully, intel will not get rid of
140    this architecture).
141    
142    COPYXXX is intended to replace the parallel COPY and FCOPY  available on
143    all the architectures. This will result in further simplification of the
144    register allocator that must be aware of them for coalescing purposes, and
145    will also simplify certain aspects of the machine description that provides
146    callbacks related to parallel copies.
147    
148    ANNOTATION should be obvious, and now INSTR represents the honest to God
149    machine instruction set!
150    
151    The <arch>/instructions/<arch>Instr.sml files define certain utility
152    functions for making porting easier -- essentially converting upper case
153    to lower case. All machine instructions (of type instr) are in upper case,
154    and the lower case form generates an MLRISC instruction. For example on
155    the alpha we have:
156    
157      datatype instr =
158         LDA of {r:cell, b:cell, d:operand}
159       | ...
160    
161      val lda : {r:cell, b:cell, d:operand} -> instruction
162        ...
163    
164    where lda is just (INSTR o LDA), etc.
165    
166    ----------------------------------------------------------------------
167    Name: Matthias Blume
168    Date: 2001/11/22 21:40:00 EST
169    Tag: Release_110_37
170    Description:
171    
172    Release 110.37.  This time for real.
173    
174    ----------------------------------------------------------------------
175    Name: Matthias Blume
176    Date: 2001/11/21 16:35:00 EST
177    Tag: blume-20011121-foot-in-mouth
178    Description:
179    
180    Removed the "Release_110_37" tag because of a serious bug.
181    This will be re-tagged once the bug is fixed.
182    
183    ----------------------------------------------------------------------
184    Name: Matthias Blume
185    Date: 2001/11/21 16:14:00 EST
186    Tag: blume-20011121-forgottenfile
187    Description:
188    
189    Forgot to add a file.  (Just a .tex-file -- part of
190    the CM manual source.)
191    
192    ----------------------------------------------------------------------
193    Name: Matthias Blume
194    Date: 2001/11/21 16:10:00 EST
195    Tag: blume-20011121-invalid_110_37
196    Description:
197    
198    Note: I removed the original tag "Release_110_37" from this commit
199          because we found a serious bug in all non-x86 backends.
200          - Matthias
201    
202    1. Modifications to the SML/NJ code generator and to the runtime system
203       so that code object name strings are directly inserted into code
204       objects at code generation time.  The only business the runtime system
205       has with this is now to read the name strings on occasions.
206       (The encoding of the name string has also changed somewhat.)
207    
208    2. CM now implements a simple "set calculus" for specifying export lists.
209       In particular, it is now possible to refer to the export lists of
210       other libraries/groups/sources and form unions as well as differences.
211       See the latest CM manual for details.
212    
213    3. An separate notion of "proxy" libraries has again be eliminated from
214       CM's model.  (Proxy libraries are now simply a special case of using
215       the export list calculus.)
216    
217    4. Some of the existing libraries now take advantage of the new set
218       calculus.
219       (Notice that not all libraries have been converted because some
220       of the existing .cm-files are supposed to be backward compatible
221       with 110.0.x.)
222    
223    5. Some cleanup in stand-alone programs.  (Don't use "exnMessage" -- use
224       "General.exnMessage"!  The former relies on a certain hook to be
225       initialized, and that often does not happen in the stand-alone case.)
226    
227    ----------------------------------------------------------------------
228    Name: Lal George
229    Date: 2001/11/21  13:56:18 EST
230    Tag: george-2001121-pseudo-ops
231    Description:
232    
233      Implemented a complete redesign of MLRISC pseudo-ops. Now there
234      ought to never be any question of incompatabilities with
235      pseudo-op syntax expected by host assemblers.
236    
237      For now, only modules supporting GAS syntax are implemented
238      but more should follow, such as MASM, and vendor assembler
239      syntax, e.g. IBM as, Sun as, etc.
240    
241    ----------------------------------------------------------------------
242    Name: Matthias Blume
243    Date: 2001/11/14 11:52:00 EST
244    Tag: blume-20011114-srcname
245    Description:
246    
247    1. Routed the name of the current source file to mlriscgen where it
248       should be directly emitted into the code object.  (This last part
249       is yet to be done.)
250    
251    2. Some cleanup of the pgraph code to make it match the proposal that
252       I put out the other day.  (The proposal notwithstanding, things are
253       still in flux here.)
254    
255    ----------------------------------------------------------------------
256    Name: Lal George
257    Date: 2001/11/14 09:44:04 EST
258    Tag:
259    Description:
260    
261      Fix for a backpatching bug reported by Allen.
262    
263      Because the boundary between short and long span-dependent
264      instructions is +/- 128, there are an astounding number of
265      span-dependent instructions whose size is over estimated.
266    
267      Allen came up with the idea of letting the size of span
268      dependent instructions be non-monotonic, for a maxIter
269      number of times, after which the size must be monotonically
270      increasing.
271    
272      This table shows the number of span-dependent instructions
273      whose size was over-estimated as a function of maxIter, for the
274      file Parse/parse/ml.grm.sml:
275    
276         maxIter            # of instructions:
277            10                      687
278            20                      438
279            30                      198
280            40                        0
281    
282      In compiling the compiler, there is no significant difference in
283      compilation speed between maxIter=10 and maxIter=40. Actually,
284      my measurements showed that maxIter=40 was a tad faster than
285      maxIter=10! Also 96% of the  files in the compiler reach a fix
286      point within 13 iterations, so fixing maxIter at 40, while high,
287      is okay.
288    
289    ----------------------------------------------------------------------
290    Name: Matthias Blume
291    Date: 2001/10/31 15:25:00 EST
292    Tag: blume-20011031-pgraph
293    Description:
294    
295    CKIT:
296    * Changed the "Function" constructor of type Ast.ctype to carry optional
297      argument identifiers.
298    * Changed the return type of TypeUtil.getFunction accordingly.
299    * Type equality ignores the argument names.
300    * TypeUtil.composite tries to preserve argument names but gives up quickly
301      if there is a mismatch.
302    
303    installation script:
304    * attempts to use "curl" if available (unless "wget" is available as well)
305    
306    CM:
307    * has an experimental implementation of "portable graphs" which I will
308      soon propose as an implementation-independent library format
309    * there are also new libraries $/pgraph.cm and $/pgraph-util.cm
310    
311    NLFFI-LIB:
312    * some cleanup (all cosmetic)
313    
314    NLFFIGEN:
315    * temporarily disabled the mechanism that suppresses ML output for
316      C definitions whose identifiers start with an underscore character
317    * generate val bindings for enum constants
318    * user can request that only one style (light or heavy) is being used;
319      default is to use both (command-line arguments: -heavy and -light)
320    * fixed bug in handling of function types involving incomplete pointers
321    * generate ML entry points that take record arguments (i.e., using
322      named arguments) for C functions that have a prototype with named
323      arguments
324      (see changes to CKIT)
325    
326  ----------------------------------------------------------------------  ----------------------------------------------------------------------
327  Name: Allen Leung  Name: Allen Leung

Legend:
Removed from v.972  
changed lines
  Added in v.1021

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