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 961, Tue Oct 16 15:32:39 2001 UTC revision 1016, Tue Jan 15 23:10:06 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: Lal George
16    Date: 2002/01/15
17    Tag: <none>
18    Description:
19    
20    1. Since COPY instructions are no longer native to the architecture,
21       a generic functor can be used to implement the expandCopies function.
22    
23    2. Allowed EXPORT and IMPORT pseudo-op declarations to appear inside a
24       TEXT segment.
25    
26    ----------------------------------------------------------------------
27    Name: Matthias Blume
28    Date: 2002/01/15
29    Tag: blume-20020115-ffiupdates
30    Description:
31    
32    1. Fix for bug resulting in single-precision float values being returned
33       incorrectly from FFI calls.
34    
35    2. Small modifications to C FFI API:
36    
37        - memory-allocation routines return straight objects (no options)
38          and raise an exception in out-of-memory situations
39        - unsafe extensions to cast between function pointers and pointers
40          from/to ints
41        - added structure C_Debug as an alternative to structure C where
42          pointer-dereferencing (|*| and |*!) always check for null-pointers
43        - added open_lib' to DynLinkage;  open_lib' works like open_lib
44          but also takes a (possibly empty) list of existing library handles
45          that the current library depends on
46    
47    ----------------------------------------------------------------------
48    Name: Matthias Blume
49    Date: 2002/01/10
50    Tag: blume-20020110-newffigen
51    Description:
52    
53    1. Updates to portable graph code.
54    
55    2. Major update to ml-nlffigen and ml-nlffi-lib.  Things are much
56       more scalable now so that even huge interfaces such as the one
57       for GTK compile in finite time and space. :-)
58       See src/ml-nlffigen/README for details on what's new.
59    
60    ----------------------------------------------------------------------
61    Name: Lal George
62    Date: 2001/01/09 14:31:35 EST 2002
63    Tag: george-20011206-rm-native-copy
64    Description:
65    
66            Removed the native COPY and FCOPY instructions
67            from all the architectures and replaced it with the
68            explicit COPY instruction from the previous commit.
69    
70            It is now possible to simplify many of the optimizations
71            modules that manipulate copies. This has not been
72            done in this change.
73    
74    ----------------------------------------------------------------------
75    Name: Lal George
76    Date: 2001/12/06 16:50:13 EST 2001
77    Tag: george-20011206-mlrisc-instruction
78    Description:
79    
80    Changed the representation of instructions from being fully abstract
81    to being partially concrete. That is to say:
82    
83      from
84            type instruction
85    
86      to
87            type instr                              (* machine instruction *)
88    
89            datatype instruction =
90                LIVE of {regs: C.cellset, spilled: C.cellset}
91              | KILL of {regs: C.cellset, spilled: C.cellset}
92              | COPYXXX of {k: CB.cellkind, dst: CB.cell list, src: CB.cell list}
93              | ANNOTATION of {i: instruction, a: Annotations.annotation}
94              | INSTR of instr
95    
96    This makes the handling of certain special instructions that appear on
97    all architectures easier and uniform.
98    
99    LIVE and KILL say that a list of registers are live or killed at the
100    program point where they appear. No spill code is generated when an
101    element of the 'regs' field is spilled, but the register is moved to
102    the 'spilled' (which is present, more for debugging than anything else).
103    
104    LIVE replaces the (now deprecated) DEFFREG instruction on the alpha.
105    We used to generate:
106    
107            DEFFREG f1
108            f1 := f2 + f3
109            trapb
110    
111    but now generate:
112    
113            f1 := f2 + f3
114            trapb
115            LIVE {regs=[f1,f2,f3], spilled=[]}
116    
117    Furthermore, the DEFFREG (hack) required that all floating point instruction
118    use all registers mentioned in the instruction. Therefore f1 := f2 + f3,
119    defines f1 and uses [f1,f2,f3]! This hack is no longer required resulting
120    in a cleaner alpha implementation. (Hopefully, intel will not get rid of
121    this architecture).
122    
123    COPYXXX is intended to replace the parallel COPY and FCOPY  available on
124    all the architectures. This will result in further simplification of the
125    register allocator that must be aware of them for coalescing purposes, and
126    will also simplify certain aspects of the machine description that provides
127    callbacks related to parallel copies.
128    
129    ANNOTATION should be obvious, and now INSTR represents the honest to God
130    machine instruction set!
131    
132    The <arch>/instructions/<arch>Instr.sml files define certain utility
133    functions for making porting easier -- essentially converting upper case
134    to lower case. All machine instructions (of type instr) are in upper case,
135    and the lower case form generates an MLRISC instruction. For example on
136    the alpha we have:
137    
138      datatype instr =
139         LDA of {r:cell, b:cell, d:operand}
140       | ...
141    
142      val lda : {r:cell, b:cell, d:operand} -> instruction
143        ...
144    
145    where lda is just (INSTR o LDA), etc.
146    
147    ----------------------------------------------------------------------
148    Name: Matthias Blume
149    Date: 2001/11/22 21:40:00 EST
150    Tag: Release_110_37
151    Description:
152    
153    Release 110.37.  This time for real.
154    
155    ----------------------------------------------------------------------
156    Name: Matthias Blume
157    Date: 2001/11/21 16:35:00 EST
158    Tag: blume-20011121-foot-in-mouth
159    Description:
160    
161    Removed the "Release_110_37" tag because of a serious bug.
162    This will be re-tagged once the bug is fixed.
163    
164    ----------------------------------------------------------------------
165    Name: Matthias Blume
166    Date: 2001/11/21 16:14:00 EST
167    Tag: blume-20011121-forgottenfile
168    Description:
169    
170    Forgot to add a file.  (Just a .tex-file -- part of
171    the CM manual source.)
172    
173    ----------------------------------------------------------------------
174    Name: Matthias Blume
175    Date: 2001/11/21 16:10:00 EST
176    Tag: blume-20011121-invalid_110_37
177    Description:
178    
179    Note: I removed the original tag "Release_110_37" from this commit
180          because we found a serious bug in all non-x86 backends.
181          - Matthias
182    
183    1. Modifications to the SML/NJ code generator and to the runtime system
184       so that code object name strings are directly inserted into code
185       objects at code generation time.  The only business the runtime system
186       has with this is now to read the name strings on occasions.
187       (The encoding of the name string has also changed somewhat.)
188    
189    2. CM now implements a simple "set calculus" for specifying export lists.
190       In particular, it is now possible to refer to the export lists of
191       other libraries/groups/sources and form unions as well as differences.
192       See the latest CM manual for details.
193    
194    3. An separate notion of "proxy" libraries has again be eliminated from
195       CM's model.  (Proxy libraries are now simply a special case of using
196       the export list calculus.)
197    
198    4. Some of the existing libraries now take advantage of the new set
199       calculus.
200       (Notice that not all libraries have been converted because some
201       of the existing .cm-files are supposed to be backward compatible
202       with 110.0.x.)
203    
204    5. Some cleanup in stand-alone programs.  (Don't use "exnMessage" -- use
205       "General.exnMessage"!  The former relies on a certain hook to be
206       initialized, and that often does not happen in the stand-alone case.)
207    
208    ----------------------------------------------------------------------
209    Name: Lal George
210    Date: 2001/11/21  13:56:18 EST
211    Tag: george-2001121-pseudo-ops
212    Description:
213    
214      Implemented a complete redesign of MLRISC pseudo-ops. Now there
215      ought to never be any question of incompatabilities with
216      pseudo-op syntax expected by host assemblers.
217    
218      For now, only modules supporting GAS syntax are implemented
219      but more should follow, such as MASM, and vendor assembler
220      syntax, e.g. IBM as, Sun as, etc.
221    
222    ----------------------------------------------------------------------
223    Name: Matthias Blume
224    Date: 2001/11/14 11:52:00 EST
225    Tag: blume-20011114-srcname
226    Description:
227    
228    1. Routed the name of the current source file to mlriscgen where it
229       should be directly emitted into the code object.  (This last part
230       is yet to be done.)
231    
232    2. Some cleanup of the pgraph code to make it match the proposal that
233       I put out the other day.  (The proposal notwithstanding, things are
234       still in flux here.)
235    
236    ----------------------------------------------------------------------
237    Name: Lal George
238    Date: 2001/11/14 09:44:04 EST
239    Tag:
240    Description:
241    
242      Fix for a backpatching bug reported by Allen.
243    
244      Because the boundary between short and long span-dependent
245      instructions is +/- 128, there are an astounding number of
246      span-dependent instructions whose size is over estimated.
247    
248      Allen came up with the idea of letting the size of span
249      dependent instructions be non-monotonic, for a maxIter
250      number of times, after which the size must be monotonically
251      increasing.
252    
253      This table shows the number of span-dependent instructions
254      whose size was over-estimated as a function of maxIter, for the
255      file Parse/parse/ml.grm.sml:
256    
257         maxIter            # of instructions:
258            10                      687
259            20                      438
260            30                      198
261            40                        0
262    
263      In compiling the compiler, there is no significant difference in
264      compilation speed between maxIter=10 and maxIter=40. Actually,
265      my measurements showed that maxIter=40 was a tad faster than
266      maxIter=10! Also 96% of the  files in the compiler reach a fix
267      point within 13 iterations, so fixing maxIter at 40, while high,
268      is okay.
269    
270    ----------------------------------------------------------------------
271    Name: Matthias Blume
272    Date: 2001/10/31 15:25:00 EST
273    Tag: blume-20011031-pgraph
274    Description:
275    
276    CKIT:
277    * Changed the "Function" constructor of type Ast.ctype to carry optional
278      argument identifiers.
279    * Changed the return type of TypeUtil.getFunction accordingly.
280    * Type equality ignores the argument names.
281    * TypeUtil.composite tries to preserve argument names but gives up quickly
282      if there is a mismatch.
283    
284    installation script:
285    * attempts to use "curl" if available (unless "wget" is available as well)
286    
287    CM:
288    * has an experimental implementation of "portable graphs" which I will
289      soon propose as an implementation-independent library format
290    * there are also new libraries $/pgraph.cm and $/pgraph-util.cm
291    
292    NLFFI-LIB:
293    * some cleanup (all cosmetic)
294    
295    NLFFIGEN:
296    * temporarily disabled the mechanism that suppresses ML output for
297      C definitions whose identifiers start with an underscore character
298    * generate val bindings for enum constants
299    * user can request that only one style (light or heavy) is being used;
300      default is to use both (command-line arguments: -heavy and -light)
301    * fixed bug in handling of function types involving incomplete pointers
302    * generate ML entry points that take record arguments (i.e., using
303      named arguments) for C functions that have a prototype with named
304      arguments
305      (see changes to CKIT)
306    
307    ----------------------------------------------------------------------
308    Name: Allen Leung
309    Date: 2001/10/27 20:34:00 EDT
310    Tag: leunga-20011027-x86-fast-fp-call
311    Description:
312    
313       Fixed the bug described in blume-20010920-slowfp.
314    
315       The fix involves
316          1. generating FCOPYs in FSTP in ia32-svid
317          2. marking a CALL with the appropriate annotation
318    
319  ----------------------------------------------------------------------  ----------------------------------------------------------------------
320  Name: Matthias Blume  Name: Matthias Blume

Legend:
Removed from v.961  
changed lines
  Added in v.1016

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