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 1003, Fri Dec 7 02:45:32 2001 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: 2001/12/06 16:50:13 EST 2001
17    Tag: george-20011206-mlrisc-instruction
18    Description:
19    
20    Changed the representation of instructions from being fully abstract
21    to being partially concrete. That is to say:
22    
23      from
24            type instruction
25    
26      to
27            type instr                              (* machine instruction *)
28    
29            datatype instruction =
30                LIVE of {regs: C.cellset, spilled: C.cellset}
31              | KILL of {regs: C.cellset, spilled: C.cellset}
32              | COPYXXX of {k: CB.cellkind, dst: CB.cell list, src: CB.cell list}
33              | ANNOTATION of {i: instruction, a: Annotations.annotation}
34              | INSTR of instr
35    
36    This makes the handling of certain special instructions that appear on
37    all architectures easier and uniform.
38    
39    LIVE and KILL say that a list of registers are live or killed at the
40    program point where they appear. No spill code is generated when an
41    element of the 'regs' field is spilled, but the register is moved to
42    the 'spilled' (which is present, more for debugging than anything else).
43    
44    LIVE replaces the (now deprecated) DEFFREG instruction on the alpha.
45    We used to generate:
46    
47            DEFFREG f1
48            f1 := f2 + f3
49            trapb
50    
51    but now generate:
52    
53            f1 := f2 + f3
54            trapb
55            LIVE {regs=[f1,f2,f3], spilled=[]}
56    
57    Furthermore, the DEFFREG (hack) required that all floating point instruction
58    use all registers mentioned in the instruction. Therefore f1 := f2 + f3,
59    defines f1 and uses [f1,f2,f3]! This hack is no longer required resulting
60    in a cleaner alpha implementation. (Hopefully, intel will not get rid of
61    this architecture).
62    
63    COPYXXX is intended to replace the parallel COPY and FCOPY  available on
64    all the architectures. This will result in further simplification of the
65    register allocator that must be aware of them for coalescing purposes, and
66    will also simplify certain aspects of the machine description that provides
67    callbacks related to parallel copies.
68    
69    ANNOTATION should be obvious, and now INSTR represents the honest to God
70    machine instruction set!
71    
72    The <arch>/instructions/<arch>Instr.sml files define certain utility
73    functions for making porting easier -- essentially converting upper case
74    to lower case. All machine instructions (of type instr) are in upper case,
75    and the lower case form generates an MLRISC instruction. For example on
76    the alpha we have:
77    
78      datatype instr =
79         LDA of {r:cell, b:cell, d:operand}
80       | ...
81    
82      val lda : {r:cell, b:cell, d:operand} -> instruction
83        ...
84    
85    where lda is just (INSTR o LDA), etc.
86    
87    ----------------------------------------------------------------------
88    Name: Matthias Blume
89    Date: 2001/11/22 21:40:00 EST
90    Tag: Release_110_37
91    Description:
92    
93    Release 110.37.  This time for real.
94    
95    ----------------------------------------------------------------------
96    Name: Matthias Blume
97    Date: 2001/11/21 16:35:00 EST
98    Tag: blume-20011121-foot-in-mouth
99    Description:
100    
101    Removed the "Release_110_37" tag because of a serious bug.
102    This will be re-tagged once the bug is fixed.
103    
104    ----------------------------------------------------------------------
105    Name: Matthias Blume
106    Date: 2001/11/21 16:14:00 EST
107    Tag: blume-20011121-forgottenfile
108    Description:
109    
110    Forgot to add a file.  (Just a .tex-file -- part of
111    the CM manual source.)
112    
113    ----------------------------------------------------------------------
114    Name: Matthias Blume
115    Date: 2001/11/21 16:10:00 EST
116    Tag: blume-20011121-invalid_110_37
117    Description:
118    
119    Note: I removed the original tag "Release_110_37" from this commit
120          because we found a serious bug in all non-x86 backends.
121          - Matthias
122    
123    1. Modifications to the SML/NJ code generator and to the runtime system
124       so that code object name strings are directly inserted into code
125       objects at code generation time.  The only business the runtime system
126       has with this is now to read the name strings on occasions.
127       (The encoding of the name string has also changed somewhat.)
128    
129    2. CM now implements a simple "set calculus" for specifying export lists.
130       In particular, it is now possible to refer to the export lists of
131       other libraries/groups/sources and form unions as well as differences.
132       See the latest CM manual for details.
133    
134    3. An separate notion of "proxy" libraries has again be eliminated from
135       CM's model.  (Proxy libraries are now simply a special case of using
136       the export list calculus.)
137    
138    4. Some of the existing libraries now take advantage of the new set
139       calculus.
140       (Notice that not all libraries have been converted because some
141       of the existing .cm-files are supposed to be backward compatible
142       with 110.0.x.)
143    
144    5. Some cleanup in stand-alone programs.  (Don't use "exnMessage" -- use
145       "General.exnMessage"!  The former relies on a certain hook to be
146       initialized, and that often does not happen in the stand-alone case.)
147    
148    ----------------------------------------------------------------------
149    Name: Lal George
150    Date: 2001/11/21  13:56:18 EST
151    Tag: george-2001121-pseudo-ops
152    Description:
153    
154      Implemented a complete redesign of MLRISC pseudo-ops. Now there
155      ought to never be any question of incompatabilities with
156      pseudo-op syntax expected by host assemblers.
157    
158      For now, only modules supporting GAS syntax are implemented
159      but more should follow, such as MASM, and vendor assembler
160      syntax, e.g. IBM as, Sun as, etc.
161    
162    ----------------------------------------------------------------------
163    Name: Matthias Blume
164    Date: 2001/11/14 11:52:00 EST
165    Tag: blume-20011114-srcname
166    Description:
167    
168    1. Routed the name of the current source file to mlriscgen where it
169       should be directly emitted into the code object.  (This last part
170       is yet to be done.)
171    
172    2. Some cleanup of the pgraph code to make it match the proposal that
173       I put out the other day.  (The proposal notwithstanding, things are
174       still in flux here.)
175    
176    ----------------------------------------------------------------------
177    Name: Lal George
178    Date: 2001/11/14 09:44:04 EST
179    Tag:
180    Description:
181    
182      Fix for a backpatching bug reported by Allen.
183    
184      Because the boundary between short and long span-dependent
185      instructions is +/- 128, there are an astounding number of
186      span-dependent instructions whose size is over estimated.
187    
188      Allen came up with the idea of letting the size of span
189      dependent instructions be non-monotonic, for a maxIter
190      number of times, after which the size must be monotonically
191      increasing.
192    
193      This table shows the number of span-dependent instructions
194      whose size was over-estimated as a function of maxIter, for the
195      file Parse/parse/ml.grm.sml:
196    
197         maxIter            # of instructions:
198            10                      687
199            20                      438
200            30                      198
201            40                        0
202    
203      In compiling the compiler, there is no significant difference in
204      compilation speed between maxIter=10 and maxIter=40. Actually,
205      my measurements showed that maxIter=40 was a tad faster than
206      maxIter=10! Also 96% of the  files in the compiler reach a fix
207      point within 13 iterations, so fixing maxIter at 40, while high,
208      is okay.
209    
210    ----------------------------------------------------------------------
211    Name: Matthias Blume
212    Date: 2001/10/31 15:25:00 EST
213    Tag: blume-20011031-pgraph
214    Description:
215    
216    CKIT:
217    * Changed the "Function" constructor of type Ast.ctype to carry optional
218      argument identifiers.
219    * Changed the return type of TypeUtil.getFunction accordingly.
220    * Type equality ignores the argument names.
221    * TypeUtil.composite tries to preserve argument names but gives up quickly
222      if there is a mismatch.
223    
224    installation script:
225    * attempts to use "curl" if available (unless "wget" is available as well)
226    
227    CM:
228    * has an experimental implementation of "portable graphs" which I will
229      soon propose as an implementation-independent library format
230    * there are also new libraries $/pgraph.cm and $/pgraph-util.cm
231    
232    NLFFI-LIB:
233    * some cleanup (all cosmetic)
234    
235    NLFFIGEN:
236    * temporarily disabled the mechanism that suppresses ML output for
237      C definitions whose identifiers start with an underscore character
238    * generate val bindings for enum constants
239    * user can request that only one style (light or heavy) is being used;
240      default is to use both (command-line arguments: -heavy and -light)
241    * fixed bug in handling of function types involving incomplete pointers
242    * generate ML entry points that take record arguments (i.e., using
243      named arguments) for C functions that have a prototype with named
244      arguments
245      (see changes to CKIT)
246    
247    ----------------------------------------------------------------------
248    Name: Allen Leung
249    Date: 2001/10/27 20:34:00 EDT
250    Tag: leunga-20011027-x86-fast-fp-call
251    Description:
252    
253       Fixed the bug described in blume-20010920-slowfp.
254    
255       The fix involves
256          1. generating FCOPYs in FSTP in ia32-svid
257          2. marking a CALL with the appropriate annotation
258    
259  ----------------------------------------------------------------------  ----------------------------------------------------------------------
260  Name: Matthias Blume  Name: Matthias Blume

Legend:
Removed from v.961  
changed lines
  Added in v.1003

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