Diff of /sml/trunk/HISTORY

Parent Directory | Revision Log | Patch

	revision 972, Sun Oct 28 00:37:02 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	Name: Allen Leung

---

Colored Diff Long Colored Diff Unidiff Context Diff Side by Side

Legend: Removed from v.972   changed lines   Added in v.1003

---

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