Diff of /sml/trunk/HISTORY

Parent Directory | Revision Log | Patch

	revision 1137, Tue Mar 12 22:28:55 2002 UTC	revision 1174, Sat Mar 23 21:14:40 2002 UTC
#	Line 11	Line 11
11	Date: yyyy/mm/dd	Date: yyyy/mm/dd
12	Tag: <post-commit CVS tag>	Tag: <post-commit CVS tag>
13	Description:	Description:
14
15		----------------------------------------------------------------------
16		Name: Allen Leung
17		Date: 2002/03/23 15:50:00 EST
18		Tag: leunga-20020323-flint-cps-rcc-primops
19		Description:
20
21		1. Changes to FLINT primops:
22
23		(* make a call to a C-function;
24		* The primop carries C function prototype information and specifies
25		* which of its (ML-) arguments are floating point. C prototype
26		* information is for use by the backend, ML information is for
27		* use by the CPS converter. *)
28		| RAW_CCALL of { c_proto: CTypes.c_proto,
29		ml_args: ccall_type list,
30		ml_res_opt: ccall_type option,
31		reentrant : bool
32		} option
33		(* Allocate uninitialized storage on the heap.
34		* The record is meant to hold short-lived C objects, i.e., they
35		* are not ML pointers.  With the tag, the representation is
36		* the same as RECORD with tag tag_raw32 (sz=4), or tag_fblock (sz=8)
37		*)
38		| RAW_RECORD of {tag:bool,sz:int}
39		and ccall_type = CCALL_INT32 | CCALL_REAL64 | CCALL_ML_PTR
40
41		2.  These CPS primops are now overloaded:
42
43		rawload of {kind:numkind}
44		rawstore of {kind:numkind}
45
46		The one argument form is:
47
48		rawload {kind} address
49
50		The two argument form is:
51
52		rawload {kind} [ml object, byte-offset]
53
54		3. RAW_CCALL/RCC now takes two extra arguments:
55
56		a. The first is whether the C call is reentrant, i.e., whether
57		ML state should be saved and restored.
58		b. The second argument is a string argument specifying the name of
59		library and the C function.
60
61		These things are currently not handled in the code generator, yet.
62
63		4. In CProto,
64
65		An encoding type of "bool" means "ml object" and is mapped into
66		C prototype of PTR.  Note that "bool" is different than "string",
67		even though "string" is also mapped into PTR, because "bool"
68		is assigned an CPS type of BOGt, while "string" is assigned INT32t.
69
70		5. Pickler/unpicker
71
72		Changed to handle RAW_RECORD and newest RAW_CCALL
73
74		6. MLRiscGen,
75
76		1. Changed to handle the new rawload/rawstore/rawrecord operators.
77		2. Code for handling C Calls has been moved to a new module CPSCCalls,
78		in the file CodeGen/cpscompile/cps-c-calls.sml
79
80		7. Added the conditional move operator
81
82		condmove of branch
83
84		to cps.  Generation of this is still buggy so it is currently
85		disabled.
86
87		----------------------------------------------------------------------
88		Name: Lal George
89		Date: 2002/03/22 14:18:25 EST
90		Tag: george-20020322-cps-branch-prob
91		Description:
92
93		Implemented the Ball-Larus branch prediction-heuristics, and
94		incorporated graphical viewers for control flow graphs.
95
96		Ball-Larus Heuristics:
97		---------------------
98		See the file compiler/CodeGen/cpscompile/cpsBranchProb.sml.
99
100		By design it uses the Dempster-Shafer theory for combining
101		probabilities.  For example, in the function:
102
103		fun f(n,acc) = if n = 0 then acc else f(n-1, n*acc)
104
105		the ball-larus heuristics predicts that the n=0 is unlikely
106		(OH-heuristic), and the 'then' branch is unlikely because of the
107		RH-heuristic -- giving the 'then' branch an even lower combined
108		probability using the Dempster-Shafer theory.
109
110		Finally, John Reppy's loop analysis in MLRISC, further lowers the
111		probability of the 'then' branch because of the loop in the else
112		branch.
113
114
115		Graphical Viewing:
116		------------------
117		I merely plugged in Allen's graphical viewers into the compiler. The
118		additional code is not much. At the top level, saying:
119
120		Control.MLRISC.getFlag "cfg-graphical-view" := true;
121
122		will display the graphical view of the control flow graph just before
123		back-patching.  daVinci must be in your path for this to work. If
124		daVinci is not available, then the default viewer can be changed
125		using:
126
127		Control.MLRISC.getString "viewer"
128
129		which can be set to "dot" or "vcg" for the corresponding viewers. Of
130		course, these viewers must be in your path.
131
132		The above will display the compilation unit at the level of clusters,
133		many of which are small, boring, and un-interesting. Also setting:
134
135		Control.MLRISC.getInt "cfg-graphical-view_size"
136
137		will display clusters that are larger than the value set by the above.
138
139
140		----------------------------------------------------------------------
141		Name: Matthias Blume
142		Date: 2002/03/21 22:20:00 EST
143		Tag: blume-20020321-kmp-bugfix
144		Description:
145
146		Changed the interface to the KMP routine in PreString and fixed
147		a minor bug in one place where it was used.
148
149		----------------------------------------------------------------------
150		Name: Allen Leung
151		Date: 2002/03/21 20:30:00 EST
152		Tag: leunga-20020321-cfg
153		Description:
154
155		Fixed a potential problem in cfg edge splitting.
156
157		----------------------------------------------------------------------
158		Name: Allen Leung
159		Date: 2002/03/21 17:15:00 EST
160		Tag: leunga-20020321-x86-fp-cfg
161		Description:
162
163		1. Recoded the buggy parts of x86-fp.
164
165		a. All the block reordering code has been removed.
166		We now depend on the block placement phases to do this work.
167
168		b. Critical edge splitting code has been simplified and moved into the
169		CFG modules, as where they belong.
170
171		Both of these were quite buggy and complex.  The code is now much, much
172		simpler.
173
174		2. X86 backend.
175
176		a. Added instructions for 64-bit support.  Instruction selection for
177		64-bit has not been committed, however, since that
178		requires changes to MLTREE which haven't been approved by
179		Lal and John.
180
181		b. Added support for FUCOMI and FUCOMIP when generating code for
182		PentiumPro and above.  We only generate these instructions in
183		the fast-fp mode.
184
185		c. Added cases for JP and JNP in X86FreqProps.
186
187		3. CFG
188
189		CFG now has a bunch of methods for edge splitting and merging.
190
191		4. Machine description.
192
193		John's simplification of MLTREE_BASIS.fcond broke a few machine
194		description things:
195
196		rtl-build.{sig,sml} and hppa.mdl fixed.
197
198		NOTE: the machine description stuff in the repository is still broken.
199		Again, I can't put my fixes in because that involves
200		changes to MLTREE.
201
202		----------------------------------------------------------------------
203		Name: Matthias Blume
204		Date: 2002/03/20 15:55:00 EST
205		Tag: blume-20020320-kmp
206		Description:
207
208		Implemented Knuth-Morris-Pratt string matching in PreString and used
209		it for String.isSubstring, Substring.isSubstring, and
210		Substring.position.
211
212		(Might need some stress-testing.  Simple examples worked fine.)
213
214		----------------------------------------------------------------------
215		Name: Matthias Blume
216		Date: 2002/03/19 16:37:00 EST
217		Tag: blume-20020319-witnesses
218		Description:
219
220		Added a structure C.W and functions convert/Ptr.convert to ml-nlffi-lib.
221
222		This implements a generic mechanism for changing constness qualifiers
223		anywhere within big C types without resorting to outright "casts".
224		(So far, functions such as C.rw/C.ro or C.Ptr.rw/C.Ptr.ro only let you
225		modify the constness at the outermost level.)
226		The implementation of "convert" is based on the idea of "witness"
227		values -- values that are not used by the operation but whose types
228		"testify" to their applicability.  On the implementation side, "convert"
229		is simply a projection (returning its second curried argument).  With
230		cross-module inlining, it should not result in any machine code being
231		generated.
232
233		----------------------------------------------------------------------
234		Name: Matthias Blume
235		Date: 2002/03/15 16:40:00 EST
236		Tag: blume-20020315-basis
237		Description:
238
239		Provided (preliminary?) implementations for
240
241		{String,Substring}.{concatWith,isSuffix,isSubstring}
242
243		and
244
245		Substring.full
246
247		Those are in the Basis spec but they were missing in SML/NJ.
248
249		----------------------------------------------------------------------
250		Name: Matthias Blume
251		Date: 2002/03/14 21:30:00 EST
252		Tag: blume-20020314-controls
253		Description:
254
255		Controls:
256		---------
257
258		1. Factored out the recently-added Controls : CONTROLS stuff and put
259		it into its own library $/controls-lib.cm.  The source tree for
260		this is under src/smlnj-lib/Controls.
261
262		2. Changed the names of types and functions in this interface, so they
263		make a bit more "sense":
264
265		module -> registry
266		'a registry -> 'a group
267
268		3. The interface now deals in ref cells only.  The getter/setter interface
269		is (mostly) gone.
270
271		4. Added a function that lets one register an already-existing ref cell.
272
273		5. Made the corresponding modifications to the rest of the code so that
274		everything compiles again.
275
276		6. Changed the implementation of Controls.MLRISC back to something closer
277		to the original.  In particular, this module (and therefore MLRISC)
278		does not depend on Controls.  There now is some link-time code in
279		int-sys.sml that registers the MLRISC controls with the Controls
280		module.
281
282		CM:
283		---
284
285		* One can now specify the lambda-split aggressiveness in init.cmi.
286
287		----------------------------------------------------------------------
288		Name: Allen Leung
289		Date: 2002/03/13 17:30:00 EST
290		Tag: leunga-20020313-x86-fp-unary
291		Description:
292
293		Bug fix for:
294
295		> leunga@weaselbane:~/Yale/tmp/sml-dist{21} bin/sml
296		> Standard ML of New Jersey v110.39.1 [FLINT v1.5], March 08, 2002
297		> - fun f(x,(y,z)) = Real.~ y;
298		> [autoloading]
299		> [autoloading done]
300		>       fchsl   (%eax), 184(%esp)
301		> Error: MLRisc bug: X86MCEmitter.emitInstr
302		>
303		> uncaught exception Error
304		>   raised at: ../MLRISC/control/mlriscErrormsg.sml:16.14-16.19
305
306		The problem was that the code generator did not generate any fp registers
307		in this case, and the ra didn't know that it needed to run the X86FP phase to
308		translate the pseudo fp instruction.   This only happened with unary fp
309		operators in certain situations.
310
311		----------------------------------------------------------------------
312		Name: Matthias Blume
313		Date: 2002/03/13 14:00:00 EST
314		Tag: blume-20020313-overload-etc
315		Description:
316
317		1. Added _overload as a synonym for overload for backward compatibility.
318		(Control.overloadKW must be true for either version to be accepted.)
319
320		2. Fixed bug in install script that caused more things to be installed
321		than what was requested in config/targets.
322
323		3. Made CM aware of the (_)overload construct so that autoloading
324		works.
325
326		----------------------------------------------------------------------
327		Name: Matthias Blume
328		Date: 2002/03/12 22:03:00 EST
329		Tag: blume-20020312-url
330		Description:
331
332		Forgot to update BOOT and srcarchiveurl.
333
334	----------------------------------------------------------------------	----------------------------------------------------------------------
335	Name: Matthias Blume	Name: Matthias Blume
336	Date: 2002/03/12 17:30:00 EST	Date: 2002/03/12 17:30:00 EST
#	Line 27	Line 347
347	provides a bit more "stability":  Once CM has seen a compilation	provides a bit more "stability":  Once CM has seen a compilation
348	unit, it keeps its identity constant (as long as you do not delete	unit, it keeps its identity constant (as long as you do not delete
349	those crucial CM/GUID/* files).  This means that when you change	those crucial CM/GUID/* files).  This means that when you change
350	and interface, compiler, then go back to the old interface, and	an interface, compile, then go back to the old interface, and
351	compile again, you arrive at the original pid.	compile again, you arrive at the original pid.
352
353	There now also is a mechanism that instructs CM to use the plain	There now also is a mechanism that instructs CM to use the plain
#	Line 102	Line 422
422	----------------------------------------------------------------------	----------------------------------------------------------------------
423	Name: Allen Leung	Name: Allen Leung
424	Date: 2002/03/11 10:30:00 EST	Date: 2002/03/11 10:30:00 EST
425	Tag: leunga-20020310-runtime-string0	Tag: leunga-20020311-runtime-string0
426	Description:	Description:
427
428	The representation of the empty string now points to a	The representation of the empty string now points to a

---

Colored Diff Long Colored Diff Unidiff Context Diff Side by Side

Legend: Removed from v.1137   changed lines   Added in v.1174

---

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