Diff of /sml/trunk/HISTORY

Parent Directory | Revision Log | Patch

	revision 1158, Fri Mar 22 01:21:25 2002 UTC	revision 1238, Mon Jun 10 20:35:28 2002 UTC
#	Line 13	Line 13
13	Description:	Description:
14
15	----------------------------------------------------------------------	----------------------------------------------------------------------
16		Name: Matthias Blume
17		Date: 2002/06/10 16:35:00 EDT
18		Tag: blume-20020610-ieeereal
19		Description:
20
21		I back-ported my implementation of IEEEReal.fromString from the
22		idlbasis-devel branch so that we can test it.
23
24		Another small change is that ppDec tries to give more information
25		than just "<sig>" in the case of functors.  However, this code is
26		broken in some mysterious way if the functor's body's signature
27		has not been declared by ascription but gets inferred from the
28		implementation.  This needs fixing...
29
30		----------------------------------------------------------------------
31		Name: Matthias Blume
32		Date: 2002/05/31
33		Tag: blume-20020531-btrace-mode
34		Description:
35
36		Resurrected SMLofNJ.Internals.BTrace.mode.  (It accidentally fell by
37		the wayside when I switched over to using Controls everywhere.)
38
39		----------------------------------------------------------------------
40		Name: Lal George
41		Date: 2002/05/23 12:21:40 EDT
42		Tag: george-20020523-visual-labels
43		Description:
44
45		Labels are now displayed in the graphical output to make
46		the fall-through and target blocks obvious.
47
48		----------------------------------------------------------------------
49		Name: Matthias Blume
50		Date: 2002/05/22 11:03:00 EDT
51		Tag: blume-20020522-shrink
52		Description:
53
54		John tweaked yesterday's fix for 1131 to handle an out-of-memory
55		situation that comes up when allocating huge arrays.
56
57		----------------------------------------------------------------------
58		Name: Matthias Blume
59		Date: 2002/05/21 16:00:00 EDT
60		Tag: Release_110_40
61		Description:
62
63		New working release (110.40).  New bootfiles.
64
65		[Also: John Reppy fixed GC bug 1131.]
66
67		----------------------------------------------------------------------
68		Name: Matthias Blume
69		Date: 2002/05/21 12:35:00 EDT
70		Tag: blume-20020521-cmdoc
71		Description:
72
73		CM documentation update.
74
75		----------------------------------------------------------------------
76		Name: Matthias Blume
77		Date: 2002/05/21 10:55:00 EDT
78		Tag: blume-20020521-misc
79		Description:
80
81		- John tweaked runtime to be silent on heap export (except when
82		GC messages are on).
83		- I added a few more things (cross-compiling versions of CMB) to
84		config/preloads (as suggestions).
85
86		----------------------------------------------------------------------
87		Name: Matthias Blume
88		Date: 2002/05/20 22:25:00 EDT
89		Tag: blume-20020520-controls
90		Description:
91
92		- Added ControlUtil structure to control-lib.cm.
93		- Use it throughout.
94		- Used Controls facility to define MLRISC controls (as opposed to
95		registering MLRISC control ref cells with Controls after the fact)
96		- Fixed messed-up controls priorities.
97
98		* Removed again all the stuff from config/preloads that one wouldn't
99		be able to preload at the time the initial heap image is built.
100		(Many libraries, e.g., CML, do not exist yet at this time.  The
101		only libraries that can be preloaded via config/preloads are those
102		that come bundled with the bootfiles.)
103
104		----------------------------------------------------------------------
105		Name: Matthias Blume
106		Date: 2002/05/20 10:59:00 EDT
107		Tag: blume-20020520-preloads
108		Description:
109
110		Added a lot of commented-out suggestions for things to be included
111		in config/preloads.
112
113		----------------------------------------------------------------------
114		Name: Allen Leung
115		Date: 2002/05/18 14:20:00 EDT
116		Tag: leunga-20020518-mdl
117		Description:
118
119		o Made the mdl tool stuff compile and run again.
120		o I've disabled all the stuff that depends on RTL specifications; they
121		are all badly broken anyway.
122
123		----------------------------------------------------------------------
124		Name: Matthias Blume
125		Date: 2002/05/17 16:49:00 EDT
126		Tag: blume-20020517-controls
127		Description:
128
129		0. John Reppy made several modifications to the SML/NJ library.
130		In particular, there is a shiny new controls-lib.cm.
131
132		1. Pushed new controls interface through compiler so that everything
133		compiles again.
134
135		2. Added FormatComb and FORMAT_COMB to the CML version of the
136		SML/NJ library (so that CML compiles again).
137
138		3. Modified init scripts because XXX_DEFAULT environment variables
139		are no longer with us.  (Boot-time initialization is now done
140		using the same environment variables that are also used for
141		startup-time initialization of controls.)
142
143		----------------------------------------------------------------------
144		Name: Lal George
145		Date: 2002/05/15 09:20:10 EDT
146		Tag: george-20020515-pseudo-op-decls
147		Description:
148
149		All pseudo-ops emitted before the first segment declaration
150		such as TEXT, DATA, and BSS directives are assumed to be global
151		declarations and are emitted first in the assembly file. This is
152		useful in a number of situations where one has pseudo-ops that are not
153		specific to any segment, and also works around the constraint that one
154		cannot have client pseudo-ops in the TEXT segment.
155
156		Because no segment is associated with these declarations it is
157		an error to allocate any space or objects before the first segment
158		directive and an exception will be raised. However, we cannot make
159		this check for client pseudo-ops.
160
161		These top level declarations are a field in the CFG graph_info.
162		In theory you can continue to add to this field after the CFG has been
163		built -- provided you know what you are doing;-)
164
165		----------------------------------------------------------------------
166		Name: Matthias Blume
167		Date: 2002/05/13 16:40:00 EDT
168		Tag: blume-20020513-pp-etc
169		Description:
170
171		A few minor bugfixes:
172
173		- Stopgap measure for bug recently reported by Elsa Gunter (ppDec).
174		(Bogus printouts for redefined bindings still occur.  Compiler
175		bug should no longer occur now.  We need to redo the prettyprinter
176		from scratch.)
177
178		- CM pathname printer now also adds escape sequences for ( and )
179
180		- commend and docu fixes for ml-nlffi
181
182		----------------------------------------------------------------------
183		Name: Matthias Blume
184		Date: 2002/05/10 16:40:00 EDT
185		Tag: blume-20020510-erg-textio
186		Description:
187
188		Applied the following bugfix provided by Emden Gansner:
189
190		Output is corrupted when outputSubstr is used rather than output.
191
192		The problem occurs when a substring
193
194		ss = (s, dataStart, dataLen)
195
196		where dataStart > 0, fills a stream buffer with avail bytes left.
197		avail bytes of s, starting at index dataStart, are copied into the
198		buffer, the buffer is flushed, and then the remaining dataLen-avail
199		bytes of ss are copied into the beginning of the buffer. Instead of
200		starting this copy at index dataStart+avail in s, the current code
201		starts the copy at index avail.
202
203		Fix:
204		In text-io-fn.sml, change line 695 from
205		val needsFlush = copyVec(v, avail, dataLen-avail, buf, 0)
206		to
207		val needsFlush = copyVec(v, dataStart+avail, dataLen-avail, buf, 0)
208
209		----------------------------------------------------------------------
210		Name: Matthias Blume
211		Date: 2002/04/12 13:55:00 EDT
212		Tag: blume-20020412-assyntax
213		Description:
214
215		1. Grabbed newer assyntax.h from the XFree86 project.
216		2. Fiddled with how to compile X86.prim.asm without warnings.
217		3. (Very) Minor cleanup in CM.
218
219		----------------------------------------------------------------------
220		Name: Matthias Blume
221		Date: 2002/04/01 (no joke!) 17:07:00 EST
222		Tag: blume-20020401-x86div
223		Description:
224
225		Added full support for div/mod/rem/quot on the x86, using the machine
226		instruction's two results (without clumsily recomputing the remainder)
227		directly where appropriate.
228
229		Some more extensive power-of-two support was added to the x86 instruction
230		selector (avoiding expensive divs, mods, and muls where they can be
231		replaced with cheaper shifts and masks).  However, this sort of thing
232		ought to be done earlier, e.g., within the CPS optimizer so that
233		all architectures benefit from it.
234
235		The compiler compiles to a fixed point, but changes might be somewhat
236		fragile nevertheless.  Please, report any strange things that you might
237		see wrt. div/mod/quot/rem...
238
239		----------------------------------------------------------------------
240		Name: Matthias Blume
241		Date: 2002/03/29 17:22:00
242		Tag: blume-20020329-div
243		Description:
244
245		Fixed my broken div/mod logic.  Unfortunately, this means that the
246		inline code for div/mod now has one more comparison than before.
247		Fast paths (quotient > 0 or remainder = 0) are not affected, though.
248		The problem was with quotient = 0, because that alone does not tell
249		us which way the rounding went.  One then has to look at whether
250		remainder and divisor have the same sign...  :(
251
252		Anyway, I replaced the bootfiles with fresh ones...
253
254		----------------------------------------------------------------------
255		Name: Matthias Blume
256		Date: 2002/03/29 14:10:00 EST
257		Tag: blume-20020329-inlprims
258		Description:
259
260		NEW BOOTFILES!!!    Version number bumped to 110.39.3.
261
262		Primops have changed. This means that the bin/boot-file formats have
263		changed as well.
264
265		To make sure that there is no confusion, I made a new version.
266
267
268		CHANGES:
269
270		* removed REMT from mltree (remainder should never overflow).
271
272		* added primops to deal with divisions of all flavors to the frontend
273
274		* handled these primops all the way through so they map to their respective
275		MLRISC support
276
277		* used these primops in the implementation of Int, Int32, Word, Word32
278
279		* removed INLDIV, INLMOD, and INLREM as they are no longer necessary
280
281		* parameterized INLMIN, INLMAX, and INLABS by a numkind
282
283		* translate.sml now deals with all flavors of INL{MIN,MAX,ABS}, including
284		floating point
285
286		* used INL{MIN,MAX,ABS} in the implementation of Int, Int32, Word, Word32,
287		and Real (but Real.abs maps to a separate floating-point-only primop)
288
289
290		TODO items:
291
292		* Hacked Alpha32 instruction selection, disabling the selection of REMx
293		instructions because the machine instruction encoder cannot handle
294		them.  (Hppa, PPC, and Sparc instruction selection did not handle
295		REM in the first place, and REM is supported by the x86 machine coder.)
296
297		* Handle DIV and MOD with DIV_TO_NEGINF directly in the x86 instruction
298		selection phase.  (The two can be streamlined because the hardware
299		delivers both quotient and remainder at the same time anyway.)
300
301		* Think about what to do with "valOf(Int32.minInt) div ~1" and friends.
302		(Currently the behavior is inconsistent both across architectures and
303		wrt. the draft Basis spec.)
304
305		* Word8 should eventually be handled natively, too.
306
307		* There seems to be one serious bug in mltree-gen.sml.  It appears, though,
308		as if there currently is no execution path that could trigger it in
309		SML/NJ.  (The assumptions underlying functions arith and promotable do not
310		hold for things like multiplication and division.)
311
312		----------------------------------------------------------------------
313		Name: Matthias Blume
314		Date: 2002/03/27 16:27:00 EST
315		Tag: blume-20020327-mlrisc-divisions
316		Description:
317
318		Added support for all four division operations (ML's div, mod, quot,
319		and rem) to MLRISC.  In the course of doing so, I also rationalized
320		the naming (no more annoying switch-around of DIV and QUOT), by
321		parameterizing the operation by div_rounding_mode (which can be either
322		DIV_TO_ZERO or DIV_TO_NEGINF).
323
324		The generic MLTreeGen functor takes care of compiling all four
325		operations down to only round-to-zero div.
326
327		Missing pieces:
328
329		* Doing something smarter than relying on MLTreeGen on architectures
330		like, e.g., the x86 where hardware division delivers both quotient and
331		remainder at the same time.  With this, the implementation of the
332		round-to-neginf operations could be further streamlined.
333
334		* Remove inlining support for div/mod/rem from the frontend and replace it
335		with primops that get carried through to the backend.  Do this for all
336		int and word types.
337
338		----------------------------------------------------------------------
339		Name: Matthias Blume
340		Date: 2002/03/25 17:25:00 EST
341		Tag: blume-20020325-divmod
342		Description:
343
344		I improved (hopefully without breaking them) the implementation of Int.div,
345		Int.mod, and Int.rem.   For this, the code in translate.sml now takes
346		advantage of the following observations:
347
348		Let  q = x quot y      r = x rem y
349		d = x div  y      m = x mod y
350
351		where "quot" is the round-to-zero version of integer division that
352		hardware usually provides.  Then we have:
353
354		r = x - q * y        where neither the * nor the - will overflow
355		d = if q >= 0 orelse x = q * y then q else q - 1
356		where neither the * nor the - will overflow
357		m = if q >= 0 orelse r = 0 then r else r + y
358		where the + will not overflow
359
360		This results in substantial simplification of the generated code.
361		The following table shows the number of CFG nodes and edges generated
362		for
363		fun f (x, y) = x OPER y
364		(* with OPER \in div, mod, quot, rem *)
365
366
367		OPER | nodes(old) | edges(old) | nodes(new) | edges(new)
368		--------------------------------------------------------
369		div |         24 |         39 |         12 |         16
370		mod |         41 |         71 |         12 |         16
371		quot |          8 |         10 |          8 |         10
372		rem |         10 |         14 |          8 |         10
373
374
375		----------------------------------------------------------------------
376		Name: Matthias Blume
377		Date: 2002/03/25 22:06:00 EST
378		Tag: blume-20020325-cprotobug
379		Description:
380
381		Fixed a bug in cproto (c prototype decoder).
382
383		----------------------------------------------------------------------
384		Name: Matthias Blume
385		Date: 2002/03/25 16:00:00 EST
386		Tag: blume-20020325-raw-primops
387		Description:
388
389		I did some cleanup to Allen's new primop code and
390		replaced yesterday's bootfiles with new ones.
391		(But they are stored in the same place.)
392
393		----------------------------------------------------------------------
394		Name: Matthias Blume
395		Date: 2002/03/24 22:40:00 EST
396		Tag: blume-20020324-bootfiles
397		Description:
398
399		Made the bootfiles that Allen asked for.
400
401		----------------------------------------------------------------------
402		Name: Allen Leung
403		Date: 2002/03/23 15:50:00 EST
404		Tag: leunga-20020323-flint-cps-rcc-primops
405		Description:
406
407		1. Changes to FLINT primops:
408
409		(* make a call to a C-function;
410		* The primop carries C function prototype information and specifies
411		* which of its (ML-) arguments are floating point. C prototype
412		* information is for use by the backend, ML information is for
413		* use by the CPS converter. *)
414		| RAW_CCALL of { c_proto: CTypes.c_proto,
415		ml_args: ccall_type list,
416		ml_res_opt: ccall_type option,
417		reentrant : bool
418		} option
419		(* Allocate uninitialized storage on the heap.
420		* The record is meant to hold short-lived C objects, i.e., they
421		* are not ML pointers.  With the tag, the representation is
422		* the same as RECORD with tag tag_raw32 (sz=4), or tag_fblock (sz=8)
423		*)
424		| RAW_RECORD of {tag:bool,sz:int}
425		and ccall_type = CCALL_INT32 | CCALL_REAL64 | CCALL_ML_PTR
426
427		2.  These CPS primops are now overloaded:
428
429		rawload of {kind:numkind}
430		rawstore of {kind:numkind}
431
432		The one argument form is:
433
434		rawload {kind} address
435
436		The two argument form is:
437
438		rawload {kind} [ml object, byte-offset]
439
440		3. RAW_CCALL/RCC now takes two extra arguments:
441
442		a. The first is whether the C call is reentrant, i.e., whether
443		ML state should be saved and restored.
444		b. The second argument is a string argument specifying the name of
445		library and the C function.
446
447		These things are currently not handled in the code generator, yet.
448
449		4. In CProto,
450
451		An encoding type of "bool" means "ml object" and is mapped into
452		C prototype of PTR.  Note that "bool" is different than "string",
453		even though "string" is also mapped into PTR, because "bool"
454		is assigned an CPS type of BOGt, while "string" is assigned INT32t.
455
456		5. Pickler/unpicker
457
458		Changed to handle RAW_RECORD and newest RAW_CCALL
459
460		6. MLRiscGen,
461
462		1. Changed to handle the new rawload/rawstore/rawrecord operators.
463		2. Code for handling C Calls has been moved to a new module CPSCCalls,
464		in the file CodeGen/cpscompile/cps-c-calls.sml
465
466		7. Added the conditional move operator
467
468		condmove of branch
469
470		to cps.  Generation of this is still buggy so it is currently
471		disabled.
472
473		----------------------------------------------------------------------
474		Name: Lal George
475		Date: 2002/03/22 14:18:25 EST
476		Tag: george-20020322-cps-branch-prob
477		Description:
478
479		Implemented the Ball-Larus branch prediction-heuristics, and
480		incorporated graphical viewers for control flow graphs.
481
482		Ball-Larus Heuristics:
483		---------------------
484		See the file compiler/CodeGen/cpscompile/cpsBranchProb.sml.
485
486		By design it uses the Dempster-Shafer theory for combining
487		probabilities.  For example, in the function:
488
489		fun f(n,acc) = if n = 0 then acc else f(n-1, n*acc)
490
491		the ball-larus heuristics predicts that the n=0 is unlikely
492		(OH-heuristic), and the 'then' branch is unlikely because of the
493		RH-heuristic -- giving the 'then' branch an even lower combined
494		probability using the Dempster-Shafer theory.
495
496		Finally, John Reppy's loop analysis in MLRISC, further lowers the
497		probability of the 'then' branch because of the loop in the else
498		branch.
499
500
501		Graphical Viewing:
502		------------------
503		I merely plugged in Allen's graphical viewers into the compiler. The
504		additional code is not much. At the top level, saying:
505
506		Control.MLRISC.getFlag "cfg-graphical-view" := true;
507
508		will display the graphical view of the control flow graph just before
509		back-patching.  daVinci must be in your path for this to work. If
510		daVinci is not available, then the default viewer can be changed
511		using:
512
513		Control.MLRISC.getString "viewer"
514
515		which can be set to "dot" or "vcg" for the corresponding viewers. Of
516		course, these viewers must be in your path.
517
518		The above will display the compilation unit at the level of clusters,
519		many of which are small, boring, and un-interesting. Also setting:
520
521		Control.MLRISC.getInt "cfg-graphical-view_size"
522
523		will display clusters that are larger than the value set by the above.
524
525
526		----------------------------------------------------------------------
527		Name: Matthias Blume
528		Date: 2002/03/21 22:20:00 EST
529		Tag: blume-20020321-kmp-bugfix
530		Description:
531
532		Changed the interface to the KMP routine in PreString and fixed
533		a minor bug in one place where it was used.
534
535		----------------------------------------------------------------------
536	Name: Allen Leung	Name: Allen Leung
537	Date: 2002/03/21 20:30:00 EST	Date: 2002/03/21 20:30:00 EST
538	Tag: leunga-20020321-cfg	Tag: leunga-20020321-cfg

---

Colored Diff Long Colored Diff Unidiff Context Diff Side by Side

Legend: Removed from v.1158   changed lines   Added in v.1238

---

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