Diff of /sml/trunk/HISTORY

Parent Directory | Revision Log | Patch

	revision 577, Fri Mar 10 08:07:18 2000 UTC	revision 647, Tue May 16 02:53:02 2000 UTC
#	Line 12	Line 12
12	Tag: <post-commit CVS tag>	Tag: <post-commit CVS tag>
13	Description:	Description:
14	----------------------------------------------------------------------	----------------------------------------------------------------------
15		Name: Allen Leung
16		Date: 2000/05/15 22:53 EDT
17		Tag: leunga-20000515-alpha-x86-ra
18		Description:
19
20		1. Alpha
21
22		Slight cleanup.  Removed the instruction SGNXL
23
24		2. X86
25
26		Added the following instructions to the instruction set:
27
28		ROLx, RORx,
29		BTx, BTSx, BTLx, BTRx,
30		XCHGx, and variants with the LOCK prefix
31
32		3. Register Allocation
33
34		The module ra-rewrite-with-renaming has been improved.
35
36		These have no effect on SML/NJ.
37
38		----------------------------------------------------------------------
39		Name: Matthias Blume
40		Date: 2000/05/15 16:20:00 JST
41		Tag: blume-20000515-lightrebuild
42		Description:
43
44		1. I added an alternative to "-rebuild" to "makeml".  The difference is
45		that prior to calling CMB.make' the CM-variable "LIGHT" will be
46		defined.  In effect, the command will not build any cross-compiler
47		backends and therefore finish more quickly.
48
49		The "fixpt" script also takes a "-light" switch to be able to use
50		this new facility while compiling for a fixpoint.
51
52		2. I replaced all mentions of anchored paths in group owner specifications
53		with simple relative paths (usually starting with "..").
54		The rationale is that a library's internal workings should not be
55		compromised by the lack of some anchor.  (An anchor is necessary
56		for someone who wants to refer to the library by an anchored path,
57		but it should not be necessary to build the same library in the first
58		place.)
59
60		3. I changed the way CM's tool mechanism determines the shell command
61		string used for things like ml-yacc etc. so that it does not break
62		when CM.Control.implicit_anchors is turned off.
63
64		----------------------------------------------------------------------
65		Name: Matthias Blume
66		Date: 2000/05/12 18:20:00 JST
67		Tag: blume-20000512-ml-build
68		Description:
69
70		Fixed a bug in config/_ml-build that prevented ml-yacc and ml-lex from
71		getting installed properly (by config/install.sh).
72
73		----------------------------------------------------------------------
74		Name: Matthias Blume
75		Date: 2000/05/12 17:30:00 JST
76		Tag: blume-20000512-anchors
77		Description:
78
79		!!! NEW BOOT FILES !!!
80
81		This change is in preparation of fading out support for "implicitly
82		anchored path names".  I went through all sources and used the
83		explicit (and relatively new) $-notation.  See system/README and the
84		CM manual for more info on this.
85
86		I also modified the anchoring scheme for some things such as "smlnj",
87		"MLRISC", "cm", etc. to take advantage of the fact that explicit
88		anchors are more expressive: anchor name and first arc do not have to
89		coincide.  This entails the following user-visible change:
90
91		You have to write $smlnj/foo/bar instead of smlnj/foo/bar.  In
92		particular, when you fire up sml with a command-line argument, say,
93		e.g.:
94
95		sml '$smlnj/cmb.cm'
96
97		At the ML toplevel prompt:
98
99		CM.autoload "$smlnj/cmb.cm";
100
101		There is also a new controller in CM.Control that can be used to turn
102		off all remaining support for implicit anchors by saying:
103
104		CM.autoload "$smlnj/
105		#set CM.Control.implicit_anchors false;
106
107		This causes CM to reject implicitly anchored paths.  This is (for the
108		time being) less permissive than the "final" version where there will
109		be no more such implicit anchors and relative paths will be just that:
110		relative.
111
112		The next step (version after next version?) will be to make the
113		default for CM.Control.implicit_anchors false.  After the dust has
114		settled, I can then produce the "final" version of this...
115
116		Note: Since bootstrapping is a bit tricky, I provided new boot files.
117
118		----------------------------------------------------------------------
119		Name: Matthias Blume
120		Date: 2000/05/11 16:30:00 JST
121		Tag: blume-20000511-sources
122		Description:
123
124		The main change is that I added function CM.sources as a generalized
125		version of the earlier CM.makedepend.  This entails the following
126		additional changes:
127
128		- CM.makedepend has been dropped.
129
130		- CM manual has been updated.
131
132		- TOOLS signature and API have been changed.
133
134		----------------------------------------------------------------------
135		Name: Allen Leung
136		Date: 2000/05/10 21:17 EDT
137		Tag: leunga-20000510-moby-c--ssa
138		Description:
139
140		Various bug fixes and new features for C--, Moby and MLRISC optimizations.
141		None of these affect SML/NJ.
142
143		1. Register Allocation
144
145		a. A new ra spilling module (ra/ra-spill-with-renaming) is implemented.
146		This module tries to remove local (i.e. basic block level) redundancies
147		during spilling.
148
149		b. A new framework for performing region based register allocation.
150		Not yet entirely functional.
151
152		2. X86
153
154		a. DefUse for POP was missing the stack pointer [found by Lal]
155		b. Reload for CALL was incorrect in X86Spill [found by John]
156		c. Various fixes in X86Spill so that it can be used correctly for
157		the new spilling module.
158
159		3. SSA/IR
160
161		a. New module ir/dj-dataflow.sml implements elimination based
162		data flow analysis.
163
164		4. MLRiscGen
165
166		a. Fix for gc type annotation
167
168		5. MDGen
169
170		Various fixes for machine description -> ml code translation.  For ssa
171		only.
172
173		----------------------------------------------------------------------
174		Name: Allen Leung
175		Date: 2000/05/08 22:17 EDT
176		Tag: leunga-20000508-labexp
177		Description:
178
179		Fermin has found a few assembly problems with constant expressions
180		generated in LabelExp.  Mostly, the problems involve extra parentheses,
181		which choke on dumb assemblers.  This is his fix.
182
183		----------------------------------------------------------------------
184		Name: Dave MacQueen
185		Date: 2000/04/09 14:00 EDT
186		Tag: dbm-20000502-Version_110_28
187		Description:
188
189		1. Updated src/compiler/TopLevel/main/version.sml to version 110.28
190
191		2. Updated config/version to 110.28
192
193		3. Updated config/srcarchiveurl
194
195		3. New boot files!
196		ftp://ftp.research.bell-labs.com/dist/smlnj/working/110.28/
197
198		----------------------------------------------------------------------
199		Name: Matthias Blume
200		Date: 2000/05/01 19:05:00 JST
201		Tag: blume-20000501-noweb
202		Description:
203
204		A new noweb tool has been added.  The existing system is entirely
205		unaffected by this, but some CM users have asked for renewed noweb
206		support.  Everything is documented in the CM manual.
207
208		New (plugin) libraries:
209
210		noweb-tool.cm
211		nw-ext.cm
212
213		----------------------------------------------------------------------
214		Name: Dave MacQueen
215		Date: 2000/04/30 12:40PM EDT
216		Tag: dbm-20000430-bug_fixes
217		Description:
218
219		1. Fix for bug 1498
220		smlnj/src/system/Basis/Implementation/Unsafe/object.sig
221		smlnj/src/system/Basis/Implementation/Unsafe/object.sml
222		added toRealArray function
223		smlnj/src/compiler/MiscUtil/print/ppobj.sml
224		added check for tag Obj.RealArray to array printing case in ppObj
225
226		2. Fix for bug 1510
227		smlnj/src/compiler/Semant/types/typesutil.sml
228		fixed definition of dummyargs (used by equalTycon) so that
229		dummy args are distinct types
230
231		----------------------------------------------------------------------
232		Name: Matthias Blume
233		Date: 2000/04/30 01:00:00 JST
234		Tag: blume-20000430-versions
235		Description:
236
237		1. CM version numbering added.  This is an implementation of Lal's
238		proposal for adding version numbers and version checking to .cm
239		files.  Lal said that his proposal was just that -- a proposal.
240		For the time being I went ahead and implemented it so that people
241		can comment on it.  Everything is completely backward-compatible
242		(except for the stable library format, i.e., new bootfiles!).
243
244		As usual, see the CM manual for details.
245
246		2. An alternative syntax for anchored paths has been implemented.
247		Dave has recently voiced the same concerns that I had when I did
248		this, so there should be some support.  My take is that eventually
249		I will let support for the current syntax (where anchors are
250		"implicit") fade out in favor of the new, explicit syntax.
251		In order to be backward-compatible, both old and new syntax are
252		currently supported.
253
254		Again, see the CM manual for details.
255
256		3. Parallel make is trying to be slightly smarter:  When the master
257		process finds a "bottleneck", i.e., when there is only one
258		compilation unit that can be compiled and everybody else is
259		waiting on it, then it will simply compile it directly instead
260		of clumsily telling one of the slaves to do it.
261
262		4. Support for "unsharing" added.  This is necessary in order to be
263		able to have two different versions of the same library running
264		at the same time (e.g., for trying out a new MLRISC while still
265		having the old MLRISC linked into the current compiler, etc.)
266		See the CM manual.
267
268		5. Simple "makedepend" functionality added for generating Makefile
269		dependency information.  (This is rather crude at the moment.
270		Expect some changes here in the future.)
271
272		6. ".fun" added as a recognized suffix for ML files. Also documented
273		explicitly in the manual that the fallback behavior (unknown suffix
274		-> ML file) is not an official feature!
275
276		7. Small changes to the pickler for stable libraries.
277
278		8. Several internal changes to CM (for cleanup/improvement).
279
280
281		!!!! NEW BINFILES !!!!
282
283		----------------------------------------------------------------------
284		Name: Matthias Blume
285		Date: 2000/04/28 17:30:00 JST
286		Tag: blume-20000428-pathconfig
287		Description:
288
289		1. I changed config/install.sh to remove duplicate entries from the
290		lib/pathconfig file at the end.  Moreover, the final version of
291		lib/pathconfig is sorted alphabetically.  The same (sorting) is done
292		in src/system/installml.
293
294		2. The config/install.sh script now consistently uses relative
295		pathnames in lib/pathconfig whenever the anchor is in the lib
296		directory.  (So far this was true for the libraries that come
297		pre-compiled and bundled as part of the bootfiles but not for
298		libraries that are compiled by the script itself.)
299
300		----------------------------------------------------------------------
301		Name: Matthias Blume
302		Date: 2000/04/26 13:10:00 JST
303		Tag: blume-20000426-fun_suffix
304		Description:
305
306		Added ".fun" as a recognized file name suffix (for ML code).
307
308		----------------------------------------------------------------------
309		Name: Allen Leung
310		Date: 2000/04/25 17:00:00 EST
311		Tag: leunga-20000425-alpha-ra
312		Description:
313
314		1. Alpha
315
316		PSEUDOARITH was missing in AlphaRewrite.  This causes an endless loop
317		in C--.
318
319		2. RA
320
321		Added a flag "ra-dump-size" to print out the size of the flowgraph
322		and the interference graph.
323
324		----------------------------------------------------------------------
325		Name: Dave MacQueen
326		Date: 2000/04/25/
327		Tag: dbm-20000425-mlyacc_doc_examples
328		Description:
329		Updated mlyacc.tex sections 5 and 7 for SML '97 and CM.
330		Updated all three examples in src/ml-yacc/examples to run
331		under 110.* using CM.make.
332
333		----------------------------------------------------------------------
334		Name: Allen Leung
335		Date: 2000/04/20 23:04:00 EST
336		Tag: leunga-20000420-ssa-c---stuff
337		Description:
338
339		This update synchronizes my repository with Yale's.  Most of these
340		changes, however, do not affect SML/NJ at all (the RA is an exception).
341
342		1. Register Allocator
343
344		a. An improvement in the interference graph construction:
345		Given a copy
346
347		s <- t
348
349		no interference edge between s and t is added for this definition of s.
350
351		b. I've added two new spill heuristic modules that Fermin and I developed
352		(in the new library RA.cm). These are unused in SML/NJ but maybe
353		useful for others (Moby?)
354
355		2. X86
356
357		a. Various fixes in the backend provided by Fermin [C--] and Lal.
358
359		3. Alpha
360
361		a. Added the BSR instruction and code generation that goes with it [C--]
362		b. Other fixes too numerous to recount provided by Fermin [C--]
363
364		4. Regmaps
365
366		a. The regmaps are not initialized with the identity physical bindings
367		at creation time.  This is unneeded.
368
369		5. MLRISC Optimizations
370
371		a. The DJ-Graph module can now compute the iterated dominance frontiers
372		intersects with liveness incrementally in linear time! Woohoo!
373		This is now used in my new SSA construction algorithm.
374
375		b. THe branch reorganization module is now smarter about linear chains of
376		basic blocks.
377
378
379		----------------------------------------------------------------------
380		Name: Matthias Blume
381		Date: 2000/04/12 13:52:00 JST
382		Tag: blume_main_v110p27_1
383		Description:
384
385		Changed install.sh script to handle archive files without version number
386		and to use "boot.<arch>-<os>" instead of "sml.boot.<arch>-<os>" for the
387		name of the boot file archive.
388
389		----------------------------------------------------------------------
390		Name: Dave MacQueen
391		Date: 2000/04/09 14:00 EDT
392		Tag: dbm-20000410-Version_110_27
393		Description:
394
395		1. Updated src/compiler/TopLevel/main/version.sml to version 110.27
396
397		2. Updated src/config/version to 110.27
398
399		3. New boot files!
400
401		----------------------------------------------------------------------
402		Name: Allen Leung
403		Date: 2000/04/09 19:09:00 EST
404		Tag: leunga-20000409-misc
405		Description:
406
407		1.  Yet another fix for x86 assembly for idivl, imull, mull and friends.
408
409		2.  Miscellaneous improvements to MLRISC (unused in sml/nj)
410
411		----------------------------------------------------------------------
412		Name: Stefan
413		Date: 2000/04/07 10:00:00 EDT
414		Tag: monnier-20000406-branch-handling
415		Description:
416
417		Improved handling of branches (mostly those generated from
418		polymorphic equality), removed switchoff and changed the
419		default optimization settings (more cpsopt and less flintopt).
420
421		----------------------------------------------------------------------
422		Name: Allen Leung
423		Date: 2000/04/06 01:30:00 EST
424		Tag: leunga-20000406-peephole-x86-SSA-2
425		Description:
426
427		Forgot a few files.
428
429		----------------------------------------------------------------------
430		Name: Allen Leung
431		Date: 2000/04/06 00:36:00 EST
432		Tag: leunga-20000406-peephole-x86-SSA
433		Description:
434
435		1.  New Peephole code
436
437		2.  Minor improvement to X86 instruction selection
438
439		3.  Various fixes to SSA and machine description -> code translator
440
441		----------------------------------------------------------------------
442		Name: Matthias Blume
443		Date: 2000/04/05 12:30:00 JST
444		Tag: blume_main_v110p26p2_3
445		Description:
446
447		This update just merges three minor cosmetic updates to CM's sources
448		to get ready for the 110.27 code freeze on Friday.  No functionality
449		has changed.
450
451		----------------------------------------------------------------------
452		Name: Allen Leung
453		Date: 2000/04/04 19:39:00 EST
454		Tag: leunga-20000404-x86-asm
455		Description:
456
457		1.  Fixed a problem in X86 assembly.
458
459		Things like
460
461		jmp %eax
462		jmp (%eax)
463
464		should be output as
465
466		jmp *%eax
467		jmp *(%eax)
468
469		2.  Assembly output
470
471		Added a new flag
472
473		"asm-indent-copies" (default to false)
474
475		When this flag is on, parallel copies will be indented an extra level.
476
477		----------------------------------------------------------------------
478		Name: Allen Leung
479		Date: 2000/04/04 03:18:00 EST
480		Tag: leunga-20000404-C--Moby
481		Description:
482
483		All of these fixes are related to C--, Moby, and my own optimization
484		stuff; so they shouldn't affect SML/NJ.
485
486		1.  X86
487
488		Various fixes related floating point, and extensions.
489
490		2.  Alpha
491
492		Some extra patterns related to loads with signed/zero extension
493		provided by Fermin.
494
495		3.  Assembly
496
497		When generating assemby, resolve the value of client defined constants,
498		instead of generating symbolic values.  This is controlled by the
499		new flag "asm-resolve-constants", which is default to true.
500
501		4.  Machine Descriptions
502
503		a. The precedence parser was slightly broken when parsing infixr symbols.
504		b. The type generalizing code had the bound variables reversed, resulting
505		in a problem during arity raising.
506		c. Various fixes in machine descriptions.
507
508		----------------------------------------------------------------------
509		Name: Matthias Blume
510		Date: 2000/04/03 16:05:00 JST
511		Tag: blume_main_v110p26p2_2
512		Description:
513
514		I eliminated coreEnv from compInfo.  Access to the "Core" structure is
515		now done via the ordinary static environment that is context to each
516		compilation unit.
517
518		To this end, I arranged that instead of "structure Core" as "structure
519		_Core" is bound in the pervasive environment.  Core access is done via
520		_Core (which can never be accidentially rebound because _Core is not a
521		legal surface-syntax symbol).
522
523		The current solution is much cleaner because the core environment is
524		now simply part of the pervasive environment which is part of every
525		compilation unit's context anyway.  In particular, this eliminates all
526		special-case handling that was necessary until now in order to deal
527		with dynamic and symbolic parts of the core environment.
528
529		Remaining hackery (to bind the "magic" symbol _Core) is localized in the
530		compilation mananger's bootstrap compiler (actually: in the "init group"
531		handling).  See the comments in src/system/smlnj/init/init.cmi for
532		more details.
533
534		I also tried to track down all mentions of "Core" (as string argument
535		to Symbol.strSymbol) in the compiler and replaced them with a
536		reference to the new CoreSym.coreSym.  Seems cleaner since the actual
537		name appears in one place only.
538
539		Binfile and bootfile format have not changed, but the switchover from
540		the old "init.cmi" to the new one is a bit tricky, so I supplied new
541		bootfiles anyway.
542
543		----------------------------------------------------------------------
544		Name: Allen Leung
545		Date: 2000/04/02 21:17:00 EST
546		Tag: leunga-20000402-mltree
547		Description:
548
549		1. Renamed the constructor CALL in MLTREE by popular demand.
550		2. Added a bunch of files from my repository.  These are currently
551		used by other non-SMLNJ backends.
552
553		----------------------------------------------------------------------
554		Name: Allen Leung
555		Date: 2000/03/31 21:15:00 EST
556		Tag: leunga-20000331-aliasing
557		Description:
558
559		This update contains a rewritten (and hopefully more correct) module
560		for extracting aliasing information from CPS.
561
562		To turn on this feature:
563
564		Compiler.Control.CG.memDisambiguate := true
565
566		To pretty print the region information with assembly
567
568		Compiler.Control.MLRISC.getFlag "asm-show-region" := true;
569
570		To control how many levels of aliasing information are printed, use:
571
572		Compiler.Control.MLRISC.getInt "points-to-show-level" := n
573
574		The default of n is 3.
575
576		----------------------------------------------------------------------
577		Name: David MacQueen
578		Date: 2000/03/31 11:15:00 EST
579		Tag: dbm-20000331-runtime_fix
580		Description:
581
582		This update contains:
583
584		1. runtime/c-lib/c-libraries.c
585		includes added in revision 1.2 caused compilation errors on hppa-hpux
586
587		2. fix for bug 1556
588		system/Basis/Implementation/NJ/internal-signals.sml
589
590		----------------------------------------------------------------------
591		Name: Matthias Blume
592		Date: 2000/03/31 18:00:00 JST
593		Tag: blume_main_v110p26p2_1
594		Description:
595
596		This update contains:
597
598		1. A small change to CM's handling of stable libraries:
599		CM now maintains one "global" modmap that is used for all stable
600		libraries.  The use of such a global modmap maximizes sharing and
601		minimizes the need for re-traversing parts of environments during
602		modmap construction.  (However, this has minor impact since modmap
603		construction seems to account for just one percent or less of total
604		compile time.)
605
606		2. I added a "genmap" phase to the statistics.  This is where I got the
607		"one percent" number (see above).
608
609		3. CM's new tool parameter mechanism just became _even_ better. :)
610		- The parser understands named parameters and recursive options.
611		- The "make" and "shell" tools use these new features.
612		(This makes it a lot easier to cascade these tools.)
613		- There is a small syntax change: named parameters use a
614
615		<name> : ( <option> ... )            or
616		<name> : <string>
617
618		syntax.  Previously, named parameters were implemented in an
619		ad-hoc fashion by each tool individually (by parsing strings)
620		and had the form
621
622		<name>=<string>
623
624		See the CM manual for a full description of these issues.
625
626		----------------------------------------------------------------------
627		Name: Matthias Blume
628		Date: 2000/03/30 18:00:00 JST
629		Tag: blume_main_v110p26p2_0
630		Description:
631
632		!!!!! WARNING !!!!!!
633		!!  New binfiles  !!
634		!!!!!!!!!!!!!!!!!!!!
635
636		This update contains:
637
638		1. Moderate changes to CM:
639
640		- Changes to CM's tools mechanism.  In particular, it is now possible
641		to have tools that accept additional "command line" parameters
642		(specified in the .cm file at each instance where the tool's class is
643		used).
644
645		This was done to accomodate the new "make" and "shell" tools which
646		facilitate fairly seemless hookup to portions of code managed using
647		Makefiles or Shell scripts.
648
649		There are no classes "shared" or "private" anymore.  Instead, the
650		sharing annotation is now a parameter to the "sml" class.
651
652		There is a bit of generic machinery for implementing one's own
653		tools that accept command-line parameters.  However, I am not yet fully
654		satisfied with that part, so expect changes here in the future.
655
656		All existing tools are described in the CM manual.
657
658		- Slightly better error handling.  (CM now surpresses many followup
659		error messages that tended to be more annoying than helpful.)
660
661		2. Major changes to the compiler's static environment data structures.
662
663		- no CMStaticEnv anymore.
664		- no CMEnv, no "BareEnvironment" (actually, _only_ BareEnvironment,
665		but it is called Environment), no conversions between different
666		kinds of static environments
667
668		- There is still a notion of a "modmap", but such modmaps are generated
669		on demand at the time when they are needed.  This sounds slow, but I
670		sped up the code that generates modmaps enough for this not to lead to
671		a slowdown of the compiler (at least I didn't detect any).
672
673		- To facilitate rapid modmap generation, static environments now
674		contain an (optional) "modtree" structure.  Modtree annotations are
675		constructed by the unpickler during unpickling.  (This means that
676		the elaborator does not have to worry about modtrees at all.)
677		Modtrees have the advantage that they are compositional in the same
678		way as the environment data structure itself is compositional.
679		As a result, modtrees never hang on to parts of an environment that
680		has already been rendered "stale" by filtering or rebinding.
681
682		- I went through many, many trials and errors before arriving at the
683		current solution.  (The initial idea of "linkpaths" did not work.)
684		But the result of all this is that I have touched a lot of files that
685		depend on the "modules" and "types" data structures (most of the
686		elaborator). There were a lot of changes during my "linkpath" trials
687		that could have been reverted to their original state but weren't.
688		Please, don't be too harsh on me for messing with this code a bit more
689		than what was strictly necessary...  (I _did_ resist the tempation
690		of doing any "global reformatting" to avoid an untimely death at
691		Dave's hands. :)
692
693		- One positive aspect of the previous point:  At least I made sure that
694		all files that I touched now compile without warnings (other than
695		"polyEqual").
696
697		- compiler now tends to run "leaner" (i.e., ties up less memory in
698		redundant modmaps)
699
700		----------------------------------------------------------------------
701		Name: Allen Leung
702		Date: 2000/03/29 18:00:00
703		Tag: leunga-20000327-mlriscGen_hppa_alpha_x86
704		Boot files (optional): ftp://react-ilp.cs.nyu.edu/leunga/110.26.1-sml.boot.x86-unix-20000330.tar.gz
705		Description:
706
707		This update contains *MAJOR* changes to the way code is generated from CPS
708		in the module mlriscGen, and in various backend modules.
709
710		CHANGES
711		=======
712
713		1. MLRiscGen: forward propagation fix.
714
715		There was a bug in forward propagation introduced at about the same time
716		as the MLRISC x86 backend, which prohibits coalescing to be
717		performed effectively in loops.
718
719		Effect: speed up of loops in RISC architectures.
720		By itself, this actually slowed down certain benchmarks on the x86.
721
722		2. MLRiscGen:  forward propagating addresses from consing.
723
724		I've changed the way consing code is generated.  Basically I separated
725		out the initialization part:
726
727		store tag,   offset(allocptr)
728		store elem1, offset+4(allocptr)
729		store elem2, offset+8(allocptr)
730		...
731		store elemn, offset+4n(allocptr)
732
733		and the address computation part:
734
735		celladdr <- offset+4+alloctpr
736
737		and move the address computation part
738
739		Effect:  register pressure is generally lower as a result.  This
740		makes compilation of certain expressions much faster, such as
741		long lists with non-trivial elements.
742
743		[(0,0), (0,0), .... (0,0)]
744
745		3. MLRiscGen: base pointer elimination.
746
747		As part of the linkage mechanism, we generate the sequence:
748
749		L:  ...  <- start of the code fragment
750
751		L1:
752		base pointer <- linkreg - L1 + L
753
754		The base pointer was then used for computing relocatable addresses
755		in the code fragment.  Frequently (such as in lots of continuations)
756		this is not needed.  We now eliminate this sequence whenever possible.
757
758		For compile time efficiency, I'm using a very stupid local heuristic.
759		But in general, this should be done as a control flow analysis.
760
761		Effect:  Smaller code size.  Speed up of most programs.
762
763		4. Hppa back end
764
765		Long jumps in span dependence resolution used to depend on the existence
766		of the base pointer.
767
768		A jump to a long label L was expanded into the following sequence:
769
770		LDIL %hi(L-8192), %r29
771		LDO  %lo(L-8192)(%r29), %r29
772		ADD  %r29, baseptr, %r29
773		BV,n %r0(%r29)
774
775		In the presence of change (3) above, this will not work.  I've changed
776		it so that the following sequence of instructions are generated, which
777		doesn't mention the base pointer at all:
778
779		BL,n  L', %r29           /* branch and link, L' + 4 -> %r29 */
780		L':  ADDIL L-(L'+4), %r29     /* Compute address of L */
781		BV,n  %r0(%r29)          /* Jump */
782
783		5. Alpha back end
784
785		New alpha instructions LDB/LDW have been added, as per Fermin's
786		suggestions.   This is unrelated to all other changes.
787
788		6. X86 back end
789
790		I've changed andl to testl in the floating point test sequence
791		whenever appropriate.  The Intel optimization guide states that
792		testl is perferable to andl.
793
794		7. RA (x86 only)
795
796		I've improved the spill propagation algorithm, using an approximation
797		of maximal weighted independent sets.   This seems to be necessary to
798		alleviate the negative effect in light of the slow down in (1).
799
800		I'll write down the algorithm one of these days.
801
802		8. MLRiscGen: frequencies
803
804		I've added an annotation that states that all call gc blocks have zero
805		execution frequencies.  This improves register allocation on the x86.
806
807		BENCHMARKS
808		==========
809
810		I've only perform the comparison on 110.25.
811
812		The platforms are:
813
814		HPPA  A four processor HP machine (E9000) with 5G of memory.
815		X86   A 300Hhz Pentium II with 128M of memory, and
816		SPARC An Ultra sparc 2 with 512M of memory.
817
818		I used the following parameters for the SML benchmarks:
819
820		@SMLalloc
821		HPPA    256k
822		SPARC   512k
823		X86     256k
824
825		COMPILATION TIME
826		----------------
827		Here are the numbers comparing the compilation times of the compilers.
828		I've only compared 110.25 compiling the new sources versus
829		a fixpoint version of the new compiler compiling the same.
830
831		110.25                                  New
832		Total  Time in RA  Spill+Reload   Total  Time In RA Spill+Reload
833		HPPA   627s    116s        2684+3584     599s    95s       1003+1879
834		SPARC  892s    173s        2891+3870     708s    116s      1004+1880
835		X86    999s    315s       94006+130691   987s    296s    108877+141957
836
837		110.25         New
838		Code Size      Code Size
839		HPPA   8596736         8561421
840		SPARC  8974299         8785143
841		X86    9029180         8716783
842
843		So in summary, things are at least as good as before.   Dramatic
844		reduction in compilation is obtained on the Sparc; I can't explain it,
845		but it is reproducible.  Perhaps someone should try to reproduce this
846		on their own machines.
847
848		SML BENCHMARKS
849		--------------
850
851		On the average, all benchmarks perform at least as well as before.
852
853		HPPA         Compilation Time     Spill+Reload      Run Time
854		110.25  New            110.25    New   110.25  New
855
856		barnesHut  3.158  3.015  4.75%    1+1       0+0   2.980  2.922   2.00%
857		boyer  6.152  5.708  7.77%    0+0       0+0   0.218  0.213   2.34%
858		count-graphs  1.168  1.120  4.32%    0+0       0+0  22.705 23.073  -1.60%
859		fft  0.877  0.792 10.74%    1+3       1+3   0.602  0.587   2.56%
860		knuthBendix  3.180  2.857 11.32%    0+0       0+0   0.675  0.662   2.02%
861		lexgen  6.190  5.290 17.01%    0+0       0+0   0.913  0.788  15.86%
862		life  0.803  0.703 14.22%   25+25      0+0   0.153  0.140   9.52%
863		logic  2.048  2.007  2.08%    6+6       1+1   4.133  4.008   3.12%
864		mandelbrot  0.077  0.080 -4.17%    0+0       0+0   0.765  0.712   7.49%
865		mlyacc 22.932 20.937  9.53%  154+181    32+57  0.468  0.430   8.91%
866		nucleic  5.183  5.060  2.44%    2+2       0+0   0.125  0.120   4.17%
867		ratio-regions  3.357  3.142  6.84%    0+0       0+0  116.225 113.173 2.70%
868		ray  1.283  1.290 -0.52%    0+0       0+0   2.887  2.855   1.11%
869		simple  6.307  6.032  4.56%   28+30      5+7   3.705  3.658   1.28%
870		tsp  0.888  0.862  3.09%    0+0       0+0   7.040  6.893   2.13%
871		vliw 24.378 23.455  3.94%  106+127    25+45  2.758  2.707   1.91%
872		--------------------------------------------------------------------------
873		Average                     6.12%                                   4.09%
874
875		SPARC        Compilation Time     Spill+Reload      Run Time
876		110.25  New            110.25    New   110.25  New
877
878		barnesHut  3.778  3.592  5.20%    2+2       0+0   3.648  3.453    5.65%
879		boyer  6.632  6.110  8.54%    0+0       0+0   0.258  0.242    6.90%
880		count-graphs  1.435  1.325  8.30%    0+0       0+0  33.672 34.737   -3.07%
881		fft  0.980  0.940  4.26%    3+9       2+6   0.838  0.827    1.41%
882		knuthBendix  3.590  3.138 14.39%    0+0       0+0   0.962  0.967   -0.52%
883		lexgen  6.593  6.072  8.59%    1+1       0+0   1.077  1.078   -0.15%
884		life  0.972  0.868 11.90%   26+26      0+0   0.143  0.140    2.38%
885		logic  2.525  2.387  5.80%    7+7       1+1   5.625  5.158    9.05%
886		mandelbrot  0.090  0.093 -3.57%    0+0       0+0   0.855  0.728   17.39%
887		mlyacc 26.732 23.827 12.19%  162+189    32+57  0.550  0.560   -1.79%
888		nucleic  6.233  6.197  0.59%    3+3       0+0   0.163  0.173   -5.77%
889		ratio-regions  3.780  3.507  7.79%    0+0       0+0 133.993 131.035   2.26%
890		ray  1.595  1.550  2.90%    1+1       0+0   3.440  3.418    0.63%
891		simple  6.972  6.487  7.48%   29+32      5+7   3.523  3.525   -0.05%
892		tsp  1.115  1.063  4.86%    0+0       0+0   7.393  7.265    1.77%
893		vliw 27.765 24.818 11.87%  110+135    25+45  2.265  2.135    6.09%
894		----------------------------------------------------------------------------
895		Average                     6.94%                                    2.64%
896
897		X86          Compilation Time     Spill+Reload      Run Time
898		110.25  New            110.25    New   110.25  New
899
900		barnesHut  5.530  5.420  2.03%  593+893   597+915   3.532  3.440   2.66%
901		boyer  8.768  7.747 13.19%  493+199   301+289   0.327  0.297  10.11%
902		count-graphs  2.040  2.010  1.49%  298+394   315+457  26.578 28.660  -7.26%
903		fft  1.327  1.302  1.92%  112+209   115+210   1.055  0.962   9.71%
904		knuthBendix  5.218  5.475 -4.69%  451+598   510+650   0.928  0.932  -0.36%
905		lexgen  9.970  9.623  3.60% 1014+841  1157+885   0.947  0.928   1.97%
906		life  1.183  1.183  0.00%  162+182   145+148   0.127  0.103  22.58%
907		logic  3.285  3.512 -6.45%  514+684   591+836   5.682  5.577   1.88%
908		mandelbrot  0.147  0.143  2.33%   38+41     33+54    0.703  0.690   1.93%
909		mlyacc 35.457 32.763  8.22% 3496+4564 3611+4860  0.552  0.550   0.30%
910		nucleic  7.100  6.888  3.07%  239+168   201+158   0.175  0.173   0.96%
911		ratio-regions  6.388  6.843 -6.65% 1182+257   981+300  120.142 120.345 -0.17%
912		ray  2.332  2.338 -0.29%  346+398   402+494   3.593  3.540   1.51%
913		simple  9.912  9.903  0.08% 1475+941  1579+1168  3.057  3.178  -3.83%
914		tsp  1.623  1.532  5.98%  266+200   250+211   8.045  7.878   2.12%
915		vliw 33.947 35.470 -4.29% 2629+2774 2877+3171  2.072  1.890   9.61%
916		----------------------------------------------------------------------------
917		Average                     1.22%                                     3.36%
918
919		----------------------------------------------------------------------
920		Name: Allen Leung
921		Date: 2000/03/23 16:25:00
922		Tag: leunga-20000323-fix_x86_alpha
923		Description:
924
925		1. X86 fixes/changes
926
927		a.  The old code generated for SETcc was completely wrong.
928		The Intel optimization guide is VERY misleading.
929
930		2. ALPHA fixes/changes
931
932		a.  Added the instructions LDBU, LDWU, STB, STW as per Fermin's suggestion.
933		b.  Added a new mode byteWordLoadStores to the functor parameter to Alpha()
934		c.  Added reassociation code for address computation.
935
936		----------------------------------------------------------------------
937		Name: Allen Leung
938		Date: 2000/03/22 01:23:00
939		Tag: leunga-20000322-fix_x86_hppa_ra
940		Description:
941
942		1. X86 fixes/changes
943
944		a.  x86Rewrite bug with MUL3 (found by Lal)
945		b.  Added the instructions FSTS, FSTL
946
947		2. PA-RISC fixes/changes
948
949		a.  B label should not be a delay slot candidate!  Why did this work?
950		b.  ADDT(32, REG(32, r), LI n) now generates one instruction instead of two,
951		as it should be.
952		c.  The assembly syntax for fstds and fstdd was wrong.
953		d.  Added the composite instruction COMICLR/LDO, which is the immediate
954		operand variant of COMCLR/LDO.
955
956		3. Generic MLRISC
957
958		a.  shuffle.sml rewritten to be slightly more efficient
959		b.  DIV bug in mltree-simplify fixed (found by Fermin)
960
961		4. Register Allocator
962
963		a.  I now release the interference graph earlier during spilling.
964		May improve memory usage.
965
966		----------------------------------------------------------------------
967		Name: Matthias Blume
968		Date: 2000/03/14 14:15:32
969		Tag: blume_main_v110p26p1_2
970		Description:
971
972		1. Tools.registerStdShellCmdTool (from smlnj/cm/tool.cm) takes an
973		additional argument called "template" which is an optional string that
974		specifiel the layout of the tool command line.  See the CM manual for
975		explanation.
976
977		2. A special-purpose tool can be "regisitered" by simply dropping the
978		corresponding <...>-tool.cm (and/or <...>-ext.cm) into the same
979		directory where the .cm file lives that uses this tool.  (The
980		behavior/misfeature until now was to look for the tool description
981		files in the current working directory.)  As before, tool description
982		files could also be anchored -- in which case they can live anywhere
983		they like.  Following the recent e-mail discussion, this change should
984		make it easier to have special-purpose tools that are shipped together
985		with the sources of the program that uses them.
986
987		----------------------------------------------------------------------
988	Name: Matthias Blume	Name: Matthias Blume
989	Date: 2000/03/10 07:48:34	Date: 2000/03/10 07:48:34
990	Tag: blume_main_v110p26p1_1	Tag: blume_main_v110p26p1_1

---

Colored Diff Long Colored Diff Unidiff Context Diff Side by Side

Legend: Removed from v.577   changed lines   Added in v.647

---

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