Diff of /sml/trunk/HISTORY

Parent Directory | Revision Log | Patch

	revision 1124, Thu Mar 7 19:49:22 2002 UTC	revision 1208, Tue May 21 02:22:43 2002 UTC
#	Line 13	Line 13
13	Description:	Description:
14
15	----------------------------------------------------------------------	----------------------------------------------------------------------
16		Name: Matthias Blume
17		Date: 2002/05/20 22:25:00 EDT
18		Tag: blume-20020520-controls
19		Description:
20
21		- Added ControlUtil structure to control-lib.cm.
22		- Use it throughout.
23		- Used Controls facility to define MLRISC controls (as opposed to
24		registering MLRISC control ref cells with Controls after the fact)
25		- Fixed messed-up controls priorities.
26
27		* Removed again all the stuff from config/preloads that one wouldn't
28		be able to preload at the time the initial heap image is built.
29		(Many libraries, e.g., CML, do not exist yet at this time.  The
30		only libraries that can be preloaded via config/preloads are those
31		that come bundled with the bootfiles.)
32
33		----------------------------------------------------------------------
34		Name: Matthias Blume
35		Date: 2002/05/20 10:59:00 EDT
36		Tag: blume-20020520-preloads
37		Description:
38
39		Added a lot of commented-out suggestions for things to be included
40		in config/preloads.
41
42		----------------------------------------------------------------------
43		Name: Allen Leung
44		Date: 2002/05/18 14:20:00 EDT
45		Tag: leunga-20020518-mdl
46		Description:
47
48		o Made the mdl tool stuff compile and run again.
49		o I've disabled all the stuff that depends on RTL specifications; they
50		are all badly broken anyway.
51
52		----------------------------------------------------------------------
53		Name: Matthias Blume
54		Date: 2002/05/17 16:49:00 EDT
55		Tag: blume-20020517-controls
56		Description:
57
58		0. John Reppy made several modifications to the SML/NJ library.
59		In particular, there is a shiny new controls-lib.cm.
60
61		1. Pushed new controls interface through compiler so that everything
62		compiles again.
63
64		2. Added FormatComb and FORMAT_COMB to the CML version of the
65		SML/NJ library (so that CML compiles again).
66
67		3. Modified init scripts because XXX_DEFAULT environment variables
68		are no longer with us.  (Boot-time initialization is now done
69		using the same environment variables that are also used for
70		startup-time initialization of controls.)
71
72		----------------------------------------------------------------------
73		Name: Lal George
74		Date: 2002/05/15 09:20:10 EDT
75		Tag: george-20020515-pseudo-op-decls
76		Description:
77
78		All pseudo-ops emitted before the first segment declaration
79		such as TEXT, DATA, and BSS directives are assumed to be global
80		declarations and are emitted first in the assembly file. This is
81		useful in a number of situations where one has pseudo-ops that are not
82		specific to any segment, and also works around the constraint that one
83		cannot have client pseudo-ops in the TEXT segment.
84
85		Because no segment is associated with these declarations it is
86		an error to allocate any space or objects before the first segment
87		directive and an exception will be raised. However, we cannot make
88		this check for client pseudo-ops.
89
90		These top level declarations are a field in the CFG graph_info.
91		In theory you can continue to add to this field after the CFG has been
92		built -- provided you know what you are doing;-)
93
94		----------------------------------------------------------------------
95		Name: Matthias Blume
96		Date: 2002/05/13 16:40:00 EDT
97		Tag: blume-20020513-pp-etc
98		Description:
99
100		A few minor bugfixes:
101
102		- Stopgap measure for bug recently reported by Elsa Gunter (ppDec).
103		(Bogus printouts for redefined bindings still occur.  Compiler
104		bug should no longer occur now.  We need to redo the prettyprinter
105		from scratch.)
106
107		- CM pathname printer now also adds escape sequences for ( and )
108
109		- commend and docu fixes for ml-nlffi
110
111		----------------------------------------------------------------------
112		Name: Matthias Blume
113		Date: 2002/05/10 16:40:00 EDT
114		Tag: blume-20020510-erg-textio
115		Description:
116
117		Applied the following bugfix provided by Emden Gansner:
118
119		Output is corrupted when outputSubstr is used rather than output.
120
121		The problem occurs when a substring
122
123		ss = (s, dataStart, dataLen)
124
125		where dataStart > 0, fills a stream buffer with avail bytes left.
126		avail bytes of s, starting at index dataStart, are copied into the
127		buffer, the buffer is flushed, and then the remaining dataLen-avail
128		bytes of ss are copied into the beginning of the buffer. Instead of
129		starting this copy at index dataStart+avail in s, the current code
130		starts the copy at index avail.
131
132		Fix:
133		In text-io-fn.sml, change line 695 from
134		val needsFlush = copyVec(v, avail, dataLen-avail, buf, 0)
135		to
136		val needsFlush = copyVec(v, dataStart+avail, dataLen-avail, buf, 0)
137
138		----------------------------------------------------------------------
139		Name: Matthias Blume
140		Date: 2002/04/12 13:55:00 EDT
141		Tag: blume-20020412-assyntax
142		Description:
143
144		1. Grabbed newer assyntax.h from the XFree86 project.
145		2. Fiddled with how to compile X86.prim.asm without warnings.
146		3. (Very) Minor cleanup in CM.
147
148		----------------------------------------------------------------------
149		Name: Matthias Blume
150		Date: 2002/04/01 (no joke!) 17:07:00 EST
151		Tag: blume-20020401-x86div
152		Description:
153
154		Added full support for div/mod/rem/quot on the x86, using the machine
155		instruction's two results (without clumsily recomputing the remainder)
156		directly where appropriate.
157
158		Some more extensive power-of-two support was added to the x86 instruction
159		selector (avoiding expensive divs, mods, and muls where they can be
160		replaced with cheaper shifts and masks).  However, this sort of thing
161		ought to be done earlier, e.g., within the CPS optimizer so that
162		all architectures benefit from it.
163
164		The compiler compiles to a fixed point, but changes might be somewhat
165		fragile nevertheless.  Please, report any strange things that you might
166		see wrt. div/mod/quot/rem...
167
168		----------------------------------------------------------------------
169		Name: Matthias Blume
170		Date: 2002/03/29 17:22:00
171		Tag: blume-20020329-div
172		Description:
173
174		Fixed my broken div/mod logic.  Unfortunately, this means that the
175		inline code for div/mod now has one more comparison than before.
176		Fast paths (quotient > 0 or remainder = 0) are not affected, though.
177		The problem was with quotient = 0, because that alone does not tell
178		us which way the rounding went.  One then has to look at whether
179		remainder and divisor have the same sign...  :(
180
181		Anyway, I replaced the bootfiles with fresh ones...
182
183		----------------------------------------------------------------------
184		Name: Matthias Blume
185		Date: 2002/03/29 14:10:00 EST
186		Tag: blume-20020329-inlprims
187		Description:
188
189		NEW BOOTFILES!!!    Version number bumped to 110.39.3.
190
191		Primops have changed. This means that the bin/boot-file formats have
192		changed as well.
193
194		To make sure that there is no confusion, I made a new version.
195
196
197		CHANGES:
198
199		* removed REMT from mltree (remainder should never overflow).
200
201		* added primops to deal with divisions of all flavors to the frontend
202
203		* handled these primops all the way through so they map to their respective
204		MLRISC support
205
206		* used these primops in the implementation of Int, Int32, Word, Word32
207
208		* removed INLDIV, INLMOD, and INLREM as they are no longer necessary
209
210		* parameterized INLMIN, INLMAX, and INLABS by a numkind
211
212		* translate.sml now deals with all flavors of INL{MIN,MAX,ABS}, including
213		floating point
214
215		* used INL{MIN,MAX,ABS} in the implementation of Int, Int32, Word, Word32,
216		and Real (but Real.abs maps to a separate floating-point-only primop)
217
218
219		TODO items:
220
221		* Hacked Alpha32 instruction selection, disabling the selection of REMx
222		instructions because the machine instruction encoder cannot handle
223		them.  (Hppa, PPC, and Sparc instruction selection did not handle
224		REM in the first place, and REM is supported by the x86 machine coder.)
225
226		* Handle DIV and MOD with DIV_TO_NEGINF directly in the x86 instruction
227		selection phase.  (The two can be streamlined because the hardware
228		delivers both quotient and remainder at the same time anyway.)
229
230		* Think about what to do with "valOf(Int32.minInt) div ~1" and friends.
231		(Currently the behavior is inconsistent both across architectures and
232		wrt. the draft Basis spec.)
233
234		* Word8 should eventually be handled natively, too.
235
236		* There seems to be one serious bug in mltree-gen.sml.  It appears, though,
237		as if there currently is no execution path that could trigger it in
238		SML/NJ.  (The assumptions underlying functions arith and promotable do not
239		hold for things like multiplication and division.)
240
241		----------------------------------------------------------------------
242		Name: Matthias Blume
243		Date: 2002/03/27 16:27:00 EST
244		Tag: blume-20020327-mlrisc-divisions
245		Description:
246
247		Added support for all four division operations (ML's div, mod, quot,
248		and rem) to MLRISC.  In the course of doing so, I also rationalized
249		the naming (no more annoying switch-around of DIV and QUOT), by
250		parameterizing the operation by div_rounding_mode (which can be either
251		DIV_TO_ZERO or DIV_TO_NEGINF).
252
253		The generic MLTreeGen functor takes care of compiling all four
254		operations down to only round-to-zero div.
255
256		Missing pieces:
257
258		* Doing something smarter than relying on MLTreeGen on architectures
259		like, e.g., the x86 where hardware division delivers both quotient and
260		remainder at the same time.  With this, the implementation of the
261		round-to-neginf operations could be further streamlined.
262
263		* Remove inlining support for div/mod/rem from the frontend and replace it
264		with primops that get carried through to the backend.  Do this for all
265		int and word types.
266
267		----------------------------------------------------------------------
268		Name: Matthias Blume
269		Date: 2002/03/25 17:25:00 EST
270		Tag: blume-20020325-divmod
271		Description:
272
273		I improved (hopefully without breaking them) the implementation of Int.div,
274		Int.mod, and Int.rem.   For this, the code in translate.sml now takes
275		advantage of the following observations:
276
277		Let  q = x quot y      r = x rem y
278		d = x div  y      m = x mod y
279
280		where "quot" is the round-to-zero version of integer division that
281		hardware usually provides.  Then we have:
282
283		r = x - q * y        where neither the * nor the - will overflow
284		d = if q >= 0 orelse x = q * y then q else q - 1
285		where neither the * nor the - will overflow
286		m = if q >= 0 orelse r = 0 then r else r + y
287		where the + will not overflow
288
289		This results in substantial simplification of the generated code.
290		The following table shows the number of CFG nodes and edges generated
291		for
292		fun f (x, y) = x OPER y
293		(* with OPER \in div, mod, quot, rem *)
294
295
296		OPER | nodes(old) | edges(old) | nodes(new) | edges(new)
297		--------------------------------------------------------
298		div |         24 |         39 |         12 |         16
299		mod |         41 |         71 |         12 |         16
300		quot |          8 |         10 |          8 |         10
301		rem |         10 |         14 |          8 |         10
302
303
304		----------------------------------------------------------------------
305		Name: Matthias Blume
306		Date: 2002/03/25 22:06:00 EST
307		Tag: blume-20020325-cprotobug
308		Description:
309
310		Fixed a bug in cproto (c prototype decoder).
311
312		----------------------------------------------------------------------
313		Name: Matthias Blume
314		Date: 2002/03/25 16:00:00 EST
315		Tag: blume-20020325-raw-primops
316		Description:
317
318		I did some cleanup to Allen's new primop code and
319		replaced yesterday's bootfiles with new ones.
320		(But they are stored in the same place.)
321
322		----------------------------------------------------------------------
323		Name: Matthias Blume
324		Date: 2002/03/24 22:40:00 EST
325		Tag: blume-20020324-bootfiles
326		Description:
327
328		Made the bootfiles that Allen asked for.
329
330		----------------------------------------------------------------------
331		Name: Allen Leung
332		Date: 2002/03/23 15:50:00 EST
333		Tag: leunga-20020323-flint-cps-rcc-primops
334		Description:
335
336		1. Changes to FLINT primops:
337
338		(* make a call to a C-function;
339		* The primop carries C function prototype information and specifies
340		* which of its (ML-) arguments are floating point. C prototype
341		* information is for use by the backend, ML information is for
342		* use by the CPS converter. *)
343		| RAW_CCALL of { c_proto: CTypes.c_proto,
344		ml_args: ccall_type list,
345		ml_res_opt: ccall_type option,
346		reentrant : bool
347		} option
348		(* Allocate uninitialized storage on the heap.
349		* The record is meant to hold short-lived C objects, i.e., they
350		* are not ML pointers.  With the tag, the representation is
351		* the same as RECORD with tag tag_raw32 (sz=4), or tag_fblock (sz=8)
352		*)
353		| RAW_RECORD of {tag:bool,sz:int}
354		and ccall_type = CCALL_INT32 | CCALL_REAL64 | CCALL_ML_PTR
355
356		2.  These CPS primops are now overloaded:
357
358		rawload of {kind:numkind}
359		rawstore of {kind:numkind}
360
361		The one argument form is:
362
363		rawload {kind} address
364
365		The two argument form is:
366
367		rawload {kind} [ml object, byte-offset]
368
369		3. RAW_CCALL/RCC now takes two extra arguments:
370
371		a. The first is whether the C call is reentrant, i.e., whether
372		ML state should be saved and restored.
373		b. The second argument is a string argument specifying the name of
374		library and the C function.
375
376		These things are currently not handled in the code generator, yet.
377
378		4. In CProto,
379
380		An encoding type of "bool" means "ml object" and is mapped into
381		C prototype of PTR.  Note that "bool" is different than "string",
382		even though "string" is also mapped into PTR, because "bool"
383		is assigned an CPS type of BOGt, while "string" is assigned INT32t.
384
385		5. Pickler/unpicker
386
387		Changed to handle RAW_RECORD and newest RAW_CCALL
388
389		6. MLRiscGen,
390
391		1. Changed to handle the new rawload/rawstore/rawrecord operators.
392		2. Code for handling C Calls has been moved to a new module CPSCCalls,
393		in the file CodeGen/cpscompile/cps-c-calls.sml
394
395		7. Added the conditional move operator
396
397		condmove of branch
398
399		to cps.  Generation of this is still buggy so it is currently
400		disabled.
401
402		----------------------------------------------------------------------
403		Name: Lal George
404		Date: 2002/03/22 14:18:25 EST
405		Tag: george-20020322-cps-branch-prob
406		Description:
407
408		Implemented the Ball-Larus branch prediction-heuristics, and
409		incorporated graphical viewers for control flow graphs.
410
411		Ball-Larus Heuristics:
412		---------------------
413		See the file compiler/CodeGen/cpscompile/cpsBranchProb.sml.
414
415		By design it uses the Dempster-Shafer theory for combining
416		probabilities.  For example, in the function:
417
418		fun f(n,acc) = if n = 0 then acc else f(n-1, n*acc)
419
420		the ball-larus heuristics predicts that the n=0 is unlikely
421		(OH-heuristic), and the 'then' branch is unlikely because of the
422		RH-heuristic -- giving the 'then' branch an even lower combined
423		probability using the Dempster-Shafer theory.
424
425		Finally, John Reppy's loop analysis in MLRISC, further lowers the
426		probability of the 'then' branch because of the loop in the else
427		branch.
428
429
430		Graphical Viewing:
431		------------------
432		I merely plugged in Allen's graphical viewers into the compiler. The
433		additional code is not much. At the top level, saying:
434
435		Control.MLRISC.getFlag "cfg-graphical-view" := true;
436
437		will display the graphical view of the control flow graph just before
438		back-patching.  daVinci must be in your path for this to work. If
439		daVinci is not available, then the default viewer can be changed
440		using:
441
442		Control.MLRISC.getString "viewer"
443
444		which can be set to "dot" or "vcg" for the corresponding viewers. Of
445		course, these viewers must be in your path.
446
447		The above will display the compilation unit at the level of clusters,
448		many of which are small, boring, and un-interesting. Also setting:
449
450		Control.MLRISC.getInt "cfg-graphical-view_size"
451
452		will display clusters that are larger than the value set by the above.
453
454
455		----------------------------------------------------------------------
456		Name: Matthias Blume
457		Date: 2002/03/21 22:20:00 EST
458		Tag: blume-20020321-kmp-bugfix
459		Description:
460
461		Changed the interface to the KMP routine in PreString and fixed
462		a minor bug in one place where it was used.
463
464		----------------------------------------------------------------------
465		Name: Allen Leung
466		Date: 2002/03/21 20:30:00 EST
467		Tag: leunga-20020321-cfg
468		Description:
469
470		Fixed a potential problem in cfg edge splitting.
471
472		----------------------------------------------------------------------
473		Name: Allen Leung
474		Date: 2002/03/21 17:15:00 EST
475		Tag: leunga-20020321-x86-fp-cfg
476		Description:
477
478		1. Recoded the buggy parts of x86-fp.
479
480		a. All the block reordering code has been removed.
481		We now depend on the block placement phases to do this work.
482
483		b. Critical edge splitting code has been simplified and moved into the
484		CFG modules, as where they belong.
485
486		Both of these were quite buggy and complex.  The code is now much, much
487		simpler.
488
489		2. X86 backend.
490
491		a. Added instructions for 64-bit support.  Instruction selection for
492		64-bit has not been committed, however, since that
493		requires changes to MLTREE which haven't been approved by
494		Lal and John.
495
496		b. Added support for FUCOMI and FUCOMIP when generating code for
497		PentiumPro and above.  We only generate these instructions in
498		the fast-fp mode.
499
500		c. Added cases for JP and JNP in X86FreqProps.
501
502		3. CFG
503
504		CFG now has a bunch of methods for edge splitting and merging.
505
506		4. Machine description.
507
508		John's simplification of MLTREE_BASIS.fcond broke a few machine
509		description things:
510
511		rtl-build.{sig,sml} and hppa.mdl fixed.
512
513		NOTE: the machine description stuff in the repository is still broken.
514		Again, I can't put my fixes in because that involves
515		changes to MLTREE.
516
517		----------------------------------------------------------------------
518		Name: Matthias Blume
519		Date: 2002/03/20 15:55:00 EST
520		Tag: blume-20020320-kmp
521		Description:
522
523		Implemented Knuth-Morris-Pratt string matching in PreString and used
524		it for String.isSubstring, Substring.isSubstring, and
525		Substring.position.
526
527		(Might need some stress-testing.  Simple examples worked fine.)
528
529		----------------------------------------------------------------------
530		Name: Matthias Blume
531		Date: 2002/03/19 16:37:00 EST
532		Tag: blume-20020319-witnesses
533		Description:
534
535		Added a structure C.W and functions convert/Ptr.convert to ml-nlffi-lib.
536
537		This implements a generic mechanism for changing constness qualifiers
538		anywhere within big C types without resorting to outright "casts".
539		(So far, functions such as C.rw/C.ro or C.Ptr.rw/C.Ptr.ro only let you
540		modify the constness at the outermost level.)
541		The implementation of "convert" is based on the idea of "witness"
542		values -- values that are not used by the operation but whose types
543		"testify" to their applicability.  On the implementation side, "convert"
544		is simply a projection (returning its second curried argument).  With
545		cross-module inlining, it should not result in any machine code being
546		generated.
547
548		----------------------------------------------------------------------
549		Name: Matthias Blume
550		Date: 2002/03/15 16:40:00 EST
551		Tag: blume-20020315-basis
552		Description:
553
554		Provided (preliminary?) implementations for
555
556		{String,Substring}.{concatWith,isSuffix,isSubstring}
557
558		and
559
560		Substring.full
561
562		Those are in the Basis spec but they were missing in SML/NJ.
563
564		----------------------------------------------------------------------
565		Name: Matthias Blume
566		Date: 2002/03/14 21:30:00 EST
567		Tag: blume-20020314-controls
568		Description:
569
570		Controls:
571		---------
572
573		1. Factored out the recently-added Controls : CONTROLS stuff and put
574		it into its own library $/controls-lib.cm.  The source tree for
575		this is under src/smlnj-lib/Controls.
576
577		2. Changed the names of types and functions in this interface, so they
578		make a bit more "sense":
579
580		module -> registry
581		'a registry -> 'a group
582
583		3. The interface now deals in ref cells only.  The getter/setter interface
584		is (mostly) gone.
585
586		4. Added a function that lets one register an already-existing ref cell.
587
588		5. Made the corresponding modifications to the rest of the code so that
589		everything compiles again.
590
591		6. Changed the implementation of Controls.MLRISC back to something closer
592		to the original.  In particular, this module (and therefore MLRISC)
593		does not depend on Controls.  There now is some link-time code in
594		int-sys.sml that registers the MLRISC controls with the Controls
595		module.
596
597		CM:
598		---
599
600		* One can now specify the lambda-split aggressiveness in init.cmi.
601
602		----------------------------------------------------------------------
603		Name: Allen Leung
604		Date: 2002/03/13 17:30:00 EST
605		Tag: leunga-20020313-x86-fp-unary
606		Description:
607
608		Bug fix for:
609
610		> leunga@weaselbane:~/Yale/tmp/sml-dist{21} bin/sml
611		> Standard ML of New Jersey v110.39.1 [FLINT v1.5], March 08, 2002
612		> - fun f(x,(y,z)) = Real.~ y;
613		> [autoloading]
614		> [autoloading done]
615		>       fchsl   (%eax), 184(%esp)
616		> Error: MLRisc bug: X86MCEmitter.emitInstr
617		>
618		> uncaught exception Error
619		>   raised at: ../MLRISC/control/mlriscErrormsg.sml:16.14-16.19
620
621		The problem was that the code generator did not generate any fp registers
622		in this case, and the ra didn't know that it needed to run the X86FP phase to
623		translate the pseudo fp instruction.   This only happened with unary fp
624		operators in certain situations.
625
626		----------------------------------------------------------------------
627		Name: Matthias Blume
628		Date: 2002/03/13 14:00:00 EST
629		Tag: blume-20020313-overload-etc
630		Description:
631
632		1. Added _overload as a synonym for overload for backward compatibility.
633		(Control.overloadKW must be true for either version to be accepted.)
634
635		2. Fixed bug in install script that caused more things to be installed
636		than what was requested in config/targets.
637
638		3. Made CM aware of the (_)overload construct so that autoloading
639		works.
640
641		----------------------------------------------------------------------
642		Name: Matthias Blume
643		Date: 2002/03/12 22:03:00 EST
644		Tag: blume-20020312-url
645		Description:
646
647		Forgot to update BOOT and srcarchiveurl.
648
649		----------------------------------------------------------------------
650		Name: Matthias Blume
651		Date: 2002/03/12 17:30:00 EST
652		Tag: blume-20020312-version110392
653		Description:
654
655		Yet another version number bump (because of small changes to the
656		binfile format).  Version number is now 110.39.2.  NEW BOOTFILES!
657
658		Changes:
659
660		The new pid generation scheme described a few weeks ago was overly
661		complicated.  I implemented a new mechanism that is simpler and
662		provides a bit more "stability":  Once CM has seen a compilation
663		unit, it keeps its identity constant (as long as you do not delete
664		those crucial CM/GUID/* files).  This means that when you change
665		an interface, compile, then go back to the old interface, and
666		compile again, you arrive at the original pid.
667
668		There now also is a mechanism that instructs CM to use the plain
669		environment hash as a module's pid (effectively making its GUID
670		the empty string).  For this, "noguid" must be specified as an
671		option to the .sml file in question within its .cm file.
672		This is most useful for code that is being generated by tools such
673		as ml-nlffigen (because during development programmers tend to
674		erase the tool's entire output directory tree including CM's cached
675		GUIDs).  "noguid" is somewhat dangerous (since it can be used to locally
676		revert to the old, broken behavior of SML/NJ, but in specific cases
677		where there is no danger of interface confusion, its use is ok
678		(I think).
679
680		ml-nlffigen by default generates "noguid" annotations.  They can be
681		turned off by specifying -guid in its command line.
682
683		----------------------------------------------------------------------
684		Name: Lal George
685		Date: 2002/03/12 12 14:42:36 EST
686		Tag: george-20020312-frequency-computation
687		Description:
688
689		Integrated jump chaining and static block frequency into the
690		compiler. More details and numbers later.
691
692		----------------------------------------------------------------------
693		Name: Lal George
694		Date: 2002/03/11 11 22:38:53 EST
695		Tag: george-20020311-jump-chain-elim
696		Description:
697
698		Tested the jump chain elimination on all architectures (except the
699		hppa).  This is on by default right now and is profitable for the
700		alpha and x86, however, it may not be profitable for the sparc and ppc
701		when compiling the compiler.
702
703		The gc test will typically jump to a label at the end of the cluster,
704		where there is another jump to an external cluster containing the actual
705		code to invoke gc. This is to allow factoring of common gc invocation
706		sequences. That is to say, we generate:
707
708		f:
709		testgc
710		ja   L1      % jump if above to L1
711
712		L1:
713		jmp L2
714
715
716		After jump chain elimination the 'ja L1' instructions is converted to
717		'ja L2'. On the sparc and ppc, many of the 'ja L2' instructions may end
718		up being implemented in their long form (if L2 is far away) using:
719
720		jbe     L3      % jump if below or equal to L3
721		jmp     L2
722		L3:
723		...
724
725
726		For large compilation units L2  may be far away.
727
728
729		----------------------------------------------------------------------
730		Name: Matthias Blume
731		Date: 2002/03/11 13:30:00 EST
732		Tag: blume-20020311-mltreeeval
733		Description:
734
735		A functor parameter was missing.
736
737		----------------------------------------------------------------------
738		Name: Allen Leung
739		Date: 2002/03/11 10:30:00 EST
740		Tag: leunga-20020311-runtime-string0
741		Description:
742
743		The representation of the empty string now points to a
744		legal null terminated C string instead of unit.  It is now possible
745		to convert an ML string into C string with InlineT.CharVector.getData.
746		This compiles into one single machine instruction.
747
748		----------------------------------------------------------------------
749		Name: Allen Leung
750		Date: 2002/03/10 23:55:00 EST
751		Tag: leunga-20020310-x86-call
752		Description:
753
754		Added machine generation for CALL instruction (relative displacement mode)
755
756		----------------------------------------------------------------------
757		Name: Matthias Blume
758		Date: 2002/03/08 16:05:00
759		Tag: blume-20020308-entrypoints
760		Description:
761
762		Version number bumped to 110.39.1.  NEW BOOTFILES!
763
764		Entrypoints: non-zero offset into a code object where execution should begin.
765
766		- Added the notion of an entrypoint to CodeObj.
767		- Added reading/writing of entrypoint info to Binfile.
768		- Made runtime system bootloader aware of entrypoints.
769		- Use the address of the label of the first function given to mlriscGen
770		as the entrypoint.  This address is currently always 0, but it will
771		not be 0 once we turn on block placement.
772		- Removed the linkage cluster code (which was The Other Way(tm) of dealing
773		with entry points) from mlriscGen.
774
775		----------------------------------------------------------------------
776		Name: Allen Leung
777		Date: 2002/03/07 20:45:00 EST
778		Tag: leunga-20020307-x86-cmov
779		Description:
780
781		Bug fixes for CMOVcc on x86.
782
783		1. Added machine code generation for CMOVcc
784		2. CMOVcc is now generated in preference over SETcc on PentiumPro or above.
785		3. CMOVcc cannot have an immediate operand as argument.
786
787		----------------------------------------------------------------------
788		Name: Matthias Blume
789		Date: 2002/03/07 16:15:00 EST
790		Tag: blume-20020307-controls
791		Description:
792
793		This is a very large but mostly boring patch which makes (almost)
794		every tuneable compiler knob (i.e., pretty much everything under
795		Control.* plus a few other things) configurable via both the command
796		line and environment variables in the style CM did its configuration
797		until now.
798
799		Try starting sml with '-h' (or, if you are brave, '-H')
800
801		To this end, I added a structure Controls : CONTROLS to smlnj-lib.cm which
802		implements the underlying generic mechanism.
803
804		The interface to some of the existing such facilities has changed somewhat.
805		For example, the MLRiscControl module now provides mkFoo instead of getFoo.
806		(The getFoo interface is still there for backward-compatibility, but its
807		use is deprecated.)
808
809		The ml-build script passes -Cxxx=yyy command-line arguments through so
810		that one can now twiddle the compiler settings when using this "batch"
811		compiler.
812
813		TODO items:
814
815		We should go through and throw out all controls that are no longer
816		connected to anything.  Moreover, we should go through and provide
817		meaningful (and correct!) documentation strings for those controls
818		that still are connected.
819
820		Currently, multiple calls to Controls.new are accepted (only the first
821		has any effect).  Eventually we should make sure that every control
822		is being made (via Controls.new) exactly once.  Future access can then
823		be done using Controls.acc.
824
825		Finally, it would probably be a good idea to use the getter-setter
826		interface to controls rather than ref cells.  For the time being, both
827		styles are provided by the Controls module, but getter-setter pairs are
828		better if thread-safety is of any concern because they can be wrapped.
829
830		*****************************************
831
832		One bug fix: The function blockPlacement in three of the MLRISC
833		backpatch files used to be hard-wired to one of two possibilities at
834		link time (according to the value of the placementFlag).  But (I
835		think) it should rather sense the flag every time.
836
837		*****************************************
838
839		Other assorted changes (by other people who did not supply a HISTORY entry):
840
841		1. the cross-module inliner now works much better (Monnier)
842		2. representation of weights, frequencies, and probabilities in MLRISC
843		changed in preparation of using those for weighted block placement
844		(Reppy, George)
845
846		----------------------------------------------------------------------
847	Name: Lal George	Name: Lal George
848	Date: 2002/03/07 14:44:24 EST 2002	Date: 2002/03/07 14:44:24 EST 2002
849	Tag: george-20020307-weighted-block-placement	Tag: george-20020307-weighted-block-placement

---

Colored Diff Long Colored Diff Unidiff Context Diff Side by Side

Legend: Removed from v.1124   changed lines   Added in v.1208

---

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