Diff of /sml/trunk/HISTORY

Parent Directory | Revision Log | Patch

	revision 1073, Fri Feb 15 22:07:38 2002 UTC	revision 1158, Fri Mar 22 01:21:25 2002 UTC
#	Line 13	Line 13
13	Description:	Description:
14
15	----------------------------------------------------------------------	----------------------------------------------------------------------
16		Name: Allen Leung
17		Date: 2002/03/21 20:30:00 EST
18		Tag: leunga-20020321-cfg
19		Description:
20
21		Fixed a potential problem in cfg edge splitting.
22
23		----------------------------------------------------------------------
24		Name: Allen Leung
25		Date: 2002/03/21 17:15:00 EST
26		Tag: leunga-20020321-x86-fp-cfg
27		Description:
28
29		1. Recoded the buggy parts of x86-fp.
30
31		a. All the block reordering code has been removed.
32		We now depend on the block placement phases to do this work.
33
34		b. Critical edge splitting code has been simplified and moved into the
35		CFG modules, as where they belong.
36
37		Both of these were quite buggy and complex.  The code is now much, much
38		simpler.
39
40		2. X86 backend.
41
42		a. Added instructions for 64-bit support.  Instruction selection for
43		64-bit has not been committed, however, since that
44		requires changes to MLTREE which haven't been approved by
45		Lal and John.
46
47		b. Added support for FUCOMI and FUCOMIP when generating code for
48		PentiumPro and above.  We only generate these instructions in
49		the fast-fp mode.
50
51		c. Added cases for JP and JNP in X86FreqProps.
52
53		3. CFG
54
55		CFG now has a bunch of methods for edge splitting and merging.
56
57		4. Machine description.
58
59		John's simplification of MLTREE_BASIS.fcond broke a few machine
60		description things:
61
62		rtl-build.{sig,sml} and hppa.mdl fixed.
63
64		NOTE: the machine description stuff in the repository is still broken.
65		Again, I can't put my fixes in because that involves
66		changes to MLTREE.
67
68		----------------------------------------------------------------------
69		Name: Matthias Blume
70		Date: 2002/03/20 15:55:00 EST
71		Tag: blume-20020320-kmp
72		Description:
73
74		Implemented Knuth-Morris-Pratt string matching in PreString and used
75		it for String.isSubstring, Substring.isSubstring, and
76		Substring.position.
77
78		(Might need some stress-testing.  Simple examples worked fine.)
79
80		----------------------------------------------------------------------
81		Name: Matthias Blume
82		Date: 2002/03/19 16:37:00 EST
83		Tag: blume-20020319-witnesses
84		Description:
85
86		Added a structure C.W and functions convert/Ptr.convert to ml-nlffi-lib.
87
88		This implements a generic mechanism for changing constness qualifiers
89		anywhere within big C types without resorting to outright "casts".
90		(So far, functions such as C.rw/C.ro or C.Ptr.rw/C.Ptr.ro only let you
91		modify the constness at the outermost level.)
92		The implementation of "convert" is based on the idea of "witness"
93		values -- values that are not used by the operation but whose types
94		"testify" to their applicability.  On the implementation side, "convert"
95		is simply a projection (returning its second curried argument).  With
96		cross-module inlining, it should not result in any machine code being
97		generated.
98
99		----------------------------------------------------------------------
100		Name: Matthias Blume
101		Date: 2002/03/15 16:40:00 EST
102		Tag: blume-20020315-basis
103		Description:
104
105		Provided (preliminary?) implementations for
106
107		{String,Substring}.{concatWith,isSuffix,isSubstring}
108
109		and
110
111		Substring.full
112
113		Those are in the Basis spec but they were missing in SML/NJ.
114
115		----------------------------------------------------------------------
116		Name: Matthias Blume
117		Date: 2002/03/14 21:30:00 EST
118		Tag: blume-20020314-controls
119		Description:
120
121		Controls:
122		---------
123
124		1. Factored out the recently-added Controls : CONTROLS stuff and put
125		it into its own library $/controls-lib.cm.  The source tree for
126		this is under src/smlnj-lib/Controls.
127
128		2. Changed the names of types and functions in this interface, so they
129		make a bit more "sense":
130
131		module -> registry
132		'a registry -> 'a group
133
134		3. The interface now deals in ref cells only.  The getter/setter interface
135		is (mostly) gone.
136
137		4. Added a function that lets one register an already-existing ref cell.
138
139		5. Made the corresponding modifications to the rest of the code so that
140		everything compiles again.
141
142		6. Changed the implementation of Controls.MLRISC back to something closer
143		to the original.  In particular, this module (and therefore MLRISC)
144		does not depend on Controls.  There now is some link-time code in
145		int-sys.sml that registers the MLRISC controls with the Controls
146		module.
147
148		CM:
149		---
150
151		* One can now specify the lambda-split aggressiveness in init.cmi.
152
153		----------------------------------------------------------------------
154		Name: Allen Leung
155		Date: 2002/03/13 17:30:00 EST
156		Tag: leunga-20020313-x86-fp-unary
157		Description:
158
159		Bug fix for:
160
161		> leunga@weaselbane:~/Yale/tmp/sml-dist{21} bin/sml
162		> Standard ML of New Jersey v110.39.1 [FLINT v1.5], March 08, 2002
163		> - fun f(x,(y,z)) = Real.~ y;
164		> [autoloading]
165		> [autoloading done]
166		>       fchsl   (%eax), 184(%esp)
167		> Error: MLRisc bug: X86MCEmitter.emitInstr
168		>
169		> uncaught exception Error
170		>   raised at: ../MLRISC/control/mlriscErrormsg.sml:16.14-16.19
171
172		The problem was that the code generator did not generate any fp registers
173		in this case, and the ra didn't know that it needed to run the X86FP phase to
174		translate the pseudo fp instruction.   This only happened with unary fp
175		operators in certain situations.
176
177		----------------------------------------------------------------------
178		Name: Matthias Blume
179		Date: 2002/03/13 14:00:00 EST
180		Tag: blume-20020313-overload-etc
181		Description:
182
183		1. Added _overload as a synonym for overload for backward compatibility.
184		(Control.overloadKW must be true for either version to be accepted.)
185
186		2. Fixed bug in install script that caused more things to be installed
187		than what was requested in config/targets.
188
189		3. Made CM aware of the (_)overload construct so that autoloading
190		works.
191
192		----------------------------------------------------------------------
193		Name: Matthias Blume
194		Date: 2002/03/12 22:03:00 EST
195		Tag: blume-20020312-url
196		Description:
197
198		Forgot to update BOOT and srcarchiveurl.
199
200		----------------------------------------------------------------------
201		Name: Matthias Blume
202		Date: 2002/03/12 17:30:00 EST
203		Tag: blume-20020312-version110392
204		Description:
205
206		Yet another version number bump (because of small changes to the
207		binfile format).  Version number is now 110.39.2.  NEW BOOTFILES!
208
209		Changes:
210
211		The new pid generation scheme described a few weeks ago was overly
212		complicated.  I implemented a new mechanism that is simpler and
213		provides a bit more "stability":  Once CM has seen a compilation
214		unit, it keeps its identity constant (as long as you do not delete
215		those crucial CM/GUID/* files).  This means that when you change
216		an interface, compile, then go back to the old interface, and
217		compile again, you arrive at the original pid.
218
219		There now also is a mechanism that instructs CM to use the plain
220		environment hash as a module's pid (effectively making its GUID
221		the empty string).  For this, "noguid" must be specified as an
222		option to the .sml file in question within its .cm file.
223		This is most useful for code that is being generated by tools such
224		as ml-nlffigen (because during development programmers tend to
225		erase the tool's entire output directory tree including CM's cached
226		GUIDs).  "noguid" is somewhat dangerous (since it can be used to locally
227		revert to the old, broken behavior of SML/NJ, but in specific cases
228		where there is no danger of interface confusion, its use is ok
229		(I think).
230
231		ml-nlffigen by default generates "noguid" annotations.  They can be
232		turned off by specifying -guid in its command line.
233
234		----------------------------------------------------------------------
235		Name: Lal George
236		Date: 2002/03/12 12 14:42:36 EST
237		Tag: george-20020312-frequency-computation
238		Description:
239
240		Integrated jump chaining and static block frequency into the
241		compiler. More details and numbers later.
242
243		----------------------------------------------------------------------
244		Name: Lal George
245		Date: 2002/03/11 11 22:38:53 EST
246		Tag: george-20020311-jump-chain-elim
247		Description:
248
249		Tested the jump chain elimination on all architectures (except the
250		hppa).  This is on by default right now and is profitable for the
251		alpha and x86, however, it may not be profitable for the sparc and ppc
252		when compiling the compiler.
253
254		The gc test will typically jump to a label at the end of the cluster,
255		where there is another jump to an external cluster containing the actual
256		code to invoke gc. This is to allow factoring of common gc invocation
257		sequences. That is to say, we generate:
258
259		f:
260		testgc
261		ja   L1      % jump if above to L1
262
263		L1:
264		jmp L2
265
266
267		After jump chain elimination the 'ja L1' instructions is converted to
268		'ja L2'. On the sparc and ppc, many of the 'ja L2' instructions may end
269		up being implemented in their long form (if L2 is far away) using:
270
271		jbe     L3      % jump if below or equal to L3
272		jmp     L2
273		L3:
274		...
275
276
277		For large compilation units L2  may be far away.
278
279
280		----------------------------------------------------------------------
281		Name: Matthias Blume
282		Date: 2002/03/11 13:30:00 EST
283		Tag: blume-20020311-mltreeeval
284		Description:
285
286		A functor parameter was missing.
287
288		----------------------------------------------------------------------
289		Name: Allen Leung
290		Date: 2002/03/11 10:30:00 EST
291		Tag: leunga-20020311-runtime-string0
292		Description:
293
294		The representation of the empty string now points to a
295		legal null terminated C string instead of unit.  It is now possible
296		to convert an ML string into C string with InlineT.CharVector.getData.
297		This compiles into one single machine instruction.
298
299		----------------------------------------------------------------------
300		Name: Allen Leung
301		Date: 2002/03/10 23:55:00 EST
302		Tag: leunga-20020310-x86-call
303		Description:
304
305		Added machine generation for CALL instruction (relative displacement mode)
306
307		----------------------------------------------------------------------
308		Name: Matthias Blume
309		Date: 2002/03/08 16:05:00
310		Tag: blume-20020308-entrypoints
311		Description:
312
313		Version number bumped to 110.39.1.  NEW BOOTFILES!
314
315		Entrypoints: non-zero offset into a code object where execution should begin.
316
317		- Added the notion of an entrypoint to CodeObj.
318		- Added reading/writing of entrypoint info to Binfile.
319		- Made runtime system bootloader aware of entrypoints.
320		- Use the address of the label of the first function given to mlriscGen
321		as the entrypoint.  This address is currently always 0, but it will
322		not be 0 once we turn on block placement.
323		- Removed the linkage cluster code (which was The Other Way(tm) of dealing
324		with entry points) from mlriscGen.
325
326		----------------------------------------------------------------------
327		Name: Allen Leung
328		Date: 2002/03/07 20:45:00 EST
329		Tag: leunga-20020307-x86-cmov
330		Description:
331
332		Bug fixes for CMOVcc on x86.
333
334		1. Added machine code generation for CMOVcc
335		2. CMOVcc is now generated in preference over SETcc on PentiumPro or above.
336		3. CMOVcc cannot have an immediate operand as argument.
337
338		----------------------------------------------------------------------
339		Name: Matthias Blume
340		Date: 2002/03/07 16:15:00 EST
341		Tag: blume-20020307-controls
342		Description:
343
344		This is a very large but mostly boring patch which makes (almost)
345		every tuneable compiler knob (i.e., pretty much everything under
346		Control.* plus a few other things) configurable via both the command
347		line and environment variables in the style CM did its configuration
348		until now.
349
350		Try starting sml with '-h' (or, if you are brave, '-H')
351
352		To this end, I added a structure Controls : CONTROLS to smlnj-lib.cm which
353		implements the underlying generic mechanism.
354
355		The interface to some of the existing such facilities has changed somewhat.
356		For example, the MLRiscControl module now provides mkFoo instead of getFoo.
357		(The getFoo interface is still there for backward-compatibility, but its
358		use is deprecated.)
359
360		The ml-build script passes -Cxxx=yyy command-line arguments through so
361		that one can now twiddle the compiler settings when using this "batch"
362		compiler.
363
364		TODO items:
365
366		We should go through and throw out all controls that are no longer
367		connected to anything.  Moreover, we should go through and provide
368		meaningful (and correct!) documentation strings for those controls
369		that still are connected.
370
371		Currently, multiple calls to Controls.new are accepted (only the first
372		has any effect).  Eventually we should make sure that every control
373		is being made (via Controls.new) exactly once.  Future access can then
374		be done using Controls.acc.
375
376		Finally, it would probably be a good idea to use the getter-setter
377		interface to controls rather than ref cells.  For the time being, both
378		styles are provided by the Controls module, but getter-setter pairs are
379		better if thread-safety is of any concern because they can be wrapped.
380
381		*****************************************
382
383		One bug fix: The function blockPlacement in three of the MLRISC
384		backpatch files used to be hard-wired to one of two possibilities at
385		link time (according to the value of the placementFlag).  But (I
386		think) it should rather sense the flag every time.
387
388		*****************************************
389
390		Other assorted changes (by other people who did not supply a HISTORY entry):
391
392		1. the cross-module inliner now works much better (Monnier)
393		2. representation of weights, frequencies, and probabilities in MLRISC
394		changed in preparation of using those for weighted block placement
395		(Reppy, George)
396
397		----------------------------------------------------------------------
398		Name: Lal George
399		Date: 2002/03/07 14:44:24 EST 2002
400		Tag: george-20020307-weighted-block-placement
401
402		Tested the weighted block placement optimization on all architectures
403		(except the hppa) using AMPL to generate the block and edge frequencies.
404		Changes were required in the machine properties to correctly
405		categorize trap instructions. There is an MLRISC flag
406		"weighted-block-placement" that can be used to enable weighted block
407		placement, but this will be ineffective without block/edge
408		frequencies (coming soon).
409
410
411		----------------------------------------------------------------------
412		Name: Lal George
413		Date: 2002/03/05 17:24:48 EST
414		Tag: george-20020305-linkage-cluster
415
416		In order to support the block placement optimization, a new cluster
417		is generated as the very first cluster (called the linkage cluster).
418		It contains a single jump to the 'real' entry point for the compilation
419		unit. Block placement has no effect on the linkage cluster itself, but
420		all the other clusters  have full freedom in the manner in which they
421		reorder blocks or functions.
422
423		On the x86 the typical linkage code that is generated is:
424		----------------------
425		.align 2
426		L0:
427		addl    $L1-L0, 72(%esp)
428		jmp     L1
429
430
431		.align  2
432		L1:
433		----------------------
434
435		72(%esp) is the memory location for the stdlink register. This
436		must contain the address of the CPS function being called. In the
437		above example, it contains the address of  L0; before
438		calling L1 (the real entry point for the compilation unit), it
439		must contain the address for L1, and hence
440
441		addl $L1-L0, 72(%esp)
442
443		I have tested this on all architectures except the hppa.The increase
444		in code size is of course negligible
445
446		----------------------------------------------------------------------
447		Name: Allen Leung
448		Date: 2002/03/03 13:20:00 EST
449		Tag: leunga-20020303-mlrisc-tools
450
451		Added #[ ... ] expressions to mlrisc tools
452
453		----------------------------------------------------------------------
454		Name: Matthias Blume
455		Date: 2002/02/27 12:29:00 EST
456		Tag: blume-20020227-cdebug
457		Description:
458
459		- made types in structure C and C_Debug to be equal
460		- got rid of code duplication (c-int.sml vs. c-int-debug.sml)
461		- there no longer is a C_Int_Debug (C_Debug is directly derived from C)
462
463		----------------------------------------------------------------------
464		Name: Matthias Blume
465		Date: 2002/02/26 12:00:00 EST
466		Tag: blume-20020226-ffi
467		Description:
468
469		1. Fixed a minor bug in CM's "noweb" tool:
470		If numbering is turned off, then truly don't number (i.e., do not
471		supply the -L option to noweb).  The previous behavior was to supply
472		-L'' -- which caused noweb to use the "default" line numbering scheme.
473		Thanks to Chris Richards for pointing this out (and supplying the fix).
474
475		2. Once again, I reworked some aspects of the FFI:
476
477		A. The incomplete/complete type business:
478
479		- Signatures POINTER_TO_INCOMPLETE_TYPE and accompanying functors are
480		gone!
481		- ML types representing an incomplete type are now *equal* to
482		ML types representing their corresponding complete types (just like
483		in C).  This is still safe because ml-nlffigen will not generate
484		RTTI for incomplete types, nor will it generate functions that
485		require access to such RTTI.   But when ML code generated from both
486		incomplete and complete versions of the C type meet, the ML types
487		are trivially interoperable.
488
489		NOTE:  These changes restore the full generality of the translation
490		(which was previously lost when I eliminated functorization)!
491
492		B. Enum types:
493
494		- Structure C now has a type constructor "enum" that is similar to
495		how the "su" constructor works.  However, "enum" is not a phantom
496		type because each "T enum" has values (and is isomorphic to
497		MLRep.Signed.int).
498		- There are generic access operations for enum objects (using
499		MLRep.Signed.int).
500		- ml-nlffigen will generate a structure E_foo for each "enum foo".
501		* The structure contains the definition of type "mlrep" (the ML-side
502		representation type of the enum).  Normally, mlrep is the same
503		as "MLRep.Signed.int", but if ml-nlffigen was invoked with "-ec",
504		then mlrep will be defined as a datatype -- thus facilitating
505		pattern matching on mlrep values.
506		("-ec" will be suppressed if there are duplicate values in an
507		enumeration.)
508		* Constructors ("-ec") or values (no "-ec") e_xxx of type mlrep
509		will be generated for each C enum constant xxx.
510		* Conversion functions m2i and i2m convert between mlrep and
511		MLRep.Signed.int.  (Without "-ec", these functions are identities.)
512		* Coversion functions c and ml convert between mlrep and "tag enum".
513		* Access functions (get/set) fetch and store mlrep values.
514		- By default (unless ml-nlffigen was invoked with "-nocollect"), unnamed
515		enumerations are merged into one single enumeration represented by
516		structure E_'.
517
518		----------------------------------------------------------------------
519		Name: Allen Leung
520		Date: 2002/02/25 04:45:00 EST
521		Tag: leunga-20020225-cps-spill
522
523		This is a new implementation of the CPS spill phase.
524		The new phase is in the new file compiler/CodeGen/cpscompile/spill-new.sml
525		In case of problems, replace it with the old file spill.sml
526
527		The current compiler runs into some serious performance problems when
528		constructing a large record.  This can happen when we try to compile a
529		structure with many items.  Even a very simple structure like the following
530		makes the compiler slow down.
531
532		structure Foo = struct
533		val x_1 = 0w1 : Word32.int
534		val x_2 = 0w2 : Word32.int
535		val x_3 = 0w3 : Word32.int
536		...
537		val x_N = 0wN : Word32.int
538		end
539
540		The following table shows the compile time, from N=1000 to N=4000,
541		with the old compiler:
542
543		N
544		1000   CPS 100 spill                           0.04u  0.00s  0.00g
545		MLRISC ra                               0.06u  0.00s  0.05g
546		(spills = 0 reloads = 0)
547		TOTAL                                   0.63u  0.07s  0.21g
548
549		1100   CPS 100 spill                           8.25u  0.32s  0.64g
550		MLRISC ra                               5.68u  0.59s  3.93g
551		(spills = 0 reloads = 0)
552		TOTAL                                   14.71u  0.99s  4.81g
553
554		1500   CPS 100 spill                           58.55u  2.34s  1.74g
555		MLRISC ra                               5.54u  0.65s  3.91g
556		(spills = 543 reloads = 1082)
557		TOTAL                                   65.40u  3.13s  6.00g
558
559		2000   CPS 100 spill                           126.69u  4.84s  3.08g
560		MLRISC ra                               0.80u  0.10s  0.55g
561		(spills = 42 reloads = 84)
562		TOTAL                                   129.42u  5.10s  4.13g
563
564		3000   CPS 100 spill                           675.59u  19.03s  11.64g
565		MLRISC ra                               2.69u  0.27s  1.38g
566		(spills = 62 reloads = 124)
567		TOTAL                                   682.48u  19.61s  13.99g
568
569		4000   CPS 100 spill                           2362.82u  56.28s  43.60g
570		MLRISC ra                               4.96u  0.27s  2.72g
571		(spills = 85 reloads = 170)
572		TOTAL                                   2375.26u  57.21s  48.00g
573
574		As you can see the old cps spill module suffers from some serious
575		performance problem.  But since I cannot decipher the old code fully,
576		instead of patching the problems up, I'm reimplementing it
577		with a different algorithm.  The new code is more modular,
578		smaller when compiled, and substantially faster
579		(O(n log n) time and O(n) space).  Timing of the new spill module:
580
581		4000  CPS 100 spill                           0.02u  0.00s  0.00g
582		MLRISC ra                               0.25u  0.02s  0.15g
583		(spills=1 reloads=3)
584		TOTAL                                   7.74u  0.34s  1.62g
585
586		Implementation details:
587
588		As far as I can tell, the purpose of the CPS spill module is to make sure the
589		number of live variables at any program point (the bandwidth)
590		does not exceed a certain limit, which is determined by the
591		size of the spill area.
592
593		When the bandwidth is too large, we decrease the register pressure by
594		packing live variables into spill records.  How we achieve this is
595		completely different than what we did in the old code.
596
597		First, there is something about the MLRiscGen code generator
598		that we should be aware of:
599
600		o MLRiscGen performs code motion!
601
602		In particular, it will move floating point computations and
603		address computations involving only the heap pointer to
604		their use sites (if there is only a single use).
605		What this means is that if we have a CPS record construction
606		statement
607
608		RECORD(k,vl,w,e)
609
610		we should never count the new record address w as live if w
611		has only one use (which is often the case).
612
613		We should do something similar to floating point, but the transformation
614		there is much more complex, so I won't deal with that.
615
616		Secondly, there are now two new cps primops at our disposal:
617
618		1. rawrecord of record_kind option
619		This pure operator allocates some uninitialized storage from the heap.
620		There are two forms:
621
622		rawrecord NONE [INT n]  allocates a tagless record of length n
623		rawrecord (SOME rk) [INT n] allocates a tagged record of length n
624		and initializes the tag.
625
626		2. rawupdate of cty
627		rawupdate cty (v,i,x)
628		Assigns to x to the ith component of record v.
629		The storelist is not updated.
630
631		We use these new primops for both spilling and increment record construction.
632
633		1. Spilling.
634
635		This is implemented with a linear scan algorithm (but generalized
636		to trees).  The algorithm will create a single spill record at the
637		beginning of the cps function and use rawupdate to spill to it,
638		and SELECT or SELp to reload from it.  So both spills and reloads
639		are fine-grain operations.  In contrast, in the old algorithm
640		"spills" have to be bundled together in records.
641
642		Ideally, we should sink the spill record construction to where
643		it is needed.  We can even split the spill record into multiple ones
644		at the places where they are needed.  But CPS is not a good
645		representation for global code motion, so I'll keep it simple and
646		am not attempting this.
647
648		2. Incremental record construction (aka record splitting).
649
650		Long records with many component values which are simulatenously live
651		(recall that single use record addresses are not considered to
652		be live) are constructed with rawrecord and rawupdate.
653		We allocate space on the heap with rawrecord first, then gradually
654		fill it in with rawupdate.  This is the technique suggested to me
655		by Matthias.
656
657		Some restrictions on when this is applicable:
658		1. It is not a VECTOR record.  The code generator currently does not handle
659		this case. VECTOR record uses double indirection like arrays.
660		2. All the record component values are defined in the same "basic block"
661		as the record constructor.  This is to prevent speculative
662		record construction.
663
664		----------------------------------------------------------------------
665		Name: Allen Leung
666		Date: 2002/02/22 01:02:00 EST
667		Tag: leunga-20020222-mlrisc-tools
668
669		Minor bug fixes in the parser and rewriter
670
671		----------------------------------------------------------------------
672		Name: Allen Leung
673		Date: 2002/02/21 20:20:00 EST
674		Tag: leunga-20020221-peephole
675
676		Regenerated the peephole files.  Some contained typos in the specification
677		and some didn't compile because of pretty printing bugs in the old version
678		of 'nowhere'.
679
680		----------------------------------------------------------------------
681		Name: Allen Leung
682		Date: 2002/02/19 20:20:00 EST
683		Tag: leunga-20020219-mlrisc-tools
684		Description:
685
686		Minor bug fixes to the mlrisc-tools library:
687
688		1.  Fixed up parsing colon suffixed keywords
689		2.  Added the ability to shut the error messages up
690		3.  Reimplemented the pretty printer and fixed up/improved
691		the pretty printing of handle and -> types.
692		4.  Fixed up generation of literal symbols in the nowhere tool.
693		5.  Added some SML keywords to to sml.sty
694
695		----------------------------------------------------------------------
696		Name: Matthias Blume
697		Date: 2002/02/19 16:20:00 EST
698		Tag: blume-20020219-cmffi
699		Description:
700
701		A wild mix of changes, some minor, some major:
702
703		* All C FFI-related libraries are now anchored under $c:
704		$/c.cm      --> $c/c.cm
705		$/c-int.cm  --> $c/internals/c-int.cm
706		$/memory.cm --> $c/memory/memory.cm
707
708		* "make" tool (in CM) now treats its argument pathname slightly
709		differently:
710		1. If the native expansion is an absolute name, then before invoking
711		the "make" command on it, CM will apply OS.Path.mkRelative
712		(with relativeTo = OS.FileSys.getDir()) to it.
713		2. The argument will be passed through to subsequent phases of CM
714		processing without "going native".  In particular, if the argument
715		was an anchored path, then "make" will not lose track of that anchor.
716
717		* Compiler backends now "know" their respective C calling conventions
718		instead of having to be told about it by ml-nlffigen.  This relieves
719		ml-nlffigen from one of its burdens.
720
721		* The X86Backend has been split into X86CCallBackend and X86StdCallBackend.
722
723		* Export C_DEBUG and C_Debug from $c/c.cm.
724
725		* C type encoding in ml-nlffi-lib has been improved to model the conceptual
726		subtyping relationship between incomplete pointers and their complete
727		counterparts.  For this, ('t, 'c) ptr has been changed to 'o ptr --
728		with the convention of instantiating 'o with ('t, 'c) obj whenever
729		the pointer target type is complete.  In the incomplete case, 'o
730		will be instantiated with some "'c iobj" -- a type obtained by
731		using one of the functors PointerToIncompleteType or PointerToCompleteType.
732
733		Operations that work on both incomplete and complete pointer types are
734		typed as taking an 'o ptr while operations that require the target to
735		be known are typed as taking some ('t, 'c) obj ptr.
736
737		voidptr is now a bit "more concrete", namely "type voidptr = void ptr'"
738		where void is an eqtype without any values.  This makes it possible
739		to work on voidptr values using functions meant to operate on light
740		incomplete pointers.
741
742		* As a result of the above, signature POINTER_TO_INCOMPLETE_TYPE has
743		been vastly simplified.
744
745		----------------------------------------------------------------------
746		Name: Matthias Blume
747		Date: 2002/02/19 10:48:00 EST
748		Tag: blume-20020219-pqfix
749		Description:
750
751		Applied Chris Okasaki's bug fix for priority queues.
752
753		----------------------------------------------------------------------
754	Name: Matthias Blume	Name: Matthias Blume
755	Date: 2002/02/15 17:05:00	Date: 2002/02/15 17:05:00
756	Tag: Release_110_39	Tag: Release_110_39

---

Colored Diff Long Colored Diff Unidiff Context Diff Side by Side

Legend: Removed from v.1073   changed lines   Added in v.1158

---

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