Diff of /sml/trunk/HISTORY

Parent Directory | Revision Log | Patch

	revision 757, Thu Dec 21 14:54:56 2000 UTC	revision 909, Fri Aug 24 17:48:53 2001 UTC
#	Line 11	Line 11
11	Date:	Date:
12	Tag: <post-commit CVS tag>	Tag: <post-commit CVS tag>
13	Description:	Description:
14
15		----------------------------------------------------------------------
16		Name: Matthias Blume
17		Date: 2001/08/24 13:47:18 EDT 2001
18		Tag: george-20010824-MLRISC-graphs
19		Description:
20
21		removed clusters from MLRISC completely and replaced with graphs.
22
23		----------------------------------------------------------------------
24		Name: Matthias Blume
25		Date: 2001/08/23 17:50:00 EDT
26		Tag: blume-20010823-toplevel
27		Description:
28
29		- some reorganization of the code that implements various kinds of
30		environments in the compiler (static, dynamic, symbolic, combined)
31		- re-implemented the EnvRef module so that evalStream works properly
32		(if the stream contains references to "use", "CM.make", etc.)
33		- cleaned up evalloop.sml and interact.sml (but they need more cleaning)
34
35		----------------------------------------------------------------------
36		Name: Matthias Blume
37		Date: 2001/08/20 15:50 EDT
38		Tag: blume20010820-slipup
39		Description:
40
41		I forgot to commit a few files.  Here they are...
42
43		----------------------------------------------------------------------
44		Name: Matthias Blume
45		Date: 2001/08/20 15:35:00 EDT
46		Tag: blume-20010820-debugprof
47		Description:
48
49		!!!! NEW BOOTFILES !!!!
50
51		This is another round of reorganizing the compiler sources.  This
52		time the main goal was to factor out all the "instrumentation"
53		passes (for profiling and backtracing) into their own library.
54		The difficulty was to do it in such a way that it does not depend
55		on elaborate.cm but only on elabdata.cm.
56
57		Therefore there have been further changes to both elaborate.cm and
58		elabdata.cm -- more "generic" things have been moved from the former
59		to the latter.  As a result, I was forced to split the assignment
60		of numbers indicating "primtyc"s into two portions: SML-generic and
61		SML/NJ-specific.  Since it would have been awkward to maintain,
62		I bit the bullet and actually _changed_ the mapping between these
63		numbers and primtycs.  The bottom line of this is that you need
64		a new set of bin- and bootfiles.
65
66		I have built new bootfiles for all architectures, so doing a fresh
67		checkout and config/install.sh should be all you need.
68
69		The newly created library's name is
70
71		$smlnj/viscomp/debugprof.cm
72
73		and its sources live under
74
75		src/compiler/DebugProf
76
77		----------------------------------------------------------------------
78		Name: Matthias Blume
79		Date: 2001/08/15 17:15:00 EDT
80		Tag: blume-20010815-compreorg
81		Description:
82
83		This is a first cut at reorganizing the CM libraries that make up the
84		core of the compiler.  The idea is to separate out pieces that could
85		be used independently by tools, e.g., the parser, the typechecker, etc.
86
87		The current status is a step in this direction, but it is not quite
88		satisfactory yet.  Expect more changes in the future.
89
90		Here is the current (new) organization...
91
92		What used to be $smlnj/viscomp/core.cm is now divided into
93		six CM libraries:
94
95		$smlnj/viscomp/basics.cm
96		/parser.cm
97		/elabdata.cm
98		/elaborate.cm
99		/execute.cm
100		/core.cm
101
102		The CM files for these libraries live under src/system/smlnj/viscomp.
103		All these libraries are proxy libraries that contain precisely
104		one CM library component.  Here are the locations of the components
105		(all within the src/compiler tree):
106
107		Basics/basics.cm
108		Parse/parser.cm
109		ElabData/elabdata.cm
110		Elaborator/elaborate.cm
111		Execution/execute.cm
112		core.cm
113
114		[This organization is the same that has been used already
115		for a while for the architecture-specific parts of the visible
116		compiler and for the old version of core.cm.]
117
118		As you will notice, many source files have been moved from their
119		respective original locations to a new home in one of the above
120		subtrees.
121
122		The division of labor between the new libraries is the following:
123
124		basics.cm:
125		- Simple, basic definitions that pertain to many (or all) of
126		the other libraries.
127		parser.cm:
128		- The SML parser, producing output of type Ast.dec.
129		- The type family for Ast is also defined and exported here.
130		elabdata.cm:
131		- The datatypes that describe input and output of the elaborator.
132		This includes types, absyn, and static environments.
133		elaborator.cm:
134		- The SML/NJ type checker and elaborator.
135		This maps an Ast.dec (with a given static environment) to
136		an Absyn.dec (with a new static environment).
137		- This libraries implements certain modules that used to be
138		structures as functors (to remove dependencies on FLINT).
139		execute.cm:
140		- Everything having to do with executing binary code objects.
141		- Dynamic environments.
142		core.cm:
143		- SML/NJ-specific instantiations of the elaborator and MLRISC.
144		- Top-level modules.
145		- FLINT (this should eventually become its own library)
146
147		Notes:
148
149		I am not 100% happy with the way I separated the elaborator (and its
150		data structures) from FLINT.  Two instances of the same problem:
151
152		1. Data structures contain certain fields that carry FLINT-specific
153		information.  I hacked around this using exn and the property list
154		module from smlnj-lib.  But the fact that there are middle-end
155		specific fields around at all is a bit annoying.
156
157		2. The elaborator calculates certain FLINT-related information.  I tried
158		to make this as abstract as I could using functorization, but, again,
159		the fact that the elaborator has to perform calculations on behalf
160		of the middle-end at all is not nice.
161
162		3. Having to used exn and property lists is unfortunate because it
163		weakens type checking.  The other alternative (parameterizing
164		nearly *everything*) is not appealing, though.
165
166		I removed the "rebinding =" warning hack because due to the new organization
167		it was awkward to maintain it.  As a result, the compiler now issues some of
168		these warnings when compiling init.cmi during bootstrap compilation. On
169		the plus side, you also get a warning when you do, for example:
170		val op = = Int32.+
171		which was not the case up to now.
172
173		I placed "assign" and "deref" into the _Core structure so that the
174		code that deals with the "lazy" keyword can find them there.  This
175		removes the need for having access to the primitive environment
176		during elaboration.
177
178		----------------------------------------------------------------------
179		Name: Matthias Blume
180		Date: 2001/08/13
181		Tag: blume-20010813-closures
182		Description:
183
184		This fix was sent to us by Zhong Shao.  It is supposed to improve the
185		performance of certain loops by avoiding needless closure allocation.
186
187		----------------------------------------------------------------------
188		Name: Lal George
189		Date: 2001/07/31 10:03:23 EDT 2001
190		Tag: george-20010731-x86-fmalloc
191		Description: Fixed bug in x86 calls
192
193		There was a bug where call instructions would mysteriously
194		vanish. The call instruction had to be one that returned
195		a floating point value.
196
197		----------------------------------------------------------------------
198		Name: Lal George
199		Date: 2001/07/19 16:36:29 EDT 2001
200		Tag: george-20010719-simple-cells
201		Description:
202
203		I have dramatically simplified the interface for CELLS in MLRISC.
204
205		In summary, the cells interface is broken up into three parts:
206
207		1. CellsBasis : CELLS_BASIS
208
209		CellsBasis is a top level structure and common for all
210		architectures.  it contains the definitions of basic datatypes
211		and utility  functions over these types.
212
213		2. functor Cells() : CELLS
214
215		Cells generates an interface for CELLS that incorporates the
216		specific resources on the target architecture, such as the
217		presence of special register classes, their number and size,
218		and various useful substructures.
219
220		3. <ARCH>CELLS
221
222		e.g. SparcCells: SPARCCELLS
223
224		<ARCH>CELLS usually contains additional bindings for special
225		registers  on the architecture, such as:
226
227		val r0 : cell           (* register zero *)
228		val y : cell            (* Y register *)
229		val psr : cell          (* processor status register *)
230		...
231
232		The structure returned by applying the Cells functor is opened
233		in this interface.
234
235		The main implication of all this is that the datatypes for cells is
236		split between CellsBasis and CELLS -- a fairly simple change for user
237		code.
238
239		In the old scheme the CELLS interface had a definitional binding of
240		the form:
241
242		signature CELLS = sig
243
244		structure CellsBasis = CellsBasis
245
246		...
247
248		end
249
250		With all the sharing constraints that goes on in MLRISC, this old
251		design  quickly leads to errors such as:
252
253		"structure definition spec inside of sharing ... "
254
255
256		and appears to require an unacceptable amount of sharing and where
257		constraint hackery.
258
259		I think this error message (the interaction of definitional specs and
260		sharing) requires more explanation on our web page.
261
262		----------------------------------------------------------------------
263		Name: Matthias Blume
264		Date: 2001/07/19 15:00:00 EDT
265		Tag: blume-20010719-libreorg
266		Description:
267
268		This update puts together a fairly extensive but straightforward change
269		to the way the libraries that implement the interactive system are
270		organized:
271
272		The biggest change is the elimination of structure Compiler.  As a
273		replacement for this structure, there is now a CM library
274		(known as $smlnj/compiler.cm or $smlnj/compiler/current.cm)
275		that exports all the substructures of the original structure Compiler
276		directly.  So instead of saying Compiler.Foo.bar one now simply
277		says Foo.bar.  (The CM libraries actually export a collection of
278		structures that is richer than the collection of substructures of
279		structure Compiler.)
280
281		To make the transition smooth, there is a separate library called
282		$smlnj/compiler/compiler.cm which puts together and exports the
283		original structure Compiler (or at least something very close to it).
284
285		There are five members of the original structure Compiler
286		that are not exported directly but which instead became members
287		of a new structure Backend (described by signature BACKEND).  These are:
288		structure Profile (: PROFILE), structure Compile (: COMPILE), structure
289		Interact (: INTERACT), structure Machine (: MACHINE), and val
290		architecture (: string).
291
292		Structure Compiler.Version has become structure CompilerVersion.
293
294		Cross-compilers for alpha32, hppa, ppc, sparc, and x86 are provided
295		by $smlnj/compiler/<arch>.cm where <arch> is alpha32, hppa, ppc, sparc,
296		or x86, respectively.
297		Each of these exports the same frontend structures that
298		$smlnj/compiler.cm exports.  But they do not have a structure Backend
299		and instead export some structure <Arch>Backend where <Arch> is Alpha32,
300		Hppa, PPC, Sparc, or X86, respectively.
301
302		Library $smlnj/compiler/all.cm exports the union of the exports of
303		$smlnj/compiler/<arch>.cm
304
305		There are no structures <Arch>Compiler anymore, use
306		$smlnj/compiler/<arch>.cm instead.
307
308		Library host-compiler-0.cm is gone.  Instead, the internal library
309		that instantiates CM is now called cm0.cm.  Selection of the host
310		compiler (backend) is no longer done here but. (Responsibility for it
311		now lies with $smlnj/compiler/current.cm.  This seems to be more
312		logical.)
313
314		Many individual files have been moved or renamed.  Some files have
315		been split into multiple files, and some "dead" files have been deleted.
316
317		Aside from these changes to library organization, there are also changes
318		to the way the code itself is organized:
319
320		Structure Binfile has been re-implemented in such a way that it no
321		longer needs any knowledge of the compiler.  It exclusively deals
322		with the details of binfile layout.  It no longer invokes the
323		compiler (for the purpose of creating new prospective binfile
324		content), and it no longer has any knowledge of how to interpret
325		pickles.
326
327		Structure Compile (: COMPILE) has been stripped down to the bare
328		essentials of compilation.  It no longer deals with linking/execution.
329		The interface has been cleaned up considerably.
330
331		Utility routines for dealing with linking and execution have been
332		moved into their own substructures.
333
334		(The ultimate goal of these changes is to provide a light-weight
335		binfile loader/linker (at least for, e.g., stable libraries) that
336		does not require CM or the compiler to be present.)
337
338		CM documentation has been updated to reflect the changes to library
339		organization.
340
341		----------------------------------------------------------------------
342		Name: Matthias Blume
343		Date: 2001/07/10 17:30:00 EDT
344		Tag: Release_110_34
345		Description:
346
347		Minor tweak to 110.34 (re-tagged):
348
349		- README.html file added to CVS repository
350		- runtime compiles properly under FreeBSD 3.X and 4.X
351
352		----------------------------------------------------------------------
353		Name: Matthias Blume
354		Date: 2001/07/10 17:30:00 EDT
355		Tag: Release_110_34
356		Description:
357
358		New version number (110.34). New bootfiles.
359
360		----------------------------------------------------------------------
361		Name: Matthias Blume
362		Date: 2001/07/09 16:00:00 EDT
363		Tag: blume-20010709-more-varargs
364		Description:
365
366		I changed the handling of varargs in ml-nlffigen again:
367		The ellipsis ... will now simply be ignored (with an accompanying warning).
368
369		The immediate effect is that you can actually call a varargs function
370		from ML -- but you can't actually supply any arguments beyond the ones
371		specified explicitly.  (For example, you can call printf with its format
372		string, but you cannot pass additional arguments.)
373
374		This behavior is only marginally more useful than the one before, but
375		it has the advantage that a function or, more importantly, a function
376		type never gets dropped on the floor, thus avoiding follow-up problems with
377		other types that refer to the offending one.
378
379		----------------------------------------------------------------------
380		Name: Matthias Blume
381		Date: 2001/07/09 11:25:00 EDT
382		Tag: blume-20010709-varargs
383		Description:
384
385		1. ckit-lib.cm now exports structure Error
386		2. ml-nlffigen reports occurences of "..." (i.e., varargs function types)
387		with a warning accompanied by a source location.  Moreover, it
388		merely skips the offending function or type and proceeds with the
389		rest of its work.u  As a result, one can safely feed C code containing
390		"..." to ml-nlffigen.
391		3. There are some internal improvements to CM, providing slightly
392		more general string substitutions in the tools subsystem.
393
394		----------------------------------------------------------------------
395		Name: Matthias Blume
396		Date: 2001/06/27 15:10:00 EDT
397		Tag: blume-20010627-concur
398		Description:
399
400		Fixed a small bug in CM's handling of parallel compilation.
401		(You could observe the bug by Control-C-interrupting an ordinary
402		CMB.make or CM.stabilize and then attaching some compile servers.
403		The result was that all of a sudden the previously interrupted
404		compilation would continue on its own.  This was because of
405		an over-optimization: CM did not bother to clean out certain queues
406		when no servers were attached "anyway", resulting in the contents
407		of these queues to grab control when new servers did get attached.)
408
409		There is also another minor update to the CM manual.
410
411		----------------------------------------------------------------------
412		Name: Matthias Blume
413		Date: 2001/06/26 16:15:00 EDT
414		Tag: blume-20010626-cmdoc
415		Description:
416
417		Minor typo fixed in CM manual (syntax diagram for libraries).
418
419		----------------------------------------------------------------------
420		Name: Matthias Blume
421		Date: 2001/06/25 22:55:00 EDT
422		Tag: blume-20010625-x86pc
423		Description:
424
425		Fixed a nasty bug in the X86 assembly code that caused signal
426		handlers to fail (crash) randomly.
427
428		----------------------------------------------------------------------
429		Name: Matthias Blume
430		Date: 2001/06/25 12:05:00 EDT
431		Tag: blume-20010625-nlffigen
432		Description:
433
434		This update fixes a number of minor bugs in ml-nlffigen as reported by
435		Nick Carter <nbc@andrew.cmu.edu>.
436
437		1. Silly but ok typedefs of the form "typedef void myvoid;" are now accepted.
438		2. Default names for generated files are now derived from the name of
439		the C file *without its directory*.  In particular, this causes generated
440		files to be placed locally even if the C file is in some system directory.
441		3. Default names for generated signatures and structures are also derived
442		from the C file name without its directory.  This avoids silly things
443		like "structure GL/GL".
444		(Other silly names are still possible because ml-nlffigen does not do
445		a thorough check of whether generated names are legal ML identifiers.
446		When in doubt, use command line arguments to force particular names.)
447
448		----------------------------------------------------------------------
449		Name: Matthias Blume
450		Date: 2001/06/21 12:25:00 EDT
451		Tag: blume-20010621-eXene
452		Description:
453
454		eXene now compiles and (sort of) works again.
455
456		The library name (for version > 110.33) is $/eXene.cm.
457
458		I also added an new example in src/eXene/examples/nbody.  See the
459		README file there for details.
460
461		----------------------------------------------------------------------
462		Name: Matthias Blume
463		Date: 2001/06/20 16:40:00 EDT
464		Tag: blume-20010620-cml
465		Description:
466
467		CML now compiles and works again.
468
469		Libraries (for version > 110.33):
470
471		$cml/cml.cm            Main CML library.
472		$cml/basis.cm          CML's version of $/basis.cm.
473		$cml/cml-internal.cm   Internal helper library.
474		$cml/core-cml.cm       Internal helper library.
475		$cml-lib/trace-cml.cm  Tracing facility.
476		$cml-lib/smlnj-lib.cm  CML's version of $/smlnj-lib.cm
477
478		The installer (config/install.sh) has been taught how to properly
479		install this stuff.
480
481		----------------------------------------------------------------------
482		Name: Matthias Blume
483		Date: 2001/06/19 17:55:00 EDT
484		Tag: blume-20010619-instantiate
485		Description:
486
487		This un-breaks the fix for bug 1432.
488		(The bug was originally fixed in 110.9 but I broke it again some
489		time after that.)
490
491		----------------------------------------------------------------------
492		Name: Matthias Blume
493		Date: 2001/06/19 17:25:00 EDT
494		Tag: blume-20010619-signals
495		Description:
496
497		This should (hopefully) fix the long-standing signal handling bug.
498		(The runtime system was constructing a continuation record with an
499		incorrect descriptor which would cause the GC to drop data on the floor...)
500
501		----------------------------------------------------------------------
502		Name: Matthias Blume
503		Date: 2001/06/15 15:05:00 EDT
504		Tag: blume-20010615-moresparc
505		Description:
506
507		Here is a short late-hour update related to Sparc c-calls:
508
509		-- made handling of double-word arguments a bit smarter
510
511		-- instruction selection phase tries to collapse certain clumsily
512		constructed ML-Trees; typical example:
513
514		ADD(ty,ADD(_,e,LI d1),LI d2)  ->  ADD(ty,e,LI(d1+d2))
515
516		This currently has no further impact on SML/NJ since mlriscGen does
517		not seem to generate such patterns in the first place, and c-calls
518		(which did generate them in the beginning) has meanwhile been fixed
519		so as to avoid them as well.
520
521		----------------------------------------------------------------------
522		Name: Matthias Blume
523		Date: 2001/06/15 15:05:00 EDT
524		Tag: blume-20010615-sparc
525		Description:
526
527		The purpose of this update is to provide an implementation of NLFFI
528		on Sparc machines.
529
530		Here are the changes in detail:
531
532		* src/MLRISC/sparc/c-calls/sparc-c-calls.sml is a new file containing
533		the Sparc implementation of the c-calls API.
534		* The Sparc backend of SML/NJ has been modified to uniformely use %fp
535		for accessing the ML frame.  Thus, we have a real frame pointer and
536		can freely modify %sp without need for an omit-frame-ptr phase.
537		The vfp logic in src/compiler/CodeGen/* has been changed to accomodate
538		this case.
539		* ml-nlffigen has been taught to produce code for different architectures
540		and calling conventions.
541		* In a way similar to what was done in the x86 case, the Sparc
542		backend uses its own specific extension to mltree.  (For example,
543		it needs to be able to generate UNIMP instructions which are part
544		of the calling convention.)
545		* ml-nlffi-lib was reorganized to make it more modular (in particular,
546		to make it easier to plug in new machine- and os-dependent parts).
547
548		There are some other fairly unrelated bug fixes and cleanups as well:
549
550		* I further hacked the .cm files for MLRISC tools (like MDLGen) so
551		that they properly share their libraries with existing SML/NJ libraries.
552		* I fixed a minor cosmetic bug in CM, supressing certain spurious
553		follow-up error messages.
554		* Updates to CM/CMB documentation.
555
556		TODO items:
557
558		* MLRISC should use a different register as its asmTemp on the Sparc.
559		(The current %o2 is a really bad choice because it is part of the
560		calling conventions, so things might interfere in unexpected ways.)
561
562		----------------------------------------------------------------------
563		Name: Matthias Blume
564		Date: 2001/06/07
565		Tag: blume-20010607-calls
566		Description:
567
568		A number of internal changes related to C calls and calling conventions:
569
570		1. ML-Tree CALL statements now carry a "pops" field.  It indicates the
571		number of bytes popped implicitly (by the callee).  In most cases
572		this field is 0 but on x86/win32 it is some non-zero value.  This
573		is information provided for the benefit of the "omit-frameptr" pass.
574		2. The CALL instruction on the x86 carries a similar "pops" field.
575		The instruction selection phase copies its value from the ML-Tree
576		CALL statement.
577		3. On all other architectures, the instruction selection phase checks
578		whether "pops=0" and complains if not.
579		4. The c-calls implementation for x86 now accepts two calling conventions:
580		"ccall" and "stdcall".  When "ccall" is selected, the caller cleans
581		up after the call and pops is set to 0.  For "stdcall", the caller
582		does nothing, leaving the cleanup to the callee; pops is set to
583		the number of bytes that were pushed onto the stack.
584		5. The cproto decoder (compiler/Semant/types/cproto.sml) now can
585		distinguish between "ccall" and "stdcall".
586		6. The UNIMP instruction has been added to the supported Sparc instruction
587		set. (This is needed for implementing the official C calling convention
588		on this architecture.)
589		7. I fixed some of the .cm files under src/MLRISC/Tools to make them
590		work with the latest CM.
591
592		----------------------------------------------------------------------
593		Name: Matthias Blume
594		Date: 2001/06/05 15:10:00 EDT
595		Tag: blume-20010605-cm-index
596		Description:
597
598		0. The "lambdasplit" parameter for class "sml" in CM has been documented.
599
600		1. CM can now generate "index files".  These are human-readable files
601		that list on a per-.cm-file basis each toplevel symbol defined or
602		imported.  The location of the index file for
603		<p>/<d>.cm is <p>/CM/INDEX/<d>.cm.
604		To enable index-file generation, set CM.Control.generate_index to true
605		or export an environment-symbol: export CM_GENERATE_INDEX=true.
606
607		The CM manual has been updated accordingly.
608
609		2. I made some slight modifications to the c-calls API in MLRISC.
610
611		a) There is now a callback to support saving/restoring of
612		dedicated but caller-save registers around the actual call
613		instruction.
614		b) One can optionally specify a comment-annotation for the
615		call instruction.
616
617		3. SML/NJ (mlriscGen.sml) uses this new API for the rawccall primop.
618		(For example, the comment annotation shows the C prototype of
619		the function being called.)
620
621		----------------------------------------------------------------------
622		Name: Matthias Blume
623		Date: 2001/06/01 13:30:00 EDT
624		Tag: blume-20010601-nlffi-cleanup
625		Description:
626
627		This is mostly a cleanup of MLFFI stuff:
628
629		- some signature files have been put into a more exposed place
630		- the ugly 'f type parameter is gone (simplifies types tremendously!)
631		- ml-nlffigen changed accordingly
632		- tutorial updated
633
634		Other changes:
635
636		- author's affiliation in CM manual(s) updated
637		- some more recognized keywords added to Allen's sml.sty
638
639		----------------------------------------------------------------------
640		Name: Matthias Blume
641		Date: 2001/05/25 15:30:00 EDT
642		Tag: blume-20010525-iptr
643		Description:
644
645		- put the official 110.33-README (as it appears on the ftp server) under
646		CVS
647		- fixed a small bug related to incomplete pointer types in
648		ml-nlffigen
649		- small cosmetic change to the ml-nlffi-lib's "arr" type constructor
650		(it does not need the 'f type parameter)
651
652		----------------------------------------------------------------------
653		Name: Matthias Blume
654		Date: 2001/05/23 14:30:00 EDT
655		Tag: Release_110_33
656		Description:
657
658		New version number (110.33).  New bootfiles.
659
660		----------------------------------------------------------------------
661		Name: Matthias Blume
662		Date: 2001/05/22 18:06:00 EDT
663		Tag: blume-20010522-targets
664		Description:
665
666		Made install.sh use file config/targets.customized if it exists, falling
667		back to config/targets if it doesn't.  This way one can have a customized
668		version of the targets file without touching the "real thing", thus
669		eliminating the constant fear of accidentally checking something bogus
670		back into the CVS repository...  (File config/targets.customized must
671		not be added to the repository!)
672
673		----------------------------------------------------------------------
674		Name: Matthias Blume
675		Date: 2001/05/22 16:30:00 EDT
676		Tag: blume-20010522-minitut
677		Description:
678
679		1. Bug fix in ml-nlffigen; now (hopefully) correctly handling
680		struct returns.
681		2. Added src/ml-nlffi-lib/Doc/mini-tutorial.txt.  This is some very
682		incomplete, preliminary documentation for NLFFI.
683
684		----------------------------------------------------------------------
685		Name: Matthias Blume
686		Date: 2001/05/14 11:30:00 EDT
687		Tag: blume-20010514-script
688		Description:
689
690		Some bugs in install script fixed.
691
692		In addition to that I also made a slight change to the NLFFI API:
693		Functors generated by ml-nlffigen now take the dynamic library as a
694		straight functor argument, not as a suspended one.  (The original
695		functor code used to force the suspension right away anyway, so there
696		was nothing gained by this complication of the interface.)
697
698		----------------------------------------------------------------------
699		Name: Matthias Blume
700		Date: 2001/05/11 14:35:00 EDT
701		Tag: blume-20010511-ml-nlffi
702		Description:
703
704		I finally took the plunge and added my new FFI code to the main
705		repository.  For x86-linux it is now ready for prime-time.
706
707		There are two new subdirectories of "src":
708
709		- ml-nlffi-lib:
710		The utility library for programs using the FFI interface.
711		Here is the implementation of $/c.cm and its associated low-level
712		partners $/c-int.cm and $/memory.cm.
713		- ml-nlffigen:
714		A stand-alone program for generating ML glue code from C source
715		code.
716
717		Building ml-nlffigen requires $/ckit-lib.cm.
718
719		The config/install.sh script has been updates to do the Right Thing
720		(hopefully).
721
722		Notice that the source tree for the C-Kit will not be put under "src"
723		but directly under the installation root directory.  (This is the
724		structure that currently exists on the CVS server when you check out
725		module "sml".)  Fortunately, config/install.sh knows about this oddity.
726
727		Bugs: No documentation yet.
728
729		----------------------------------------------------------------------
730		Name: Matthias Blume
731		Date: 2001/05/09 16:35:00 EDT
732		Tag: blume-20010509-cpscontract
733		Description:
734
735		Fixed a bug in the accounting code in cpsopt/contract.sml.  (The
736		wrapper/unwrapper elimination did not decrement usage counts and some
737		dead variables got overlooked by the dead-up logic.)
738
739		----------------------------------------------------------------------
740		Name: Lal George
741		Date: 2001/05/08  17:26:09 EDT
742		Tag: george-20010508-omit-frameptr
743		Description:
744
745		Changes to implement the omit-frame-pointer optimization to support
746		raw C calls. For now, there is only support on the Intel x86, but
747		other architectures will follow as more experience is gained with this.
748
749
750		----------------------------------------------------------------------
751		Name: Matthias Blume
752		Date: 2001/05/07 14:40:00 EDT
753		Tag: blume-20010507-proxies
754		Description:
755
756		I made into "proxy libraries" all libraries that qualify for such a
757		change.  (A qualifying library is a library that has another library or
758		groups as its sole member and repeats that member's export list
759		verbatim.  A proxy library avoids this repetition by omitting its export
760		list, effectively inheriting the list that its (only) member exports.
761		See the CM manual for more explanation.)
762		The main effect is that explicit export lists for these libraries
763		do not have to be kepts in sync, making maintenance a bit easier.
764
765		I also added copyright notices to many .cm-files.
766
767		Last but not least, I made a new set of bootfiles.
768
769		----------------------------------------------------------------------
770		Name: Matthias Blume
771		Date: 2001/05/04 17:00:00 EDT
772		Tag: blume-20010504-cm-lsplit
773		Description:
774
775		0. John merged pending changes to $/smlnj-lib.cm
776
777		1. Allen's previous change accidentally backed out of one of Lal's
778		earlier changes.  I undid this mistake (re-introducing Lal's change).
779
780		2. I used the new topOrder' function from graph-scc.sml (from $/smlnj-lib.cm)
781		within the compiler where applicable.  There is some code simplification
782		because of that.
783
784		3. The "split" phase (in FLINT) is now part of the default list of phases.
785		Compiler.Control.LambdaSplitting.* can be used to globally control the
786		lambda-splitting (cross-module-inlining) engine.  In addition to that,
787		it can now also be controlled on a per-source basis: CM has been taught
788		a new tool parameter applicable to ML source files.
789
790		- To turn lambda-splitting off completely:
791		local open Compiler.Control.LambdaSplitting in
792		val _ = set Off
793		end
794		- To make "no lambda-splitting" the global default (but allow per-source
795		overriding); this is the initial setting:
796		local open Compiler.Control.LambdaSplitting in
797		val _ = set (Default NONE)
798		end
799		- To make "lambda-splitting with aggressiveness a" the global default
800		(and allow per-source overriding):
801		local open Compiler.Control.LambdaSplitting in
802		val _ = set (Default (SOME a))
803		end
804
805		- To turn lambda-splitting off for a given ML souce file (say: a.sml)
806		write (in the respective .cm-file):
807		a.sml (lambdasplitting:off)
808		- To turn lambda-splitting for a.sml on with minimal aggressiveness:
809		a.sml (lambdasplitting:on)
810		- To turn lambda-splitting for a.sml on with aggressiveness <a> (where
811		<a> is a decimal non-negative integer):
812		a.sml (lambdasplitting:<a>)
813		- To turn lambda-splitting for a.sml on with maximal aggressiveness:
814		a.sml (lambdasplitting:infinity)
815		- To use the global default for a.sml:
816		a.sml (lambdasplitting:default)
817		or simply
818		a.sml
819
820		----------------------------------------------------------------------
821		Name: Allen Leung
822		Date: 2001/05/04 01:57:00 EDT
823		Tag: leunga-20010504-sync
824		Description:
825
826		MLRISC features.
827
828		1. Fix to CMPXCHG instructions.
829		2. Changed RA interface to allow annotations in callbacks.
830		3. Added a new method to the stream interface to allow annotations updates.
831
832		----------------------------------------------------------------------
833		Name: Matthias Blume
834		Date: 2001/05/01 11:45:00 EDT
835		Tag: blume-20010501-pcedittmp
836		Description:
837
838		Changed install.sh to use the current working directory instead of
839		/usr/tmp for a temporary file (pcedittmp).  The previous choice
840		of /usr/tmp caused trouble with MacOS X because of file premission
841		problems.
842
843		----------------------------------------------------------------------
844		Name: Matthias Blume
845		Date: 2001/04/20 11:10:00 EDT
846		Tag: blume-20010420-inMLflag
847		Description:
848
849		- added vp_limitPtrMask to vproc-state.h
850		(for use by the raw-C-calls mechanism to implement proper interrupt
851		handling)
852		- made the ML compiler aware of various data-structure offsets so it
853		can generate code for accessing the vp_inML flag and vp_limitPtrMask
854		- tweaked mlriscGen.sml to have it emit interrupt-handling code for
855		raw C-calls
856
857		----------------------------------------------------------------------
858		Name: Lal George
859		Date: 2001/04/20 09:15:28 EDT
860		Tag: george-20010420-macosX
861		Description:
862
863		- Changes to port to Mac OS X; Darwin.
864
865		- In the process I found that sqrt was broken on the PPC, because the
866		fsqrt instruction is not implemented.
867
868		----------------------------------------------------------------------
869		Name: Matthias Blume
870		Date: 2001/04/18 12:45:00 EDT
871		Tag: blume-20010418-ccalls
872		Description:
873
874		- fixed two off-by-4 errors in the x86-specific c-calls implementation
875		(this bug prevented structure arguments containing pointers from being
876		passed correctly)
877		- changed the raw-C-call code in mlriscGen.sml in such a way that
878		structure arguments are represented as a pointer to the beginning
879		of the structure (instead of having a series of synthesized arguments,
880		one for each structure member)
881
882		- made makeml script's verbosity level configurable via environment
883		variable (MAKEML_VERBOSITY)
884
885		- eliminated placeholder implementations for f32l, w16s, i16s, and f32s
886		in rawmem-x86.sml; we are now using the real thing
887
888		----------------------------------------------------------------------
889		Name: Matthias Blume
890		Date: 2001/03/22 16:25:00 EST
891		Tag: blume-20010322-bootfiles
892		Description:
893
894		Created a new set of bootfiles (for your automatic installation convenience).
895
896		----------------------------------------------------------------------
897		Name: Matthias Blume
898		Date: 2001/03/22 15:10:00 EST
899		Tag: blume-20010322-rawmem-parcm
900		Description:
901
902		1. All "raw memory access" primitives for the new FFI are implemented now
903		(at least on the x86).
904		2. Some further cleanup of CM's parallel make mechanism.
905
906		----------------------------------------------------------------------
907		Name: Matthias Blume
908		Date: 2001/03/19 17:53:00 EST
909		Tag: blume-20010319-parallel
910		Description:
911
912		Parallel make (using compile servers) now works again.
913
914		To this end, CM.stabilize and CMB.make have been modified to work in
915		two passes when compile servers are attached:
916		1. Compile everything, do not perform stabilization; this pass
917		uses compile servers
918		2. Stabilize everything; this pass does not use compile servers
919		If there are no compile servers, the two passes are combined into one
920		(as before).  Splitting the passes increases the inherent parallelism
921		in the dependency graph because the entire graph including all
922		libraries is available at the same time.  This, in turn, improves
923		server utilization.  The downside is that the master process will
924		have to do some extra work after compilation is done (because for
925		technical reasons it must re-read all the binfiles during stabilization).
926
927		----------------------------------------------------------------------
928		Name: Matthias Blume
929		Date: 2001/03/16 12:22:00 EST
930		Tag: blume-20010316-bootfiles
931		Description:
932
933		Created a new set of bootfiles (for your automatic installation convenience).
934
935		----------------------------------------------------------------------
936		Name: Matthias Blume
937		Date: 2001/03/16 11:00:00 EST
938		Tag: blume-20010316-MLTREE-fixup
939		Description:
940
941		This is a minor fixup for an (untagged) earlier commit by Allen.
942		(A file was missing).
943
944		----------------------------------------------------------------------
945		Name: Allen Leung
946		Date: Mon Mar  5 18:54:57 EST 2001
947		Tag: leunga-20010305-cut-support
948
949		1. New support for alternative control-flow in MLTREE.
950		Currently we support
951
952		FLOW_TO(CALL ...., [k1,...,kn])
953
954		This is needed for 'cuts to' in C-- and try/handle-like constructs
955		in Moby
956
957		New assembler flag "asm-show-cutsto" to turn on control-flow debugging.
958
959		2. Register Allocator
960
961		Changes in interface [from Fermin, John]
962
963		3. Alpha 8-bit SLL support [Fermin]
964
965		4. All architectures
966
967		A new module (ClusterExpandCopies) for expanding parallel copies.
968
969		----------------------------------------------------------------------
970		Name: Allen Leung
971		Date: 2001/02/27 23:07:00 EST
972		Tag: leunga-20010227-minor-stuff
973
974		1. Alpha bug fix for CMOVNE
975		2. Handle mltree COND(..,FCMP ...,...)
976		3. Bug fix in simplifier
977
978		----------------------------------------------------------------------
979		Name: Matthias Blume
980		Date: 2001/01/30 17:50:00 EST
981		Tag: blume-20010130-sync
982		Description:
983
984		This is just a minor update to sync my devel branch with the main brach.
985		The only visible change is the addition of some README files.
986
987		----------------------------------------------------------------------
988		Name: Matthias Blume
989		Date: 2001/01/12 23:30:00 JST
990		Tag: blume-20010112-bootfiles
991		Description:
992
993		Made a new set of bootfiles that goes with the current state of the
994		repository.
995
996		----------------------------------------------------------------------
997		Name: Matthias Blume
998		Date: 2001/01/12 21:20:00 JST
999		Tag: blume-20010112-sync
1000		Description:
1001
1002		I am just flushing out some minor changes that had accumulated in
1003		my private branch in order to sync with the main tree.  (This is
1004		mainly because I had CVS trouble when trying to merge _into_ my
1005		private branch.)
1006
1007		Most people should be completely unaffected by this.
1008
1009		----------------------------------------------------------------------
1010		Name: Allen Leung
1011		Date: Thu Jan 11 21:03:00 EST 2001
1012		Tag: leunga-20010111-labexp=mltree
1013		Description:
1014
1015		1.  Removed the type LabelExp and replace it by MLTree.
1016		2.  Rewritten mltree-simplify with the pattern matcher tool.
1017		3.  There were some bugs in alpha code generator which would break
1018		64-bit code generation.
1019		4.  Redo the tools to generate code with the
1020		5.  The CM files in MLRISC (and in src/system/smlnj/MLRISC)
1021		are now generated by perl scripts.
1022
1023		----------------------------------------------------------------------
1024		Name: Matthias Blume
1025		Date: 2001/01/10 21:55:00 JST
1026		Tag: blume-20010110-rcc
1027		Description:
1028
1029		The RCC stuff now seems to work (but only on the x86).
1030		This required hacking of the c-calls interface (and -implementation) in
1031		MLRISC.
1032
1033		Normal compiler users should be unaffected.
1034
1035		----------------------------------------------------------------------
1036		Name: Matthias Blume
1037		Date: 2001/01/09 01:20:00 JST
1038		Tag: blume-20010109-rcc
1039		Description:
1040
1041		This is a fairly big patch, flushing out a large number of pending
1042		changes that I made to my development copy over the last couple of days.
1043
1044		Of practical relevance at this moment is a workaround for a pickling
1045		bug that Allen ran into the other day.  The cause of the bug itself is
1046		still unknown and it might be hard to fix it properly, but the
1047		workaround has some merits of its own (namely somewhat reducing pickling
1048		overhead for certain libraries).  Therefore, I think this solution should
1049		be satisfactory at this time.
1050
1051		The rest of the changes (i.e., the vast majority) has to do with my
1052		ongoing efforts of providing direct support for C function calls from
1053		ML.  At the moment there is a new primop "RAW_CCALL", typing magic
1054		in types/cproto.sml (invoked from FLINT/trans/translate.sml), a new
1055		case in the FLINT CPS datatype (RCC), changes to cps/convert.sml to
1056		translate uses of RAW_CCALL into RCC, and changes to mlriscGen.sml to
1057		handle RCC.
1058
1059		The last part (the changes to mlriscGen.sml) are still known to be
1060		wrong on the x86 and not implemented on all other architectures.  But
1061		the infrastructure is in place. I had to change a few functor
1062		signatures in the backend to be able to route the CCalls interface
1063		from MLRISC there, and I had to specialize the mltree type (on the
1064		x86) to include the necessary extensions. (The extensions themselves
1065		were already there and redy to go in MLRISC/x86).
1066
1067		Everything should be very happy as soon as someone helps me with
1068		mlriscGen.sml...
1069
1070		In any case, nothing of this should matter to anyone as long as the
1071		new primop is not being used (which is going to be the case unless you
1072		find it where I hid it :). The rest of the compiler is completely
1073		unaffected.
1074
1075		----------------------------------------------------------------------
1076		Name: Matthias Blume
1077		Date: 2001/01/05 00:30:00 JST
1078		Tag: blume-20010105-primops
1079		Description:
1080
1081		Added some experimental support for work that I am doing right now.
1082		These changes mostly concern added primops, but there is also a new
1083		experimental C library in the runtime system (but currently not enabled
1084		anywhere except on Linux/X86).
1085
1086		In the course of adding primops (and playing with them), I discovered that
1087		Zhong's INL_PRIM hack (no type info for certain primops) was, in fact, badly
1088		broken.  (Zhong was very right he labeled this stuff as "major gross hack".)
1089		To recover, I made type information in INL_PRIM mandatory and changed
1090		prim.sml as well as built-in.sml accordingly.  The InLine structure now
1091		has complete, correct type information (i.e., no bottom types).
1092
1093		Since all these changes mean that we need new binfiles, I also bumped the
1094		version number to 110.32.1.
1095
1096		----------------------------------------------------------------------
1097		Name: Matthias Blume
1098		Date: 2000/12/30 22:10:00 JST
1099		Tag: blume-20001230-various
1100		Description:
1101
1102		Added proxy libraries for MLRISC and let MLRISC libraries refer
1103		to each other using path anchors.  (See CM manual for explanation.)
1104
1105		Updated CM documentation.
1106
1107		Fixed some bugs in CM.
1108
1109		Implemented "proxy" libraries (= syntactic sugar for CM).
1110
1111		Added "-quiet" option to makeml and changed runtime system accordingly.
1112
1113		Added cleanup handler for exportML to reset timers and compiler stats.
1114
1115		----------------------------------------------------------------------
1116		Name: Lal George
1117		Date: 2000/12/22 22:22:58 EST 2000
1118		Tag: Release_110_32
1119		Description:
1120
1121		Infinite precision used throughout MLRISC.
1122		see MLRISC/mltree/machine-int.sig
1123
1124		----------------------------------------------------------------------
1125		Name: Matthias Blume
1126		Date: 2000/12/22 23:16:00 JST
1127		Tag: blume-20001222-warn
1128		Description:
1129
1130		Corrected wording and formatting of some CM warning message which I
1131		broke in my previous patch.
1132
1133		----------------------------------------------------------------------
1134		Name: Matthias Blume
1135		Date: 2000/12/22 21:20:00 JST
1136		Tag: blume-20001222-anchorenv
1137		Description:
1138
1139		Fixed CM's handling of anchor environments in connection with CMB.make.
1140
1141		----------------------------------------------------------------------
1142		Name: Matthias Blume
1143		Date: 2000/12/22 13:15:00 JST
1144		Tag: blume-20001222-cleanup
1145		Description:
1146
1147		Removed src/cm/ffi which does not (and did not) belong here.
1148
1149	----------------------------------------------------------------------	----------------------------------------------------------------------
1150	Name: Matthias Blume	Name: Matthias Blume
1151	Date: 2000/12/21 23:55:00 JST	Date: 2000/12/21 23:55:00 JST

---

Colored Diff Long Colored Diff Unidiff Context Diff Side by Side

Legend: Removed from v.757   changed lines   Added in v.909

---

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