SCM Repository

[smlnj] Diff of /sml/trunk/HISTORY

[smlnj] / sml / trunk / HISTORY

Diff of /sml/trunk/HISTORY

-revision 916, Fri Aug 24 21:13:58 2001 UTC
+revision 1036, Fri Jan 25 22:05:44 2002 UTC
 Line 8
  The form of an entry should be:
  Name:
- Date:
+ Date: yyyy/mm/dd
  Tag: <post-commit CVS tag>
  Description:
  ----------------------------------------------------------------------
  Name: Matthias Blume
+ Date: 2002/01/25
+ Tag: blume-20020125-ffi
+ Description:
+ A large number of tweaks and improvements to ml-nlffi-lib and
+ ml-nlffigen:
+    - ML represenation types have been streamlined
+    - getter and setter functions work with concrete values, not abstract
+      ones where possible
+    - ml-nlffigen command line more flexible (see README file there)
+    - some bugs have been fixed (hopefully)
+ ----------------------------------------------------------------------
+ Name: Lal George
+ Date: 2002/01/24
+ Tag: george-20020124-risc-ra-interface
+ Description:
+    There is a dramatic simplification in the interface to the
+    register allocator for RISC architectures as a result of making
+    parallel copy instructions explicit.
+ ----------------------------------------------------------------------
+ Name: Matthias Blume
+ Date: 2002/01/22
+ Tag: blume-20020122-x86-ccalls
+ Description:
+ Bug fix for c-calls on x86 (having to do with how char- and
+ short-arguments are being handled).
+ ----------------------------------------------------------------------
+ Name: Matthias Blume
+ Date: 2002/01/21
+ Tag: blume-20020121-ff
+ Description:
+ Another day of fiddling with the FFI...
+. Bug fix/workaround:  CKIT does not complain about negative array
+    dimensions, so ml-nlffigen has to guard itself against this possibility.
+    (Otherwise a negative dimension would send it into an infinite loop.)
+. Some of the abstract types (light objects, light pointers, most "base"
+    types) in structure C are now eqtypes.
+. Added constructors and test functions for NULL function pointers.
+ ----------------------------------------------------------------------
+ Name: Matthias Blume
+ Date: 2002/01/18
+ Tag: blume-20020118-ready-for-new-release
+ Description:
+ Made config/srcarchiveurl point to a new place.  (Will provide boot
+ files shortly.)
+ Maybe we christen this to be 110.38?
+ ----------------------------------------------------------------------
+ Name: Matthias Blume
+ Date: 2002/01/18
+ Tag: blume-20020118-more-ffifiddle
+ Description:
+ Today's FFI fiddling:
+   - Provided a structure CGetSet with "convenient" versions of C.Get.* and
+     C.Set.* that use concrete (MLRep.*) arguments and results instead
+     of abstract ones.
+   - Provided word-style bit operations etc. for "int" representation
+     types in MLRep.S<Foo>Bitops where <Foo> ranges over Char, Int, Short,
+     and Long.
+ ----------------------------------------------------------------------
+ Name: Matthias Blume
+ Date: 2002/01/18
+ Tag: blume-20020118-use-x86-fp
+ Description:
+ Now that x86-fast-fp seems to be working, I turned it back on again
+ by default.  (Seems to work fine now, even with the FFI.)
+ Other than that, I added some documentation about the FFI to
+ src/ml-nlffigen/README and updated the FFI test examples in
+ src/ml-nlffi-lib/Tests/*.
+ ----------------------------------------------------------------------
+ Name: Allen Leung
+ Date: 2002/01/17
+ Tag: leunga-20020117-x86-fast-fp-call
+ Description:
+. Fixed a problem with handling return fp values when x86's fast fp
+       mode is turned on.
+. Minor pretty printing fix for cellset.  Print %st(0) as %st(0) instead
+       of %f32.
+. Added a constructor INT32lit to the ast of MLRISC tools.
+ ----------------------------------------------------------------------
+ Name: Matthias Blume
+ Date: 2002/01/16
+ Tag: blume-20020116-ffifiddle
+ Description:
+ More fiddling with the FFI interface:
+  - Make constness 'c instead of rw wherever possible.  This eliminates
+    the need for certain explicit coercions.  (However, due to ML's
+    value polymorphism, there will still be many cases where explicit
+    coercions are necessary.  Phantom types are not the whole answer
+    to modeling a subtyping relationship in ML.)
+  - ro/rw coersions for pointers added.  (Avoids the detour through */&.)
+  - "printf" test example added to src/ml-nlffi-lib/Tests.  (Demonstrates
+    clumsy workaround for varargs problem.)
+ ----------------------------------------------------------------------
+ Name: Lal George
+ Date: 2002/01/15
+ Tag: <none>
+ Description:
+. Since COPY instructions are no longer native to the architecture,
+    a generic functor can be used to implement the expandCopies function.
+. Allowed EXPORT and IMPORT pseudo-op declarations to appear inside a
+    TEXT segment.
+ ----------------------------------------------------------------------
+ Name: Matthias Blume
+ Date: 2002/01/15
+ Tag: blume-20020115-ffiupdates
+ Description:
+. Fix for bug resulting in single-precision float values being returned
+    incorrectly from FFI calls.
+. Small modifications to C FFI API:
+     - memory-allocation routines return straight objects (no options)
+       and raise an exception in out-of-memory situations
+     - unsafe extensions to cast between function pointers and pointers
+       from/to ints
+     - added structure C_Debug as an alternative to structure C where
+       pointer-dereferencing (|*| and |*!) always check for null-pointers
+     - added open_lib' to DynLinkage;  open_lib' works like open_lib
+       but also takes a (possibly empty) list of existing library handles
+       that the current library depends on
+ ----------------------------------------------------------------------
+ Name: Matthias Blume
+ Date: 2002/01/10
+ Tag: blume-20020110-newffigen
+ Description:
+. Updates to portable graph code.
+. Major update to ml-nlffigen and ml-nlffi-lib.  Things are much
+    more scalable now so that even huge interfaces such as the one
+    for GTK compile in finite time and space. :-)
+    See src/ml-nlffigen/README for details on what's new.
+ ----------------------------------------------------------------------
+ Name: Lal George
+ Date: 2001/01/09 14:31:35 EST 2002
+ Tag: george-20011206-rm-native-copy
+ Description:
+         Removed the native COPY and FCOPY instructions
+         from all the architectures and replaced it with the
+         explicit COPY instruction from the previous commit.
+         It is now possible to simplify many of the optimizations
+         modules that manipulate copies. This has not been
+         done in this change.
+ ----------------------------------------------------------------------
+ Name: Lal George
+ Date: 2001/12/06 16:50:13 EST 2001
+ Tag: george-20011206-mlrisc-instruction
+ Description:
+ Changed the representation of instructions from being fully abstract
+ to being partially concrete. That is to say:
+   from
+         type instruction
+   to
+         type instr                              (* machine instruction *)
+         datatype instruction =
+             LIVE of {regs: C.cellset, spilled: C.cellset}
+           | KILL of {regs: C.cellset, spilled: C.cellset}
+           | COPYXXX of {k: CB.cellkind, dst: CB.cell list, src: CB.cell list}
+           | ANNOTATION of {i: instruction, a: Annotations.annotation}
+           | INSTR of instr
+ This makes the handling of certain special instructions that appear on
+ all architectures easier and uniform.
+ LIVE and KILL say that a list of registers are live or killed at the
+ program point where they appear. No spill code is generated when an
+ element of the 'regs' field is spilled, but the register is moved to
+ the 'spilled' (which is present, more for debugging than anything else).
+ LIVE replaces the (now deprecated) DEFFREG instruction on the alpha.
+ We used to generate:
+         DEFFREG f1
+         f1 := f2 + f3
+         trapb
+ but now generate:
+         f1 := f2 + f3
+         trapb
+         LIVE {regs=[f1,f2,f3], spilled=[]}
+ Furthermore, the DEFFREG (hack) required that all floating point instruction
+ use all registers mentioned in the instruction. Therefore f1 := f2 + f3,
+ defines f1 and uses [f1,f2,f3]! This hack is no longer required resulting
+ in a cleaner alpha implementation. (Hopefully, intel will not get rid of
+ this architecture).
+ COPYXXX is intended to replace the parallel COPY and FCOPY  available on
+ all the architectures. This will result in further simplification of the
+ register allocator that must be aware of them for coalescing purposes, and
+ will also simplify certain aspects of the machine description that provides
+ callbacks related to parallel copies.
+ ANNOTATION should be obvious, and now INSTR represents the honest to God
+ machine instruction set!
+ The <arch>/instructions/<arch>Instr.sml files define certain utility
+ functions for making porting easier -- essentially converting upper case
+ to lower case. All machine instructions (of type instr) are in upper case,
+ and the lower case form generates an MLRISC instruction. For example on
+ the alpha we have:
+   datatype instr =
+      LDA of {r:cell, b:cell, d:operand}
+    | ...
+   val lda : {r:cell, b:cell, d:operand} -> instruction
+     ...
+ where lda is just (INSTR o LDA), etc.
+ ----------------------------------------------------------------------
+ Name: Matthias Blume
+ Date: 2001/11/22 21:40:00 EST
+ Tag: Release_110_37
+ Description:
+ Release 110.37.  This time for real.
+ ----------------------------------------------------------------------
+ Name: Matthias Blume
+ Date: 2001/11/21 16:35:00 EST
+ Tag: blume-20011121-foot-in-mouth
+ Description:
+ Removed the "Release_110_37" tag because of a serious bug.
+ This will be re-tagged once the bug is fixed.
+ ----------------------------------------------------------------------
+ Name: Matthias Blume
+ Date: 2001/11/21 16:14:00 EST
+ Tag: blume-20011121-forgottenfile
+ Description:
+ Forgot to add a file.  (Just a .tex-file -- part of
+ the CM manual source.)
+ ----------------------------------------------------------------------
+ Name: Matthias Blume
+ Date: 2001/11/21 16:10:00 EST
+ Tag: blume-20011121-invalid_110_37
+ Description:
+ Note: I removed the original tag "Release_110_37" from this commit
+       because we found a serious bug in all non-x86 backends.
+       - Matthias
+. Modifications to the SML/NJ code generator and to the runtime system
+    so that code object name strings are directly inserted into code
+    objects at code generation time.  The only business the runtime system
+    has with this is now to read the name strings on occasions.
+    (The encoding of the name string has also changed somewhat.)
+. CM now implements a simple "set calculus" for specifying export lists.
+    In particular, it is now possible to refer to the export lists of
+    other libraries/groups/sources and form unions as well as differences.
+    See the latest CM manual for details.
+. An separate notion of "proxy" libraries has again be eliminated from
+    CM's model.  (Proxy libraries are now simply a special case of using
+    the export list calculus.)
+. Some of the existing libraries now take advantage of the new set
+    calculus.
+    (Notice that not all libraries have been converted because some
+    of the existing .cm-files are supposed to be backward compatible
+    with 110.0.x.)
+. Some cleanup in stand-alone programs.  (Don't use "exnMessage" -- use
+    "General.exnMessage"!  The former relies on a certain hook to be
+    initialized, and that often does not happen in the stand-alone case.)
+ ----------------------------------------------------------------------
+ Name: Lal George
+ Date: 2001/11/21  13:56:18 EST
+ Tag: george-2001121-pseudo-ops
+ Description:
+   Implemented a complete redesign of MLRISC pseudo-ops. Now there
+   ought to never be any question of incompatabilities with
+   pseudo-op syntax expected by host assemblers.
+   For now, only modules supporting GAS syntax are implemented
+   but more should follow, such as MASM, and vendor assembler
+   syntax, e.g. IBM as, Sun as, etc.
+ ----------------------------------------------------------------------
+ Name: Matthias Blume
+ Date: 2001/11/14 11:52:00 EST
+ Tag: blume-20011114-srcname
+ Description:
+. Routed the name of the current source file to mlriscgen where it
+    should be directly emitted into the code object.  (This last part
+    is yet to be done.)
+. Some cleanup of the pgraph code to make it match the proposal that
+    I put out the other day.  (The proposal notwithstanding, things are
+    still in flux here.)
+ ----------------------------------------------------------------------
+ Name: Lal George
+ Date: 2001/11/14 09:44:04 EST
+ Tag:
+ Description:
+   Fix for a backpatching bug reported by Allen.
+   Because the boundary between short and long span-dependent
+   instructions is +/- 128, there are an astounding number of
+   span-dependent instructions whose size is over estimated.
+   Allen came up with the idea of letting the size of span
+   dependent instructions be non-monotonic, for a maxIter
+   number of times, after which the size must be monotonically
+   increasing.
+   This table shows the number of span-dependent instructions
+   whose size was over-estimated as a function of maxIter, for the
+   file Parse/parse/ml.grm.sml:
+      maxIter            # of instructions:
+                     687
+                     438
+                     198
+                       0
+   In compiling the compiler, there is no significant difference in
+   compilation speed between maxIter=10 and maxIter=40. Actually,
+   my measurements showed that maxIter=40 was a tad faster than
+   maxIter=10! Also 96% of the  files in the compiler reach a fix
+   point within 13 iterations, so fixing maxIter at 40, while high,
+   is okay.
+ ----------------------------------------------------------------------
+ Name: Matthias Blume
+ Date: 2001/10/31 15:25:00 EST
+ Tag: blume-20011031-pgraph
+ Description:
+ CKIT:
+ * Changed the "Function" constructor of type Ast.ctype to carry optional
+   argument identifiers.
+ * Changed the return type of TypeUtil.getFunction accordingly.
+ * Type equality ignores the argument names.
+ * TypeUtil.composite tries to preserve argument names but gives up quickly
+   if there is a mismatch.
+ installation script:
+ * attempts to use "curl" if available (unless "wget" is available as well)
+ CM:
+ * has an experimental implementation of "portable graphs" which I will
+   soon propose as an implementation-independent library format
+ * there are also new libraries $/pgraph.cm and $/pgraph-util.cm
+ NLFFI-LIB:
+ * some cleanup (all cosmetic)
+ NLFFIGEN:
+ * temporarily disabled the mechanism that suppresses ML output for
+   C definitions whose identifiers start with an underscore character
+ * generate val bindings for enum constants
+ * user can request that only one style (light or heavy) is being used;
+   default is to use both (command-line arguments: -heavy and -light)
+ * fixed bug in handling of function types involving incomplete pointers
+ * generate ML entry points that take record arguments (i.e., using
+   named arguments) for C functions that have a prototype with named
+   arguments
+   (see changes to CKIT)
+ ----------------------------------------------------------------------
+ Name: Allen Leung
+ Date: 2001/10/27 20:34:00 EDT
+ Tag: leunga-20011027-x86-fast-fp-call
+ Description:
+    Fixed the bug described in blume-20010920-slowfp.
+    The fix involves
+. generating FCOPYs in FSTP in ia32-svid
+. marking a CALL with the appropriate annotation
+ ----------------------------------------------------------------------
+ Name: Matthias Blume
+ Date: 2001/10/16 11:32:00 EDT
+ Tag: blume-20011016-netbsd
+ Description:
+ Underscore patch from Chris Richards (fixing problem with compiling
+ runtime system under recent NetBSD).
+ ----------------------------------------------------------------------
+ Name: Allen Leung
+ Date: 2001/10/12 17:18:32 EDT 2001
+ Tag: leung-20011012-x86-printflowgraph
+ Description:
+ X86RA now uses a valid (instead of dummy) PrintFlowgraph module.
+ ----------------------------------------------------------------------
+ Name: Lal George
+ Date: 2001/10/11 23:51:34 EDT
+ Tag: george-20011011-too-many-instrs
+ Description:
+ The representation of a program point never expected to see more
+ than 65536 instructions in a basic block!
+ ----------------------------------------------------------------------
+ Name: Lal George
+ Date: 2001/10/09 09:41:37 EDT
+ Tag: george-20011008-mlrisc-labels
+ Description:
+ Changed the machine description files to support printing of
+ local and global labels in assembly code, based on host assembler
+ conventions.
+ ----------------------------------------------------------------------
+ Name: Matthias Blume
+ Date: 2001/09/25 15:25:00 EDT
+ Tag: blume-20010925-exninfo
+ Description:
+ I provided a non-hook implementation of exnName (at the toplevel) and
+ made the "dummy" implementation of exnMessage (at the toplevel) more
+ useful: if nothing gets "hooked in", then at least you are going to
+ see the exception name and a message indicating why you don't see more.
+ [For the time being, programs that need exnMessage and want to use
+ ml-build should either use General.exnMessage (strongly recommended) or
+ refer to structure General at some other point so that CM sees a
+ static dependency.]
+ [Similar remarks go for "print" and "use":  If you want to use their
+ functionality in stand-alone programs generated by ml-build, then use
+ TextIO.output and Backend.Interact.useFile (from $smlnj/compiler.cm).]
+ ----------------------------------------------------------------------
+ Name: Matthias Blume
+ Date: 2001/09/20 17:28:00 EDT
+ Tag: blume-20010920-slowfp
+ Description:
+ Allen says that x86-fast-fp is not safe yet, so I turned it off again...
+ ----------------------------------------------------------------------
+ Name: Matthias Blume
+ Date: 2001/09/20 17:20:00 EDT
+ Tag: blume-20010920-canonicalpaths
+ Description:
+. Updated the BOOT file (something that I forgot to do earlier).
+. Small internal change to CM so that it avoids "/../" in filenames
+    as much as possible (but only where it is safe).
+. Changed config/_run-sml (resulting in a changed bin/.run-sml) so
+    that arguments that contain delimiters are passed through correctly.
+    This change also means that all "special" arguments of the form
+    @SMLxxx... must come first.
+. Changed install script to put relative anchor names for tool commands
+    into pathconfig.
+ ----------------------------------------------------------------------
+ Name: Matthias Blume
+ >>>>>>> 1.169
+ Date: 2001/09/18 15:35:00 EDT
+ Tag: blume-20010918-readme11036
+ Description:
+ Added README files.
+ ----------------------------------------------------------------------
+ Name: Matthias Blume
+ Date: 2001/09/18 11:45:00 EDT
+ Tag: Release_110_36 (retag)
+ Description:
+ Fixed mistake in config/preloads. Retagged as 110.36.
+ ----------------------------------------------------------------------
+ Name: Matthias Blume
+ Date: 2001/09/18 09:40:00 EDT
+ Tag: Release_110_36_orig (tag changed)
+ Description:
+ New version (110.36).  New bootfiles.
+ ----------------------------------------------------------------------
+ Name: Matthias Blume
+ Date: 2001/09/14 16:15:00 EDT
+ Tag: blume-20010914-x86fastfp
+ Description:
+ John committed some changes that Allen made, in particular a (hopefully)
+ correctly working version of the x86-fp module.
+ I changed the default setting of the Control.MLRISC.getFlag "x86-fast-fp"
+ flag to "true".  Everything seems to compile to a fixpoint ok, and
+ "mandelbrot" speeds up by about 15%.
+ ----------------------------------------------------------------------
+ Name: Matthias Blume
+ Date: 2001/09/13 11:20:00 EDT
+ Tag: blume-20010913-minimal
+ Description:
+. Stefan Monnier's patch to fix a miscompilation problem that
+    was brought to light by John Reppy's work on Moby.
+. Implemented a minimal "structure Compiler" that contains just
+    "version" and "architecture".  The minimal version will be
+    available when the full version is not.  This is for backward-
+    compatibility with code that wants to test Compiler.version.
+ ----------------------------------------------------------------------
+ Name: Matthias Blume
+ Date: 2001/08/28 14:03:00 EDT
+ Tag: blume-20010828-ml-lex
+ Description:
+ Fix for bug 1581, received from Neophytos Michael.
+ ----------------------------------------------------------------------
+ Name: Matthias Blume
+ Date: 2001/08/27 11:20:00 EDT
+ Tag: blume-20010827-readme11035
+ Description:
+ Fleshed out the README file for 110.35.
+ ----------------------------------------------------------------------
+ Name: Matthias Blume
  Date: 2001/08/24 17:10:00 EDT
  Tag: Release_110_35
  Description:
-Line 2923
+Line 3503
       elaborator). There were a lot of changes during my "linkpath" trials
       that could have been reverted to their original state but weren't.
       Please, don't be too harsh on me for messing with this code a bit more
-      than what was strictly necessary...  (I _did_ resist the tempation
+      than what was strictly necessary...  (I _did_ resist the temptation
       of doing any "global reformatting" to avoid an untimely death at
       Dave's hands. :)

 Legend:



Removed from v.916
 


changed lines


 
Added in v.1036
 Legend:



Removed from v.916
 


changed lines


 
Added in v.1036
-Removed from v.916
+Added in v.1036

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