Added <stdlib.h> include to fix warning.

improved error handling in CM

fixed problem with config/preloads (html-lib)

*** empty log message ***

(sml-find-match-forward): Avoid infinite looping if the
construct is not properly closed (yet).

moved release tag

fixed installation/configuration glitches

(sml-mode-menu): Remove left over obsolete entries.

(inferior-sml-mode): Preserve the menu-bar of the
compilation minor mode, if any.

added note about eXene

new working version 110.57; new bootfiles

fixed problem with bogus exn message when using btrace

added implementation of Array2.copy

fixed bootstrap problem with previous change

improved the way certain libraries are bound into the compiler

fixed bug in xxxArraySlice.update; pushed Basis changes through

  Remove Substring.all.

  Replaced uses of Substring.all with Substring.full.

  Switch uses of Substring.all to Substring.full.

  Tweaking of the interval set API.

  Tweaking of the interval set API; see CHANGES file for details.

updated runtime system, CML, and eXene to reflect recent changes to Basis

  Rename Pack* structures to PackWord*.

  Create heap and lib directories when needed.

  Fixed bug in intersect function.

  Basis fix.

  Basis semantics fix: hexidecimal digits should be upper case.

added readme files for 110.56

BOOT and HISTORY entries

freeze for 110.56

  Fix typo.

  Tracking SML/NJ library changes.

  Adding interval sets to CM file.

  SML/NJ library additions.

  Resolve conflict.

Add Zhong Shao's fix for datatype equality functions.

    Fixed the floating point bug that Carl Hauser discovered.
    Bug was a typo in the spilling code.

   Removed some debugging printout code from x86Asm.sml.
   It looked like something left over from previous bug fixes.

*** empty log message ***

  Renamed graph-scc.sml to graph-scc-fn.sml.

  Fixed ml-lex to recognize "\r" in REs.

  Fixed ml-yacc to work with non-native eol encodings.

  Fix some odds and ends with the PP library.

added readmes for 110.55

  HPUX 11 support.

  Add HPUX 11 support.

  New makefile to support HPUX 11.

version number bump

  Changed interface to setting mode.

  New file.

  Describe atom type.

  Update for ANSITerm pretty printing support.

  Added PP stream based on ANSITermDev.  ANSITermDev now sets the initial
  mode based on the kind of the output stream (TTY enables styles).

pushed minor library changes into CML

  Changes to SML/NJ library.

  New file.

  Added ANSITerm structure.

  Added ANSITerm structure.

  Refactored PP library.

fixed problem (reversed logic) in copy code in array slice modules

     Fixed a typo in the cygwin code.

     Updated cygwin's signal/fault handling to match window's code.

  Bug fix: the calculation of required space did not take sequence headers
  into account.

new Mac OS X instructions

added READMEs for 110.54

fixed problems with new signal handling code (win32 version was incomplete)

freeze for release 110.54

added some script support for Mac OS X package installer

workaround for overloading bug

added mechanism for re-dumping heap image for interactive system

fixed Word64.fromString

  Add Tiger (MacOS X 10.4) support.

  Build ckit by default.

  Added singleton function to MONO_PRIORITYQ interface.

  Allow heap triming to be turned off.  WARNING: doing so may result in
  unbounded heap-growth, even for constant-space programs!

  Change in sizing calculation to reduce number of major GCs.

     GC protocol checking phase added.

fixed bug in IntInf.div

added join to parser combinators

  Remove debugging flag.

  New scheme for managing signals in the runtime system that should avoid
  race conditions (thanks to Cynbe for the design).

  Removing makefiles for unsupported architectures

  Removing asm code for unsupported architectures.

  Remove spurious generated file.

  Bug fix: Pointers to heap objects were not being preserved across GC.

  Include faulting PC in bogus fault error message.

  Fix for spilling single-precision floats.

  Remove excess newline.

  Fixed bug in GetOpt.

Fallback from amd64 to x86 under GNU/Linux.
The same should probably be done for other OSes.

fixed silly bug in CML:Atom

minor update to readme

dlopen trick does not work(yet) for win32

readmes for 110.53

new working version (110.53)

brought back SMLofNJ.Susp

fixed/improved structure Atom in CML

moved TraceDebugProf where it belongs

  Two bugs fixed: the floating-point stack needs to be killed by calls
  and %st(0) was being returned as the result register (instead of the
  freshly allocated pseudo register).

  Added Atom.same and Atom.lexCompare functions.

long long results implemented

long long args now seem to work

long long arguments to C functions (still not working properly)

experimental code for passing long long arguments via NLFFI (untested)

fixed bug in ml-yacc (as patterns)

fixed problem with time profiler

removed line from mk.x86-linux as it does not work as intended

fixed typo

new experimental heap2exec facility

fixed off-by-one error in globals.c

robustified link-sml script against arguments with spaces

update to ml-build script

missing files

(partially) implemented long long in nlffi

cleanup of c/memory library

   .run-sml now converts SMLNJ_HOME to a POSIX pathname on cygwin.

missing line in HISTORY

last-minute changes incorporated

updated README files to reflect last-minute-change

bug fix in statistics module

minor modifications to smlnj-lib (concerning recent changes)

  Changed HashString.hashString' to hashSubstring.

added README files

freeze for 110.52

adjusted winlibinstall.sml to match new interface for generic installer

some bullet-proofing of .bat scripts against spaces in filenames

tried to make installer cope with spaces in filenames (incomplete)

fixed bug in ml-makedepend

  Fixed infinite loop.

  Adding Allen's changes to support extensions on the PPC.

attempt to restore backward compatibility with 110.0.x (CML)

added two simple stats modules to smlnj-lib

*** empty log message ***

minor modifications/improvements to hashing and atoms

made CML 64-bit-position compliant

fixed typo in HISTORY

support for 64-bit Position.int (not enabled by default)

*** empty log message ***

*** empty log message ***

problem with unhelpful error messages addressed

(sml-font-lock-symbols-alist): Add yet more silly entries.

README files for 110.51

fixed awk bug in ml-makedepend

freeze for working version 110.51

enabled dlopen (for NLFFI) on FreeBSD

    Added support to LIVE and KILL for all the architectures.

minor improvement to the version tool

cosmetic issue

(sml-font-lock-symbols-alist): Add entries for simple type variables.
(sml-font-lock-compose-symbol): Add support for non-punctuation syms.

(sml-font-lock-symbols-alist): Add entries for >= and <=.

(sml-font-lock-symbols, sml-font-lock-symbols-alist): New vars.
(sml-font-lock-compose-symbol, sml-font-lock-symbols-keywords): New funs.
(sml-font-lock-keywords): Use them.

(sml-error-regexp-alist): Merge regexps.
(sml-update-cursor): Check sml-temp-file is non-nil.

(sml-backward-sexp, sml-forward-sexp): Remove unused var.

(temp-directory): Get rid of warning.

(sml-calculate-indentation): Remove unused var.
(sml-get-indent): Change first arg's meaning.
(sml-get-sym-indent): Adjust call.
(sml-forms-menu): Simplify.

(sml-make-error): Use match-string.

(install_startup): Don't add to load-path any more.

improved the version tool by simplifying it

fixed an issue with polyequal

added full implementation of Int64

more 64-bit hacking (see HISTORY)

added 64-bit mult and partial implementation of div

most of 64-bit word support is there now

more 64-bit hacking (still in the middle of it, though)

started building support for 64-bit types

some code cleanup

  Add missing cases for stm pretty printing.

  Add support for NEG in LABEXPs.

README files for 110.50

missing entry in BOOT

code freeze for 110.50

  Don't bother to adjust the stack pointer by zero.

  Added documentation of calling convention for struct result functions.

improved BackTrace.monitor

HISTORY entry for x86 c-calls updates

hot-fix for struct-return problem (seems to work)

  Removed debugging code that got committed by mistake.

  Fix regions in case they mean something someday.

added missing LOAD

  Possible fix for struct parameters in C calls.

    GAS syntax fix for FUCOMIP and friends.

eliminated redundant test in Int31.scan

deal with minInt correctly in Int31.fromString

specialized version of Int.fromString added (2x as fast)

support for back-trace in standalone programs

eliminated group owners; added warning

correct type of tdp_register

further TDP reorganization; coverage tool added

overhaul of Core and SMLofNJ support for TDP

spelling error

some renaming and cleanup (trace/debug/profile instrumentation)

minor code improvement

additional improvement to region info propagation

more fiddling with imprecise region annotations

eliminate unneeded argument

better region info propagation in elaborator

fixed problem with keyword handling in .cm files

minor cleanup

cleanup work

minor cleanup in absyn and btrace

made slave mode work with version tool

cleanup of the auto-version stuff

integrated versiontool

automagic generation of version.sml

added comment

some cleanup of the controls code

improved Controls interface; minor changes and bug fixes in CM

CM support for lazy sml and per-file control settings

turned MLRISC message (PPC) about long branches off by default;
made setting configurable

rounding mode patch for OS X

tweaked interface of ControlRegistry; adde command-line switch to sml

README files for 110.49

working version 110.49

turned target mlrisc on by default again

  Fixed the signal masking code to properly nest mask/unmask operations
  on a per-signal basis.

bumped heap image magic number

   Added a patch to _arch-n-opsys to enable the Cygwin runtime.
   The Cygwin runtime is turned on by setting the environment
   variable SMLNJ_CYGWIN_RUNTIME to 1.

added some exports to core.cm

installer improvements

added ptreql to InlineT

   Another bug fix to the graph library

   Minor PPC syntax fix.

  Fixing one more bug with allocation of the argument area.  Seems to
  work now (at least for Moby).

  Fix bug: argument area allocation not getting generated.

fixed wrong URL

  Get the sizes of integer return values right.

fixed typo in BOOT

added html version of README for 110.48

README for 110.48

freeze for Release 110.48

  Partial fix for signal masking code.

applied bug fix provided by Benjamin Ylvisaker

    Bug fix to dijkstra's algorithm.

  Fixed bug in mandelbrot program.  Note that this change affects the
  running time of the benchmark!

fixed pattern-matching bug and added missing cases

some improvements to handling of cmd-line arguments

  Initial support for PPC extensions.

     Added IBM syntax support for PowerPC.

moved 2 items from compiler internals to MLRISC (where they belong)

fixed obsolete comment in targets

removed warning about nlffi from WININSTALL

removed mention of e-mail address

added README files for 110.47

freeze for 110.47

added support for multiple C calling conventions

HISTORY entry

rolled back change to get_file_time (seems wrong)

dldummy is gone!

more silly typos

more missing edits

added missing update to dlclose.c

added support for nlffi under win32

assorted bug fixes

more manual hacking

generate E_* for all enums, not just the ones that are referenced

sync

slowly fleshing out the nlffi manual

added naturalIntSz to dummy implementation of c-calls

  Fixed alignment of empty struct/union types to be 1 (not 4).

  Added comment to C_CALLS signature documenting the responsibility of the
  client to handle sign extension.  Also added a "naturalIntSz" value
  that the client can use to determine the integer promotion size.
  Updated the implementations to match the changed signature and removed
  sign extension from the x86 implementation.

minor tweak to ml-nlffigen (treatment of typedef);
added description of typedef to manual

  Reimplementation of CCalls for x86.  This version uses the layout function
  to assign argument locations.

  Added documentation.

sync some changes

added some (slightly) better error reporting to ml-nlffigen

nlffi documentation (a start); minor bugfix in ml-nlffigen

template for a NLFFI manual added

Changes to allow multiple lines
in the asm: specification. Now it is possible
to write:
	asm: '' ...''
	     ...
             '' ... ''

handled C_UNION on the Sparc

moved dummy implementation of c-calls into MLRISC tree

  Rewrote handling of C_UNION type.

implemented C_UNION for ppc-macosx;
fixed bug in propagating C_UNION to backend

added C_UNION to c-types.sml

fixed bug: functions fpr and gpr were not in scope where they were used

   More PPC instructions added.

  Minor edits.

  Support for struct arguments.

added some more testcases for nlffi

fixed bug in usage of paramAlloc

   Added LWARX and STWCX to the Power PC instruction set.

NLFFI-related code tweaks

  Changed signature of layout function; added paramAlloc callback to
  PPC genCall function; eliminated non-exhaustive matches.

  Fixed spelling error in comment and remove unecessary list reversal.

nlffi under mac os x; NOW BOOTFILES; version 110.46.1

  Working on adding support for structure arguments.

  Additional work on PPC C-calls and minor change to the return type
  of the layout function in the C_CALLS interface.

updated comment

  Fixed type errors in PPC C-calls.

  Working on PPC C-Calls.

  Add missing components to C-calls stub.

  Working on PPC C calls.

  Working on C calls.

  Working on C calls.

  Working on C Calls for PPC/MacOS X.

  Added layout function to C_CALLS signature.

fixed typo

changed implementation of structure Unix to non-compliant (but perhaps
saner) behavior

added README files for 110.46

bumped version number to 110.46

structure Unix now up to (the broken) spec

structure Timer now up to spec

  Better error message.

  Handle exec failure.

  Getting example programs to compile again.

  Fixe type error introduced by switch to InfInt.int.

incremental Basis update

handling of ownership rule for CM groups changed

minor code cleanup

made win32 version work again

cleanup code added to ml-build.bat

made ml-build clean up after itself better

added missing implementations of Posix.IO.mk{Bin,Text}{Reader,Writer}

added message from David Hansel to HISTORY file

added missing socket-functionality for win32

fixed bug in installml

added support for NLFFI under NetBSD

extended some export lists

fixed installer code so that src-smlnj works again

fixed IEEEReal.scan

minor revision of MAP file

fixed problem with ml-build not returning error status

fixed problem with two different versions of exception Option

(sml-error-regexp-alist): Use new compile.el features if available.
(sml-send-region): Remove unused var `file'.
(sml-drag-region): Remove unused function.
(sml-update-cursor): Use compilation-fake-loc if available.
(compilation-goto-locus): Only advise if necessary.

Add self to load-path in sml-mode-startup.el.
(sml-def-skeleton): Try to use the new `system' abbrev flag.

updated MAP and added INDEX in src/compiler

* testcases.sml: Add a nasty case that is still wrong.

* sml-proc.el (sml-error-regexp-alist): Tune the regexp for sml/nj.

* sml-mode.el (sml-mode): Batter handle paragraphs in comments.
(sml-mode-variables): Clean up paragraph settings.
(sml-electric-pipe): Fix a boundary bug.
(sml-indent-starter, sml-get-sym-indent, sml-find-matching-starter):
Fix indentation algorithm.  Can't remember what it was about.
This code is nasty, I just can't understand what's doing what.
(completion-ignored-extensions): Add the new .cm directory.

*** empty log message ***

    Fixed the cygwin linking process so that ml-nlffi works.
    CVS tag: leunga-20040319-cygwin-nlffi

adding comments

fixed problem with IntInf.fmt

added HISTORY entry for previous change

fixed symlink problem in bin directory reported by Vesa A Norrman

added README files for 110.45

working version 110.45

  Fix CPP command in x86 makefiles (remove -ansi flag)

improved handling of exceptions at interactive toplevel

type of top-level app (HISTORY entry)

type of top-level app

  PPC CCalls hacking.

  Fixed spelling mistakes.

  Cross machine sync (working on PPC C calls)

  A bit of work on the PPC/MacOS X C Calls implementation.

  Added case for comparisons between int32 constants.

Another Subscript exception in getpair!
Is there any chance that I am the only one using this function ;-)

Lal
----------

Fixed a bug in which getpair raises an Subscript exception!

Lal
-----

Made changes to getpair.
Not all available pairs were checked by the find routine.

Further, some architectures have a requirement that pairs
be even registers and others don't.
When a pair is found, by setting the next point to search
from using:

      next = found + 1

a client could call this routine again if the even/odd
requirement is not met the first time.

Lal

*** empty log message ***

(sml-mode-menu): Fix typo.  Use std names.
(sml-begin-syms-re): Reduce redundancy.

HISTORY entry

POSIX_FLAGS -> BIT_FLAGS

tentatively changed interface of Timer to what has been discussed
on the Basis mailing list

fix spelling of URLs

added README files for 110.44

Working version 110.44

  Fixed script to be less sensitive to the order in which include
  files are examined for constants.

eliminated dont_move_libraries directive in config/targets

made installer honor INSTALLDIR variable properly

  Added case for MacOS X 10.3 (Panther).

MLRISC bug fix from Lal

typo in HISTORY

PRIM_IO; .bat-files

added a missing privilege wrapper

minor version bump; new bootfiles

implemented autoloading on behalf of prettyprinter

fixed typo in Win32 install script

installer tweaks (see HISTORY)

removed non-blocking ops from CML-specific Socket implementation

sync of socket implementation with Basis spec;
minor updates to Real64 code

incorrect tag name fixed

new interim version 110.43.2, new bootfiles

CM -> .cm; new-install.sh -> install.sh (see HISTORY and INSTALL for details)

made interim version 110.43.1 (with new bootfiles)

some improvements to modules dealing with time- and real values

some win32-specific updates related to previous commit

various Basis-related changes (see HISTORY)

code that actually implements int32->real64 (in code generator)

added primop for int32->real64 conversion

fixed bug in Real.fromLargeInt

undoing accidental commit of files intended for primop-branch

eliminate type from inline info

eliminate types from primop info

minor bugfix in new installer script

type of inputLine changed; minor bugfixes; minor installer improvements

added link to sources to README files

added MACOSXINSTALL; updated other installation instructions

Release_110_43

renamed unpack.sh as unpack to avoid a strange problem with CVS

major installer improvements: portable install script,
written in SML, which works for both Unix and Win32

fixed bug in OS.FileSys.fullPath (win32 version)

another typo... :-(

fixed typo

a lot of Basis hacking, among other things adding slices

dot.sml committed per request by Lal

added WININSTALL; fixed spelling mistakes in HISTORY

implemented IntInf in Basis and compiler;
new version number; new bootfiles

  New file.  This is a conversion from the Sparc version and is
  a "work in progress."

merged changes from mcz-branch (tag: dbm-20030813-mcz-merge1)

updated HISTORY

works under w32 again;
bug fix for ml-nlffigen; other bugfixes

  Added getDevice function to PP_STREAM signature and fixed coding style.

creating tag macqueen-20030617-bug895

fixed bug 895

eliminated many cases of poly-eq

major cleanup: eliminate all non-exhaustive pattern matches

  Logic cleanup in Scan.scanf.

added IF/WHILE/ANDALSO/ORELSE to Absyn (to improve error messages)

   Some MIPS backend stuff regenerated from the machine description tool.

     Cygwin port of smlnj

    Completely non-functional MIPS backend.

  Added a mechanism for thread-local data based on the Week's style
  property lists.

  Start of documentation for RunCML module.

  Added missing StreamIO input event constructors.

  Bug fix for CleanUp.logMailbox.

  Documentation cleanup.

added entry to CHANGES file for CML;
updated srcarchiveurl to point to Chicago

avoid name conflict with CM metadata on case-insensitive file systems

deleted ir.cm (per instruction from Lal)

added missing elements to ListPair

added mlrisc target to install script

  Fix substitution.

  Documentation cleanup for CML.

Remove outdated `sml-load-hook'.

Remove outdated `sml-load-hook'.

  Fixed some deadlinks in the doc.

    Bug in Int.rem(x,y) when y is a power of 2.

    New files for the cygwin port.

fixed bug in RISC RA related to rewrite code for spills and reloads
of floating point registers;  nucleic now compiles again on the Sparc

merged hot-fix for a CM problem related to file IDs

exported Typecheck

  Two fixes: changed bad character error message to properly print the
  bad character and fixed skipws to use Char.isSpace test.

correct README.html (but I will not re-tag again for this, sorry)

brown-paper-bag bug fixed, 110.42 updated

  removed darwin5 for Mac OS 10.1: fixes build problem

Release_110_42

  First cut at a README for 110.42.

  Minor cleanup of MacOS X build.

  Fix configuration for MacOS X 10.1.

  opsys suffix is darwin5 for macOS 10.1

  Added missing SigReturn_t type for MacOS X 10.1 (darwin5) build.

fixed divide-by-zero problem on PPC (or so I hope...)

  Added backward compatibility for MacOS X 10.1 (called darwin5).

  Make cml and cml-lib part of the default build.  Also, remove doc target,
  since it is not part of the working distribution.

From Allen: Added a flag to dump the cfg after back patching
            and delay slot filling.

  MacOS X 10.2 porting work.

  Fixed %f format bug.

  Fixed space leak in stream I/O.

  bug fix from Allen related to creating parallel copies

  Fixed makefile for MacOS X 10.1

  Fixed code for MacOS 10.2.  This still needs to be tested on 10.1.5.

  bug in aliasing cells to cc

  added flag to preserve dead copies during spilling

added handling of SMLNJ_HOME

bug fix from Allen

exported two useful functions from MLRISC Liveness modul

  use a log scale for colors when frequencies are large

added README file for 110.41

working version 110.41

more of the same: tweak installer to look for tgz files first

tweak installer to look for tgz files first

structure BTImp exported

fixed bug in FP translation of INLMAX and INLMIN

  Sync library with master copy.

  More GC space-leak fixes with some performance tuning.

  Fix sizing policy implementation.

  Rationalize max sizes for arenas.

back-ported OS.Path.xxxUnixPath from idlbasis-devel branch

back-ported IEEEReal.fromString from idlbasis-devel branch;
small (and currently broken) update to ppDec wrt. printing of functor
signatures

The empty bitmatrix must contain an table size that is
a non-zero power of two. The table is made bigger by
doubling its size, so an initial size of zero results
in a nice infinite loop.

brought BTrace.mode back

  Added header comment.

  Added header comments.

  Fixed for old-style CM.

  Coding conventions.

  Missing include file.

  Fixed minor bug in chunk trimming.

Labels are now displayed in the graphical output to make
the fall-through and target blocks obvious.

HISTORY entry for runtime bugfix

  Made the heap shrink test more robust.

forgot to mention eXene update; retagged

working version 110.40

  Added a hack to fix bug 1131.  The basic idea is to force an extra
  GC in certain places, which then has the effect of shrinking the heap.

CM documentation update

more preloads tweaks

  Modified heap exporting so that it is quiet unless GCMessages are enabled.

  Fixed bug in scanning of hex literals.

more controls cleanup; removed some wrongly-added things from config/preloads

  New files.

  Upload changes to controls library.

  fixed machine code bug in right shift instructions

added suggestions and comments to config/preloads

     Changed mdl gen to make it compile.
     x86 asssembly fix for fcmpi.

	The MLRISC Tools directory was not aware of the changes
to MLTREE.DIVS, MLTREE.REMS, and MLTREE.DIVT that added
rounding modes to each of these operators.

pushed new controls interface through compiler

  Bug fixes.

  Sync with master repository.

adjust to new DIV/MOD with rounding mode

  fixed syntax error.

  Added missing case.

  Added tyName field to ValueSyntax exception in Controls module.

  Sync with master repository (see CHANGES for details).

  Bringing the Yale repository upto date (see CHANGES file for details).

        All pseudo-ops emitted before the first segment declaration
such as TEXT, DATA, and BSS directives are assumed to be global
declarations and are emitted first in the assembly file. This is
useful in a number of situations where one has pseudo-ops that are not
specific to any segment, and also works around the constraint that one
cannot have client pseudo-ops in the TEXT segment.

        Because no segment is associated with these declarations it is
an error to allocate any space or objects before the first segment
directive and an exception will be raised. However, we cannot make
this check for client pseudo-ops.

        These top level declarations are a field in the CFG graph_info.
In theory you can continue to add to this field after the CFG has been
built -- provided you know what you are doing;-)

deleted bogus fc directory again

some minor bugfixes

textio bugfix from Emden applied

  XTime bug fix and new scrollbar implementation from Allen Stoughton.

  Added missing feature flags for Darwin (MacOS X).

make X86.prim.asm compile without warning on RH7.x systems

more div-related improvements to the x86 instruction selector

fixed bug in new div/mod logic

division primops, inline primops for min/max/abs

  Switch to CFG.exitId.

provided MLRISC support for all four division ops (div/mod/quot/rem)

improved implementation of div, mod, and rem

fixed bug in cproto

raw-primop cleanup

  Minor fixes to get the block list in a consistent form.

  inserted block with a single jump instruction to break multiple fallthroughs

made a new set of bootfiles

    Added some new primops for creating/manipulating temporary "C" objects
    allocated on the ml heap.

Edge colors  during graphical viewing is based on
execution frequency of the edge. I used the same
color scale employed in yahoo weather maps.

  Added newNode function.

 fixed some of my typos

added missing file

  Fix to handle fallsthru edges to block 0.

Implemented the Ball-Larus branch prediction-heuristic, and
incorporated graphical viewers for control flow graphs.

Ball-Larus Heuristic:
---------------------
See the file compiler/CodeGen/cpscompile/cpsBranchProb.sml.

By design it uses the Dempster-Shafer theory for combining
probabilities.  For example, in the function:

    fun f(n,acc) = if n = 0 then acc else f(n-1, n*acc)

the ball-larus heuristics predicts that the n=0 is unlikely
(OH-heuristic), and the 'then' branch is unlikely because of the
RH-heuristic -- giving the 'then' branch an even lower combined
probability using the Dempster-Shater theory.

Finally, John Reppy's loop analysis in MLRISC, further lowers the
probability of the 'then' branch because of the loop in the else
branch.


Graphical Viewing:
------------------
I merely plugged in Allen's graphical viewers into the compiler. The
additional code is not much. At the top level, saying:

	Control.MLRISC.getFlag "cfg-graphical-view" := true;

will display the graphical view of the control flow graph just before
back-patching.  daVinci must be in your path for this to work. If
daVinci is not available, then the default viewer can be changed
using:

	Control.MLRISC.getString "viewer"

which can be set to "dot" or "vcg" for the corresponding viewers. Of
course, these viewers must be in your path.

The above will display the compilation unit at the level of clusters,
many of which are small, boring, and un-interesting. Also setting:

	Control.MLRISC.getInt "cfg-graphical-view_size"

will display clusters that are larger than the value set by the above.

  rewrote the cfg dump to use CFG.dump

  Check EXIT edges too.

 treat entry and exit nodes specially

  Use CFG.entry and CFG.exit functions.

  Added code to check JUMP and BRANCH true edges.

  Added methods to get unique entry and exit blocks.

  New file.

  New file -- debugging code to check correctness of block placements.

kmp bugfix

     Improvements to cfg

  New default block placement that respects edge kinds.

     Bug fixes and improvements to x86 and cfg

KMP for string/substring matching

Switch probability module to use IntInf.

  Added not function and modified combineProb2 to avoid potential
  overflow.

implemented generic conversions in ml-nlffi-lib based on witness values

  Fixed flag names for dumping CFG.

  Fixed LEH (got the sense backwards).

  New version implements both Loop Branch and Look Exit heuristics.

provided missing String and Substring functions

  Added functor for computing loop-related branch probabilities.

  Added copyright.

moved Controls module into its own library;
changed Controls interface to make it saner;
made MLRISC independent of Controls;
minor changes to CM

  Added a function (combineProb2) for computing the combination of
  probability estimates using Dempster-Shafer theory.

  Allen's fixes to maintain frequency information.

    X86 code generator did not correctly signal the ra to run.

  adapted to new jump chaining interface

bug in installer fixed; overload tweaks

  Added flags to functor parameters to control hte behavior of the
  algorithm.  The reverseDirection flag is currently a nop.

BOOT and srcarchiveurl

implemented better, simpler PID allocation scheme;
binfile format has changed, version number is 110.39.2; NEW BOOTFILES

Integrated jump chaining and static block frequency into the
compiler. More details and numbers later.

  Added functor to compute block and edge frequencies from edge
  probabilities.

  pick between default and weighted block placement

Tested the jump chain elimination on all architectures (except the
hppa).  This is on by default right now and is profitable for the
alpha and x86, however, it may not be profitable for the sparc and ppc
when compiling the compiler.

The gc test will typically jump to a label at the end of the cluster,
where there is another jump to an external cluster containing the actual
code to invoke gc. This is to allow factoring of common gc invocation
sequences. That is to say, we generate:

	f:
	   testgc
	   ja	L1	% jump if above to L1

	L1:
	   jmp L2


After jump chain elimination the 'ja L1' instructions is converted to
'ja L2'. On the sparc and ppc, many of the 'ja L2' instructions may end
up being implemented in their long form (if L2 is far away) using:

	jbe	L3	% jump if below or equal to L3
        jmp 	L2
     L3:
	...


For large compilation units L2  may be far away.

fixed problem with missing functor parameter

     ML's empty string "" now points to a legal null terminated C string.

      Machine code generation for CALL (relative displacement mode)

  Fixed jump chaining bug.  Indentation fixes.

implemented entry points; NEW BOOTFILES

      Bug fixes for CMOVcc on x86

implemented generic Controls module and used it for
all compiler flags/tuneable knobs/...;
more command-line options accepted (for setting the above controls);
some minor bug fixes

  Changed representation of probabilities and frequencies.

Tested the weighted block placement optimization on all architectures
(except the hppa) using AMPL to generate the block and edge frequencies.
Changes were required in the machine properties to correctly
categorize trap instructions. There is an MLRISC flag
"weighted-block-placement" that can be used to enable weighted block
placement, but this will be ineffective without block/edge
frequencies (coming soon).

*** empty log message ***

(FSplit.split.stfn): Check "split all" before checking
"split nothing" in case the TFN has nothing but RET[v].

fixed downstream problems caused by those Freq type change in MLRISC

  Simplified the fcond type in ml-tree-basis.  Put in a temporary
  fix for a type error in weighted-block-placement-fn.sml.

  Better error messages.

  Added dumpBlock printing routine.

  Uncommented support for FCMP in mltree-gen.sml (this required making
  the Cells structure a functor parameter).

In order to support the block placement optimization, the first
cluster that is generated (called the linkage cluster) contains a jump
to the entry point for the compilation unit. The linkage cluster
contains only one 'function', so block placement will have no effect on
the linkage cluster itself, but all the other clusters have full
freedom in the manner in which they reorder blocks or functions.

On the x86 the typical linkage code that is generated is:
   ----------------------
	.align 2
   L0:
        addl    $L1-L0, 72(%esp)
        jmp     L0


        .align  2
   L1:
   ----------------------

72(%esp) is the memory location for the stdlink register. This
must contain the address of the CPS function being called. In the
above example, it contains the address of  L0; before
calling L1 (the real entry point for the compilation unit), it
must contain the address for L1, and hence

	addl $L1-L0, 72(%esp)

I have tested this on all architectures except the hppa.The increase
in code size is of course negligible.

   Added #[ ... ] expressions to the mlrisc-tools library

  Added comment.

  Dump CFG in block placement order.

bug fixes

  Removed use of polymorphic equality.

  Switch to using probabilities (instead of ints) for BRANCH_PROB annotation.

  New implementation of probabilities.

  removed extra blank line when printing annotations in assembly code

  Documentation update.

  Fixed bugs and added control flags.

  Added control flags.

  Added dump function to CFG module (this can replace the PrintFlowgraph
  functor).

  Removing unused (and incomplete) file.

fix exit edge labels

  Fixed type error.

  Implemented negateConditional.

  The INTO instruction should be marked as an escaping jump.

made types in C and C_Debug coincide; simplifications

fix problem with CM noweb tool;
reworked FFI: enum types;
              the (non-)distinction between incomplete and complete types

    Backedge method.

    A new CPS spill phase.  Compilation of large structures with over
1000 items can now finish before the heat death of the universe.

*** empty log message ***

My earlier commits to install the cfgView were not tested,
however they broke the various CM files used to build the compiler.
This commit is just to fix the problem and I will do a major cleanup
later. Sorry if you stumbled on this.

  Added jump-chain elimination and weighted block placement.

  Added jump-chain elimination and fixed type errors in block placement.

  Expanded comment.

  Added comments.

added a graphical visualization module for cfg

      Minor fix to mlrisc-tools's parser

    Regenerated these files.  They didn't compile.

  Removed SIDEEXIT edge kind and HYPER block kind.  Added FLOWSTO edge kind.

  New file: implementation of Pettis-Hansen block placement.

  Changed specification of negateConditional.

  Sync with master copy.

   Typo

    Minor bug fixes to the mlrisc-tools library

a mix of changes to CM and FFI

forgot to edit CHANGES file

applied bug fix for PQ implementation

typo fixed

last-minute bug fix for installer script

Working release, version 110.39

added CM.State.showBindings (and EnvRef.listBoundSymbols)

ml-nlffigen: cpif mechanism and iptr repository implemented

added additional phantom type constructor to ml-nlffi-lib.cm(tag.sml)

Fixed the MLRISC bug sent by Markus Wenzel regarding the compilation
of Isabelle on the x86.

From Allen:
-----------
 I've found the problem:

     in ra-core.sml, I use the counter "blocked" to keep track of the
     true number of elements in the freeze queue.  When the counter goes
     to zero, I skip examining the queue.  But I've messed up the
     bookkeeping in combine():

         else ();
         case !ucol of
           PSEUDO => (if !cntv > 0 then
                 (if !cntu > 0 then blocked := !blocked - 1 else ();
                                    ^^^^^^^^^^^^^^^^^^^^^^^
                  moveu := mergeMoveList(!movev, !moveu)
                 )
              else ();

     combine() is called to coalesce two nodes u and v.
     I think I was thinking that if the move counts of u and v are both
     greater than zero then after they are coalesced then one node is
     removed from the freeze queue.  Apparently I was thinking that
     both u and v are of low degree, but that's clearly not necessarily true.


02/12/2002:
    Here's the patch.  HOL now compiles.

    I don't know how this impact on performance (compile
    time or runtime).  This bug caused the RA (especially on the x86)
    to go thru the potential spill phase when there are still nodes on the
    freeze queue.

bug fix for ml-nlffigen (fptr-rtti-)

added prio-q exports to CML version of smlnj-lib.cm

ml-nlffigen updates; minor Basis updates

  Added priority queue implementation.

added -gensym option to ml-nlffigen

minor fix for a problem with GenSML

110.38.1; new bootfiles; new PID generation scheme; installer improvements

  regenerated x86Peephole.sml

  Use isZeroLE in the definition of isZero.

Compilers that generate assembly code may produce  global labels
whose value is resolved at link time. The various peephole optimization
modules did not take this in account.

TODO. The Labels.addrOf function should really return an option
type so that clients are forced to deal with this issue, rather
than an exception being raised.

  small ommission in the last commit

   Pulled out various utility modules that were embedded in the modules
   of the register allocator. I need these modules for other purposes, but
   they are not complete enough to put into a library (just yet).

  A bug fix from Allen.

  A typo causes extra fstp %st(0)'s to be generated at compensation
edges, which might cause stack underflow traps at runtime.  This
occurs in fft where there are extraneous fstps right before the 'into'
trap instruction (in this case they are harmless since none of the
integers overflow.)

Alignment directives were only recognized as being
text segment directives.

  added support for the comment pseudo op

two small bug fixes: sparc c-calls and ml-nlffigen

     MLRISC bug fixes.  MachineInt.fromString was negating its value.

INSTALL instructions added; install script improved

bug fix and minor improvement to sparc-c-calls

c-calls API modification: paramAlloc callback

  Added hooks for parameter-area management by client to CCalls API.
  Currently, these hooks are not being used.

Now For Real(tm):  Release_110_38

110.38 RETRACTED; bug fixes (real64 and FFI)

     The 'nowhere' program generator (which translates programs
     with where clauses into legal SML) is now an installation target.

Release_110_38

more FFI tweaks and improvements

    Minor lexer bug fixed

*** empty log message ***

   There is a dramatic simplification in the interface to the
   register allocator for RISC architectures as a result of making
   parallel copy instructions explicit.

fix for x86 c-call bug

FFI bug fixes and small interface enhancements

re-pointed config/srcarchiveurl

more FFI fiddling

turned x86-fp on

  Fixed label formatting bug.

  More pseudo-op fixes; added first-cut at MacOS X assembler support.

     Fixed a problem with the handling of CALLs in x86's "fast fp" mode.

  Added Allen's fix for the FP register allocator.

  Fixing formatting of pseudo-ops.

  Fixed generated comment syntax to use C syntax (not SML!).

more fiddling with the FFI interface

  When building the control flowgraph, both comments and labels
  are sensitive to the current  segment (TEXT/non-TEXT) being
  built.

bug fix for external/import declarations

Bug fixes:

  1. Emit a .text at the beginning of an assembly listing.

  2. Implement the ASCIIZ pseudo-op using .ascii and String.toCString.

  fixed various bugs with emitting pseudo-ops

1. Since COPY instructions are no longer native to the architecture,
   a generic functor can be used to implement the expandCopies function.

2. Allowed EXPORT and IMPORT pseudo-op declarations to appear inside a
   TEXT segment.

updates/bug fixes to FFI

  Added code for BSS and SPACE pseudo-ops.

  Added copyrights.

  added BSS and SPACE as basic pseudo-ops

updates to pgraph and, more importantly, to NLFFI

version file updated (config dir)

	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.

*** empty log message ***

*** empty log message ***

*** empty log message ***

    Updated peephole modules to match latest MLTREE and instruction
    representation.

 committed missing file

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.

Simplified the algorithm used to reach a fixpoint.
This should have been part of 110.37 but I couldn't
get it to work until Allen pointed out the fix.

1. Made the stuff in MLRISC/Tools compile with the new pseudo ops changes.
2. Packaged the stuff in MLRISC/Tools as a library during installation.

added HTML version of README

Release 110.37 -- this time for real.

 endian related functions

  Fixed a bug on non-x86 architectures related to backpatching
  and the new pseudo-ops.

removed obsolete item from README

added README for 110.37

postponed 110.37 because of serious bug

added forgotten .tex-file

Release 110.37 -- see HISTORY

  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.

  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.

Initial revision

srcname routed to mlriscgen;
more fiddling with pgraph (portable library graph) code

  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:
 	10			687
 	20			438
 	30			198
        40			  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.

updates to CKIT, CM, NLFFI, and scripts (see HISTORY)

     Fixed parsing of <pat> : <ty> and <exp> : <ty>

   Improved pretty printing of sml ast

  Fixed the x86 c call fp bug described in blume-20010920-slowfp

  Synchronizing with master copy of library.

HISTORY entry for netbsd patch

commited netbsd underscore patch from Chris Richards

X86RA now uses a valid (instead of dummy) PrintFlowgraph module.
Changes from Allen.

The representation of a program point never expected to see more
than 65536 instructions in a basic block!

*** empty log message ***

Updated input to PERL scripts used to generate
MLRISC cm files.

  Changed defineLabel to report an error is one tries to define the same
  label twice in the same cluster.

 fix in assembly output converting mov to set

  Fixed CM files so that CKit builds under 110.0.7.

 removed ir.cm from RA.cm

  Commented out mention of ir.cm, which is probably obsolete.

(fcexp.fcFix.fcEta): Fix typo.

made exnName and exnMessage a bit more robust wrt. ml-build

x86-fast-fp turned off again

minor cleanup of pathname handling in CM
as well as argument handling by the driver shell script

Updated input to PERL scripts used to generate
MLRISC cm files.

Installed patch from Allen to track compiler
environment changes that affect MLRISC/Tools.

  cleanup

  cleanup

added README files

(sml-tyvarseq-re): Fix typo.

fixed mistake in config/preloads; retagged

version 110.36

set x86-fast-fp to true by default

  Added show_edge to exported values.

Added Allen's x86-fp.sml fix

Added Allen's CFG fixes

minimal version of structure Compiler

(loopify.lexp.cfun): Don't forget to empty TFS when TCP is not true.

Typo

HISTORY (fix for bug 1581)

fix for bug 1581

fleshed out README file

Release_110_35

 fixed author of changes

Initial revision

renaming ir to ir-archive

missing files

removed clusters from MLRISC

Initial revision

updates to environments, envrefs, interactive loop

fixed a minor slipup

library debugprof.cm extracted

new organization of compiler libraries

  Converted sml/nj files to use the new CellSet

  Moved CellSets from Cells to CellsBasis.

  More CellsBasis changes for files not used by SML/NJ compiler.

committed Zhongs fix

bug fix in spilling imull instructions

*** empty log message ***

    There was a bug where call instructions would mysteriously
    vanish. The call instruction had to be one that returned
    a floating point value.

*** empty log message ***

*** empty log message ***

(sml-rightalign-and): New defcustom.
(sml-tyvarseq-re): New var.
(sml-font-lock-keywords): Use it.
(sml-imenu-create-index): Don't get confused by tyvarseq's.
(sml-mode-variables): Don't set `comment-column'.
(sml-funname-of-and): New function.
(sml-electric-pipe): Use it.
(sml-find-comment-indent): Try to ignore comment-markers in strings.
(sml-calculate-indentation): Handle closing parens specially.
(sml-indent-pipe): Recognize the case where `and' defines a datatype.
(sml-dangling-sym): Make it work if the symbol is an open-paren.
(sml-indent-default): Change the behavior when preceded by `end',
although I'm not quite sure why.
Understand dangling open-parens.
Properly skip *all* subexpressions of lower precedence.
Allow use of sml-symbol-indent to outdent lines starting with , or ;.
(sml-insert-form): Use preceding-char to avoid bug at bobp.

(sml-mode-menu): Add an explicit t for always-active.
(sml-symbol-indent): Add entries for , and ; and turn into defcustom.

I have dramatically simplified the interface for CELLS in MLRISC.

In summary, the cells interface is broken up into three parts:

  1. CellsBasis : CELLS_BASIS

	CellsBasis is a top level structure and common for all
        architectures.  it contains the definitions of basic datatypes
        and utility  functions over these types.

  2. functor Cells() : CELLS

	Cells generates an interface for CELLS that incorporates the
	specific resources on the target architecture, such as the
	presence of special register classes, their number and size,
	and various useful substructures.

  3. <ARCH>CELLS

	e.g. SparcCells: SPARCCELLS

        <ARCH>CELLS usually contains additional bindings for special
	registers  on the architecture, such as:

		val r0 : cell		(* register zero *)
		val y : cell		(* Y register *)
		val psr : cell		(* processor status register *)
		...

        The structure returned by applying the Cells functor is opened
        in this interface.

The main implication of all this is that the datatypes for cells is
split between CellsBasis and CELLS -- a fairly simple change for user
code.

In the old scheme the CELLS interface had a definitional binding of
the form:

	signature CELLS = sig

 	   structure CellsBasis = CellsBasis

	   ...

        end

With all the sharing constraints that goes on in MLRISC, this old
design  quickly leads to errors such as:

	"structure definition spec inside of sharing ... "


and appears to require an unacceptable amount of sharing and where
constraint hackery.

I think this error message (the interaction of definitional specs and
sharing) requires more explanation on our web page.

Substantial simplification in the CELLS interface

(sml-proc-buffer): Save excursion when calling run-sml.

*** empty log message ***

(sml-mode-menu): Fix activation of `start compiler'.

(sml-syntax-prec): Split ; and , from `in' and `with'.

Put the entry in `Emacs' rather than `Editors'.

(BuildArch): Simplify call to `install-info'.

(sml-mode-menu): Add an explicit t for always-active.

Add more stuff.  It might help for Emacs-19.34.

(test): Use elisp files in current dir.

extensive organizational changes

Some new cases from Makholm Henning.

minor tweaks to 110.34 (and re-tagged)

working version 110.34

handling of varargs made more robust (and different)

made ml-nlffigen handle varargs functions with a bit more grace;
some minor internal improvements to CM

bug in CM fixed (related to compile servers)

typo fixed in CM manual

fixed nasty X86 assembly bug

HISTORY

bug fixes for ml-nlffigen

eXene compiles and works again

CML compiles and works again

un-break fix for bug 1432