gcl-devel
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[Gcl-devel] ANSI


From: Camm Maguire
Subject: [Gcl-devel] ANSI
Date: 29 Jun 2005 00:59:47 -0400
User-agent: Gnus/5.09 (Gnus v5.9.0) Emacs/21.2

Greetings!  Made a modest checkin to version HEAD this evening.
Changelog is included below.  A few brief thoughts here as it is late
-- more later.

1) The ansi bug reports are invaluable -- thank you!  We will get to
   them eventually.  As the work is quite extensive, prioritization is
   of even more help.
2) Paul, if *compile-verbose* defaults to t, must the output of compile-file 
have
   semi-colons? 
3) compile-file ansi changes should be fixed.  Items like this are a
   little tricky, as we don't have conditions in the cltl1 image where
   compile-file must be too.  lisp to the rescue, setf
   (symbol-function ...) to a wrapper in init_ansi_gcl.lsp.in
4) first pass at a checkin of the gabriel benchmarks.  One or two show
   marked improvement over 2.6.6, but all suffer from the large number
   of iterations needed.
5) compiler improvements to inline list search functions replacing eql
   function calls to eq where possible.  More here later.  Now member,
   assoc, rassoc, and si::memq.  Also better array inlining via
   saving and reading variable type information after variable
   structures have been replaced by cvar indices.

Take care,
=============================================================================
  * Style-warning condition
  * Save variable type information in *c-vars*, using alternate form
    (cons index type) for stack vars in inline blocks, for use in late
    inlining
  * expand result-type-from-args invocation; run compiler macros before
    c1 funs
  * Better list search function optimizations -- FIXME more here later
  * Don't  autodeclare apparently constant special variable bindings
  * compile-file ansi test fixes
  * replace eql with eq in list search functions when possible
  * preliminary support for symbol keyword and cons type optimization in
    compiler
  * Preparation for error system overhaul work
  * deftype before structure-subtype in typep
  * unused vars
  * remove ':
  * fmakunbound on undefined fn is not an error
  * heuristic to detect compiled objects from text files independent of
    filename extension
  * truename is not destructive
  * wrap compile-file  in ansi image for warnings and failures
  * Remove unused ansi_cl.lisp
  * Preliminary gabriel benchmark addition
=============================================================================


-- 
Camm Maguire                                            address@hidden
==========================================================================
"The earth is but one country, and mankind its citizens."  --  Baha'u'llah




reply via email to

[Prev in Thread] Current Thread [Next in Thread]