Re: [Sbcl-devel] compiler slowdown

[William H. N. <wil...@ai...>]

On Tue, Oct 01, 2002 at 03:02:30PM +0100, Christophe Rhodes wrote:
> So would I be right in thinking that people are a lot more worried about
> the efficiency of SUBSEQ and COPY-SEQ than they are of COERCE?  This
> would be my intuition, based loosely on what I feel is reasonable about
> the knowledge of types; I don't think that efficient SUBSEQ, for
> instance, should depend on giving a precise type declaration about the
> vector element-type.

I'm a fairly heavy user of the idiom of accumulating something in an
adjustable vector, then COERCEing the adjustable vector to
SIMPLE-VECTOR or SIMPLE-STRING, and I've always thought it's
reasonable to expect that to be (as) efficient (as anything that
conses is, anyway). But I don't think that should be a strong
constraint on your general solution, since I expect it'd be easy to
handle things like that as a trivial special case layer on top of any
general solution.

-- 
William Harold Newman <wil...@ai...>
"Stay away from my house, you freak!"
  -- <http://www.cryptonomicon.com/beginning.html>
PGP key fingerprint 85 CE 1C BA 79 8D 51 8C  B9 25 FB EE E0 C3 E5 7C