| Newsgroups |
gmane.comp.gnu.gift.general |
| Message-ID |
<[email protected]> |
On Wed, May 03, 2006 at 05:49:11PM +0200, Wolfgang Müller wrote:
> On Wednesday 03 May 2006 17:41, David Squire wrote:
> > In any case, I would prefer to see a set-up where it is easy to select
> > from a variety of feature extractors,
>
> Well, who would not prefer this. Currently we have someone contributing code
> and kicking us and we even have trouble finding the time to look at the
> contribution. Do you think it looks serious if we make bigger plans at the
> current situation? This might sound harsh, but I am just stating facts.
>
> > rather than have this clobber the
> > legacy Viper code (esp. since IIRC there are cases when this is *not*
> > faster).
>
> As I understood, this was an assumption, and the Julia's most recent mail said
> that it was not the case. Correct, Julia?
as stated in another email, yep. my assumption, my fault. did everything i
could to SSE2 optimize the thing, but failed. instead, it appears i'm
cache/pipeline optimizing. as the code runs the same on a compiler
that dosent have the sse2 tricks.
i still might be wrong. however. thats why we need testing.
eyeballs that arent mine. programmers can be nearsighted. ;)
Julia Longtin <[email protected]>