This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [RFC PATCH, vectorizer]: Vectorize int -> double conversions
- From: Dorit Nuzman <DORIT at il dot ibm dot com>
- To: "Daniel Berlin" <dberlin at dberlin dot org>
- Cc: gcc-patches <gcc-patches at gcc dot gnu dot org>, "Uros Bizjak" <ubizjak at gmail dot com>
- Date: Tue, 24 Apr 2007 00:30:48 +0300
- Subject: Re: [RFC PATCH, vectorizer]: Vectorize int -> double conversions
...
> However, the
> > vectorizer doesn't know how to use these insns yet, so I don't know if
we
> > want to introduce something more general than we can currently
generate.
>
> The problem is that it takes a *LOT* more work to remove tree codes
> than add them. Thus, if we plan to support something like unpacking
> where size != vec_size/2, then we should make the operands work for it
> now, instead of having to add the same tree code later on *anyway*,
> and then clean up all the existing uses.
Maybe. Seems like a pretty mechanical work to remove tree-codes, I don't
think it's such a big deal? (though probably not so much fun). Anyhow, I
don't have a strong preference here. I guess I'm slightly more for
generalizing incrementally as needed rather than throwing in something more
general than we can generate, but it certainly is a good idea to consider
the whole picture and think about what we may need in the future up front.
dorit
> >
> > Maybe experts of specific targets can comment as well.
> >
> > dorit
> >
> >
> > > >
> > > >
> >
> >