This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [RFC] try to generate FP and/or/xor instructions for SSE
- From: tbp <tbptbp at gmail dot com>
- To: "Paolo Bonzini" <bonzini at gnu dot org>
- Cc: "Ross Ridge" <rridge at csclub dot uwaterloo dot ca>, "GCC Patches" <gcc-patches at gcc dot gnu dot org>
- Date: Thu, 23 Aug 2007 16:43:55 +0200
- Subject: Re: [RFC] try to generate FP and/or/xor instructions for SSE
- References: <20070823013503.426FE74B95@caffeine.csclub.uwaterloo.ca> <46CD273E.6060009@gnu.org> <46CD705E.90701@gmail.com> <46CD72E0.3000005@lu.unisi.ch> <46CD7ACA.8050904@gmail.com> <46CD8E69.4060402@gnu.org> <4fc48eb10708230709v1c82d957y5b3101445d56fc00@mail.gmail.com> <46CD9AAF.7000502@gnu.org>
On 8/23/07, Paolo Bonzini <bonzini@gnu.org> wrote:
> Well, they return a mask and an all-ones floating point value does not
> make much sense. Do you get pessimizations out of that? With a patch
> like this one, though, you should not, so I could try it on a small test
> case if you sent it.
I've never seen any pessimizations because of that. But even if those
masks do not make any sense as fp values, they are meant for later
consumption by ps|pd ops without penalty, so they are for all intent &
purpose ps|pd vectors and pretending they are v4si is a
misrepresentation (needing a cast back to proper form).