[Slackbuilds-users] sse2 optimisation

korgman korgie at gmail.com
Sat Jun 26 14:45:49 UTC 2010


David Spencer wrote:
> Greetings SlackPeople, may I ask your advice?
> 
> I'm doing a SlackBuild for Luminance HDR.  Its compiler flags assume
> that SSE2 instructions are available.  That's a bad assumption, but
> it's easy to make it configurable with a variable SSE2=yes|no and a
> note in the README.
> 
> So, here's my question.  Should the SlackBuild's default be SSE2=yes,
> or SSE2=no?  There are three obvious choices.
> 
> (1) The default should match upstream's assumption => SSE2=yes
> 
> (2) The default should work on any processor => SSE2=no
> 
> (3) Determine the default from the ARCH => if x86_64 then SSE2=yes else SSE2=no
> 
> Any thoughts?  I'm guessing everyone hates (3) :-)

I prefer (3) but I think it's not preferable by the admins since there is only:
i486
i686
x86_64

on the slackbuild template.

I think the i786 also have SSE2 commands but it is absent in the Slackbuild template.
http://en.wikipedia.org/wiki/I786
http://en.wikipedia.org/wiki/SSE2

Can this software work with acceptable speed in old processors without SSE2? (i686)
-- 
Michales Michaloudes
Thessaloniki


More information about the SlackBuilds-users mailing list