Using the tool "Shark" that comes with the OS X Developer Tools, you can
track where SETI@HOME spends most of it's time and from my short test of 30
seconds it was spending most of it's time in the cos and sin functions.

It looks as though they are using cos/sin tables in the latest nightly code
but I don't know for sure if they are
doing this in the public release version or not.

The tables are periodically recalculated so perhaps I was just profiling at
the time it was recalculating the table.

There are supposed to be some special Altivec cos & sin functions that speed
those up so perhaps that would help if the code was optimised using some
Altivec rutines.

I was going to take a look at that soon as some people are already
optimising the x86 version to use SSE.

The compiler they used for the OS X version is gcc which isn't supposed to
be the best compiler out there for speed.  gcc is really all about cross
platform compilation and not speed.

There should be some speed improvments to be made with Altivec though if
someone does it.

Devon.
"Lazlo" <lmobrien@excite.com> wrote in message
news:9d8533a9.0407280606.1a353387@posting.google.com...

I'm running two dual G4's (1GHz tower with 2Gb of RAM each) and it
takes about 8hrs to process one work unit on either of them. WTF? Both
are using OS X 10.3.4 (Panther) and the Seti software appears to be
using both processors (two processors per Mac) yet I'm getting less
performance than slower Windows PC's (according to various posts in
this forum).

There is no appreciable speed up at night when the Macs are not using
any other software. Does anyone know why the Mac version (3.0.8a) of
the SetiatHome would run slower than my 800MHz single processor WINXP
PC? No other software package (like processor pigs Photoshop or Maya)
runs faster on my PC than on my Macs, so again, what gives?