Hello,
I am currently rendering a 1080P mandelbulb animation with a small computer
cluster of AMD 1090T, AMD 965 and a few Intel Core 2 Duo processors. The
production rate is about 25 frames per day total and I would like to find a way
to decrease render times.
What kind of gains would I get if I used the Intel Compiler to create an SSE 4.X
binary specific for each of my processors?
Thanks!
P.S. Of course, I would appreciate any Windows binaries you could send my way.
The work is being to submit to YouTube and I have no commercial interests in
this project.
From: Christian Froeschlin
Subject: Re: Advantages Of Optimized SSE 4.X compile?
Date: 20 Oct 2010 18:19:09
Message: <4cbf6add@news.povray.org>
rgathright wrote:
> What kind of gains would I get if I used the Intel Compiler to create an SSE 4.X> binary specific for each of my processors?
I have no personal experience but was curious enough to read up
a bit on SSE4 now. On one extreme, a DivX encoding algorithm which
was specially optimized with SSE4 using special instructions dedicated
for just that purpose gave a speed increase of about 40%. On the other
hand, a povray compile of 3.6 in 2007 using Intel compiler 10 was
reported to be 16% slower than an SSE3 version Of course, the
SSE4 support of the compiler should be better now. But I don't
think you'll be getting more than a few percent speedup.
I assume you are already using the 3.7 beta SSE2 build?
"rgathright" <rga### [at] gmailcom> wrote:
> Christian Froeschlin <chr### [at] chrfrde> wrote:> > I assume you are already using the 3.7 beta SSE2 build?>> Yes, I am on the latest SSE2 build.
As a followup, I have had success using a 64 bit build on a Windows Server 2003
64bit box. I reduced tracing times from approximately 1 hour 30 minutes to 1
hour 05 minutes.
de> wrote: