Hello Kent,

  Just my 2c: was that a single-thread compression?
  Low bandwidth may be the culprit, but...

  We've tried parallel implementations of common compression
algorithms -  "pigz" and "pbzip2", they do yield a much faster
performance on our T2000s and even on 2-4 CPU machines - if
you feed them the I/Os quickly enough. Overall CPU time often
is larger by 1.5-2 times, but the wallclock time drops a lot.

  Perhaps it may be reasonable to (dynamically) link SRSS with
such versions of libraries on multicore servers?

  Desktop framebuffers may be reasonably large to benefit...

KP> It's not so much the FP, but just the fact that the single thread
KP> performance is slower than previous single-core SPARC chips. I've 
KP> measured compression performance, which doesn't use FP at all, running
KP> over low bandwidth, and a T2000 takes about 55% longer to do an 
KP> equivalent amount of compression as my V280R. It feels commensurately 
KP> sluggish in comparison.

>> Even more interesting (I think) is a performance review of the new 
>> UltraSPARC T2 range..... anyone?

KP> We're looking into doing that.

KP> Kent


KP> _______________________________________________
KP> SunRay-Users mailing list
KP> [email protected]
KP> http://www.filibeto.org/mailman/listinfo/sunray-users



-- 
Best regards,
 Jim Klimov                            mailto:[EMAIL PROTECTED]

_______________________________________________
SunRay-Users mailing list
[email protected]
http://www.filibeto.org/mailman/listinfo/sunray-users

Reply via email to