Maarten de Boer wrote:
>>Hmm.. Does anybody have more information about this? Could I disable these
>>optimizations compiling for another CPU (386, or maybe even Pentium Classic?)
>>I suppose the performance penalty that would give would be acceptable. Latency
>>is more important.
>>
>
> Okay, I just compiled my kernel for 386 and the latency test results are a lot
> better, but still there are some nasty spikes:
>
> http://193.145.55.36/latencytest3
>
> Maarten
>
Not sure if this was mentioned yet or not, but did you go through the
steps of optimizing your hard disk? If its an IDE drive you'll need to
use the hdparm program. Make sure DMA is enabled. It looks like from
your graphs that you are still using just a 256 byte file for the disk
tests. Try putting 256000000 instead for the last parameter in ./do_tests.
--
Josh Green
Smurf Sound Font Editor (http://smurf.sourceforge.net)
