On Wed, Jan 19, 2005 at 08:31:21PM -0600, Chris Mumford wrote:
I still have to add up the columns 56.1% + 38.2% + 4.1% + 2.1% = 100.5%
which makes sense since the best synthetic hyperthreading improvement was
7%.

The improvement with HyperThreading is not a fixed amount. It depends on how much each thread being run is stalled (waiting for memory) vs. getting useful work done, and also on the mix of instructions being executed. Myth seems to get a big boost from HyperThreading when decoding MPEG2; some well-tuned software I use at work actually runs *worse* with it on.

-Doug

Attachment: signature.asc
Description: OpenPGP digital signature

_______________________________________________
mythtv-dev mailing list
[email protected]
http://mythtv.org/cgi-bin/mailman/listinfo/mythtv-dev

Reply via email to