HT kernel

[Kris Kennaway]

On Wed, Oct 27, 2004 at 02:17:48AM +0200, Alex de Kruijff wrote:
> On Tue, Oct 26, 2004 at 02:01:30PM -0700, Kris Kennaway wrote:
> > On Tue, Oct 26, 2004 at 09:54:45PM +0100, Alexandre Vieira wrote:
> > > Hello,
> > > 
> > >   I have a machine with an Intel p4 3.2ghz FSB800 w/ 1MB L2 cache and
> > > I wanted to know your opinion about some kernel options that would
> > > boost the performance of this kind of processor.
> > 
> > Note that for a lot of workloads HT decreases performance.
> 
> In what way? Does HT/SMP kernel or option do worse then a normal kernel
> or default options?

Depends on the workload.  Remember that hyperthreading isn't "a secret
extra CPU hiding inside the same chip"; if you try and execute an
instruction on the second virtual CPU that cannot be executed on the
silicon because the first virtual CPU is using that part of the
silicon (or other reasons), the virtual CPU will block.  If you're
spending most of your time blocked in this way because of the
computational workload you're running, performance is going to be
worse than the !HT case due to OS overheads.

Kris
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 187 bytes
Desc: not available
Url : http://lists.freebsd.org/pipermail/freebsd-questions/attachments/20041026/1ac27c8c/attachment.bin