Yes. I did enable SMT (Hyperthreading) scheduler support. Here is what I have for Processor Family:
"Pentium-4/Celeron(P4-based)/Pentium-4 M/Xeon)" Here is what dmesg reports: CPU0: Intel(R) Pentium(R) 4 CPU 3.40GHz stepping 09 CPU1: Intel(R) Pentium(R) 4 CPU 3.40GHz stepping 09 Are my settings correct? By the way my cpu also has 512K L2 cache. On Apr 12, 2005 2:08 PM, Richard Fish <[EMAIL PROTECTED]> wrote: > Zander Z365 wrote: > > >Thanks to all of you for helping me. I can successfully emerge X & > >KDE using an SMP kernel and hard setting the CPU frequency. However, > >using an SMP kernel 'cat /proc/cpuinfo' nor 'x86info -mhz' seem to > >show the new cpu frequency. I do have two more questions: > > > > > > Strange...my P4 3Ghz HT reports this fine in /proc/cpuinfo. What does > your dmesg output say about your processor(s)? You should have > something similar to: > > Apr 11 08:20:55 carcharias CPU0: Intel(R) Pentium(R) 4 CPU 3.00GHz > stepping 09 > Apr 11 08:20:55 carcharias CPU1: Intel(R) Pentium(R) 4 CPU 3.00GHz > stepping 09 > > Are you sure you selected the right processor family? > Also, did you enable the SMT (hyperthreading) scheduler in the kernel? > > >1. Using a Uni-processor kernel it appears I do not have to decrease > >my cpu frequency to emerge large products or compiles. Will I loose a > >great deal of performance using it since I really only have one cpu > >(even though It has HT technology)? > > > > > Well, 99% of the time, it won't make any difference. But, yes, I can > notice it on my laptop. I would say the hit is about 20%. Of course, > my CPU only has 512K of L2 cache, compared to your, what was it, 2M?, so > it might not make as big a difference for you. > > The reason is because HT really only helps when you have cache-misses, > and data must be fetched from RAM. This takes several clock cycles to > complete, during which time nothing else can run in the execution unit > of the processor. HT adds another instruction pipeline, so as long as > at least one of the two pipelines can make progress, the execution unit > is in use. So fewer cache misses for you should mean less of an impact. > > >2. Will 'speedfreq' work with SMP kernels? > > > > > > Yes. > > -Richard > > -- > gentoo-user@gentoo.org mailing list > > -- gentoo-user@gentoo.org mailing list
