[haiku-commits] Re: r34888 - haiku/trunk/src/system/kernel/vm

  • From: Ingo Weinhold <ingo_weinhold@xxxxxx>
  • To: haiku-commits@xxxxxxxxxxxxx
  • Date: Tue, 05 Jan 2010 16:11:56 +0100

On 2010-01-05 at 15:10:33 [+0100], Rene Gollent <anevilyak@xxxxxxxxx> wrote:
> On Tue, Jan 5, 2010 at 7:22 AM, Ingo Weinhold <ingo_weinhold@xxxxxx> wrote:
> > That still sounds quite slow. I can't really place the CPU 
> > performancewise,
> > though. Do you have any numbers for our competitors (Linux, FreeBSD)?
> I can get you FreeBSD numbers tonight, I'm fairly certain it's in the
> 20-30 minute range though (that system is 4 years old at this point
> and doesn't exactly have the world's best disk subsystem).

No need to. I was just wondering whether something was seriously wrong, but 
factor 2 for -j2 is still in the expected range, I suppose. This was 
KDEBUG=0 installation?

> >PS: In case someone is looking for a (small) challenge: We still don't use
> >the optimized syscall instructions for AMD CPUs.
> AMDs can't use sysenter/sysexit?

At least older models don't support those. No idea whether they've added it 
in newer ones. Though I suspect they wouldn't have done that, since they 
have their own instructions (syscall/sysret), which are quite similar. You 
should find a "initial_commpage_syscall():..." line in the syslog, telling 
you what mechanism the kernel has decided to use.

CU, Ingo

Other related posts: