Yes, smp is quite slow, but it does point out one flaw somewhere, I
haven't tracked it down to qemu or OpenBSD, but OpenBSD stops detecting
cpus at cpu19 (cpu0-cpu19 = 20 cpus) even though X86_MAXPROCS is 32 ..
Yes, the i386 -smp does not boot bsd.mp from i386, also something busted
I suppose, but I do believe this is perhaps the way isa devices are
simulated in qemu ...
I don't see any pthread functions in qemu, yet it links against
pthreads. So I wonder, is this necessary? I've tried qemu against
rthreads myself, but so long as it never inits any new threads, I've
come to realize it's hardly a useful test.. ;-)
Can Erkin Acar wrote:
> On Tue, Dec 20, 2005 at 05:30:50PM -0600, Todd T. Fries wrote:
>>Please test, verified to build on i386 and amd64 thusfar.
>>See attachment or:
>> http://todd.fries.net/ports/qemu-0.8.0.tar.gz >
> tested on an amd64 host for:
> Windows 98 on i386, using an existing installation. -- ok
> network installation from a recent snapshot for the following:
> OpenBSD 3.8/i386 -current bsd, Xorg -- ok
> OpenBSD 3.8/i386 -current bsd.mp, on dual processors (-smp 2) -- panic
> apic_intr_establish: intr 12 can't share level-triggered with edge triggered
> OpenBSD 3.8/amd64 -current bsd, Xorg -- ok
> OpenBSD 3.8/amd64 -current bsd.mp, on dual processors (-smp 2), Xorg -- ok
> The dual processor 'emulation' is quite slow.
> Have not yet tested userland networking or the new vlan stuff.
> btw. I have been using 'rthreads' during the tests.