On 24.8.2012 1:49, Adam Hraska wrote:
>> If I am not mistaken, all of those were in fact QEMU runs, right? We
>> need to test this thoroughly on bare metal as well.
> 
> Yep, they were run in qemu. In qemu with hardware virtualization.
> I would hope it does a reasonable job of mimicking running the
> code on bare metal -- CHT is not kernel-specific code and it
> just makes use of ordering memory accesses and having atomic
> instructions.

I guess the problem here is that it is not clear whether -smp 4 under
QEMU/KVM is equivalent to a 4-core bare metal process as far as
parallelism is concerned, i.e. whether guest instructions are indeed
executed in parallel on the host or just interleaved on one CPU. In the
latter case, testing would give you a somewhat weaker confidence of the
kernel stability. Testing on bare metal is therefore essential in this case.

> Agreed. I will eventually add them, but that is not to be considered
> during the final evaluation.

Right.

Jakub

_______________________________________________
HelenOS-devel mailing list
[email protected]
http://lists.modry.cz/cgi-bin/listinfo/helenos-devel

Reply via email to