On 21 June 2012 10:39, Andy Doan
<andy.doan@linaro.org> wrote:
On 06/20/2012 02:50 PM, Paul Larson wrote:
If I had to guess, I'd say that the network dependency issue ARM is
talking about is the fact that we install an image over the wire, and
that they probably want to do bare metal testing on systems that may not
even have network support at all.
I suspect it could be a bit of this and a bit of what Zach is saying. Benchmark guys work very hard to make sure the system has as little unpredictable disruptions as possible. Having an active network service doesn't help their cause.
Yes its kind of both. We would want :
1.Avoid network connectivity in few cases which will help in reducing the variance. (which may not be possible in LAVA)
2.Have Wifi connectivity on the device since few third party benchmarking apps won't run without wifi connectivity.