[MAGEEC] Energy cost of context switching

[Jeremy Bennett]

On 19/11/13 18:51, Kerstin Eder wrote:
> Hi Jeremy,
> 
> This would make an interesting student project - investigate polling vs
> interrupts for different platforms / loads.  ;)
> It could start end Jan and we would have a comprehensive set of results
> by Sept.

This would be a good idea.

I did a bit more digging and now recall where I got this idea from. It
was Michael J Pont of TTE systems, who specialize in time-triggered
approaches to hard real-time systems. The real benefit of time-triggered
models is that they give you predictability, but I remember him pointing
out that for some latency/throughput requirements time-triggered polling
could beat interrupts hands-down.

> Some indicative results would help at this stage to motivate the
> project, but clearly there are different opinions on this so it must be
> tested out.

Perhaps James could garner some indicative data with only modest effort?


Jeremy

-- 
Tel:      +44 (1590) 610184
Cell:     +44 (7970) 676050
SkypeID: jeremybennett
Twitter: @jeremypbennett
Email:   jeremy.bennett at embecosm.com
Web:     www.embecosm.com