Skip to main content

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index] [List Home]
Re: [tracecompass-dev] Per-CPU filtering the CPU Usage view

The advantage of signals is that it's easier to make dependencies. It's
also better if you have several sources. If it's a one on one, it's much
less interesting.

So we already have a listener for threads in the critical path I think.
If we need a second, this should become a signal, no? Especially if we
can later right click in the resource view and "follow cpu/follow thread"

I like it a lot. I'm thinking also a smart legend that would use the
multi-core visualizer of cdt. This would have the advantage of showing
process snaking.



On 16-01-29 01:11 PM, Bernd Hufmann wrote:
> Hello
>
> about the listener interface, I would suggest to create TmfSignals
> that are sent when a Resource (CPU), Thread is selected etc. Then
> other views just can listen to these signals and act upon reception.
> Separate listener interfaces are not necessarily needed.
>
> /Bernd
>
>
> On 01/29/2016 12:56 PM, Geneviève Bastien wrote:
>> Hi Matthew,
>>
>> I've been thinking on a potentially similar problem for threads. Cedric
>> Biancheri here at the lab implemented a cool feature with a Resources
>> view-like view where it highlights the information related to a given
>> thread. But the thread selection mechanism is a bit hackish.
>>
>> I was thinking that we could have an application-wide "Thread
>> selection". There would be providers (for instance, the Control Flow
>> View would provide thread selection when threads are selected) and
>> listeners who would get notifications when the selection changes (for
>> instance the Critical Path View. Also the Resources View could have a
>> 'Follow Thread Selection' mode for Cedric's highlights).
>>
>> My suggestion to you would this be to change 'Threads' for 'CPU' in what
>> I just wrote and have the Resources View as provider and CPU Usage view
>> as listener.
>>
>> But that is just a proposition. I would eventually have written about it
>> to ask advice myself, so for what it's worth, it's out now!
>>
>> Maybe the UX people have some comments on this?
>>
>> Geneviève
>>
>>
>>
>>
>> On 01/29/2016 09:41 AM, Matthew Khouzam wrote:
>>> Hi all,
>>>
>>> Sorry for the redundant title. I was musing how to filter the cpu usage
>>> view on a per-cpu usage. It seems trivial at first. With 4 cores. A
>>> simple toolbar would suffice. However, this is a problem with many-core
>>> systems, like the Xeon-Phi. Those beasts have over 200 cores, and this
>>> is just going to get bigger. So do you have any suggestions for how to
>>> properly filter these views?
>>>
>>> I'm leaning towards a control flow view style filter view.
>>>
>>> Thoughts?
>>>
>>> Matthew
>>> _______________________________________________
>>> tracecompass-dev mailing list
>>> tracecompass-dev@xxxxxxxxxxx
>>> To change your delivery options, retrieve your password, or
>>> unsubscribe from this list, visit
>>> https://dev.eclipse.org/mailman/listinfo/tracecompass-dev
>> _______________________________________________
>> tracecompass-dev mailing list
>> tracecompass-dev@xxxxxxxxxxx
>> To change your delivery options, retrieve your password, or
>> unsubscribe from this list, visit
>> https://dev.eclipse.org/mailman/listinfo/tracecompass-dev
>
>
> _______________________________________________
> tracecompass-dev mailing list
> tracecompass-dev@xxxxxxxxxxx
> To change your delivery options, retrieve your password, or
> unsubscribe from this list, visit
> https://dev.eclipse.org/mailman/listinfo/tracecompass-dev



Back to the top