| Re: [tracecompass-dev] Time Graph view stuck |
|
Hi Matthew, This was spot on! We are in far better shape now. I see some more scope for
optimization, shall keep in touch. Thanks, Vinod From: Matthew Khouzam <matthew.khouzam@xxxxxxxxxxxx>
Yes, if it's slowly progressing it means the analysis is slow. Are you running a regex per event? I suggest either profiling your software with visualvm or some equivalent to figure out the bottleneck... or even better, you can trace it!
🙂 Check out the TraceCompassLogUtils class for examples. How big is your trace? are we talking 100kb, or 100 tb? this can have an impact too. When you figure out how many events/second you have you can know if something is going wrong. Quite frankly, making things go fast in trace compass is one of my favorite parts of my job. Some leads: Is your event parsing slow? Do you have an O(N2) datastructure somewhere? Are you reading and writing to the hdd per event? Are you doing network operations per event? Hope this helps. Matthew. From: Vinod Appu <vinod.appu@xxxxxxxxxxx> Hi Matthew, No, I don’t see any error log. The analysis is still in progress after hours, I see. I can see new entries being added to the views including state system explorer. What could be potential reasons? ~Vinod From: Matthew Khouzam <matthew.khouzam@xxxxxxxxxxxx>
Hi all, check your error log, i bet you got an exception like a null pointer and the analysis froze. From: tracecompass-dev <tracecompass-dev-bounces@xxxxxxxxxxx>
on behalf of Vinod Appu via tracecompass-dev <tracecompass-dev@xxxxxxxxxxx> Hi, I’ve a view implementing BaseDataProviderTimeGraphView. When I tried to load ~ 50 Million events, it took more than 1hr time. Is this expected? The progress got stuck as below. Is there a benchmark done for the view? Is it possible to remove other analysis like event total/ event per type..? Does it going to give me an advantage?
~Vinod |
