search

soctrace-inria / ocelotl

Visualization module with aggregation features for Framesoc infrastructure
http://soctrace-inria.github.io/ocelotl
4 stars 0 forks source link

Canceling "RUN" not working properly #133

Open stanisic opened 9 years ago

stanisic commented 9 years ago

With a 100MB trace, when I increase "Timeslice Number" to 100 and launch analysis, computing dichotomy takes a lot of time (which makes sense). However if I then try to cancel the computation (clicking on "Cancel" button), the execution does not stop. After waiting for several minutes and no changes, I was obliged to stop it by killing java process and whole eclipse with it.

dosimont commented 9 years ago

HI Luka. There is indeed some issues with canceling. It works properly when it occurs during the trace reading and the microscopic model generation. However, the aggregation is performed by an extern C++ library that does not handle canceling ; if you try to cancel during this step, it will continue to compute until Eclipse takes the hand again. I strongly recommand to limit the spatiotemporal analysis to 30 time slices, since the algorithm complexity is cubic on the time dimension: I estimate that with 100 time slices, the aggregation computation would need at least 35 times the time duration of a 30 time slice analysis.