You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I am not fully sure if it is exposed in cupla but I think so. If it is exposed than it is simple to add to picongpu.
Never the less thr performance of the thread backend is really bad therefore it is not exposed.
I added a section in out Wiki how OpenMP can be traced with score-p.
I can have a look into it next week.
@psychocoderHPC yes, I thought so. We probably would need a threadBlocks implementation in alpaka to make this backend anywhere near to performant. (And even then, the backends with preemptive multithreading such as omp2b and boost.fiber will be way more efficient.)
Anyway, the idea here is just to check if our implementation of C++(11?) threads can be captured by Score-P.
I exposed the c++ threads in #2701 to the PIConGPU user.
Set export PIC_BACKEND=threads or run pic-configure -b threads to select the C++11 thread backend. WARNING: Your program will run really slow with this backend.
I like to use the C++ thread backend of Alpaka. Right now I can only choose
cuda
,omp2b
,serial
,tbb
inpic-configure
.The text was updated successfully, but these errors were encountered: