Aaron Boxer
Aaron Boxer
ahhh I see. So, you can repro, great.
> Yes, tf::Taskflow::for_each is a drop-in replacement - by default, it uses the guided scheduling algorithm similar to OpenMP. What does each iteration look like in your case? The source...
sorry about that, I set it up for Wayland, as I run Fedora. If you pull now, it will work on X11. You may have to delete the cmake cache...
I think this will fix that (assume you're on ubuntu) `apt-get install xorg-dev libglu1-mesa-dev`
Great! try this - no viewer will get launched `./OptCuts_bin 100 ../input/bimba_i_f10000.obj `
great, thanks again for looking into this. The program does run for quite a while. `TBB BENCHMARK: Symmetric Dirichlet Energy : average duration 0.008541 ` is the important line -...
> OK - let me wait for a complete run to finish and dig into it a bit more. Could you please let me know what each step of the...
Very interesting, I will do some more investigation - will keep you posted.
no, I don't have any recursive tasking. I only introduce taskflow on a single loop - there is no recursion, and it's a simple loop.
here's an interesting data point: I removed all code from the loop, and it still takes 0.04 ms on average. TBB takes 0.001 ms on average.