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
Octo-Tiger currently does not work when using the current HPX master. We get a segfault, usually at the beginning of the application (for instance when loading of options or during the first timestep -- it varies). Note that this does not always happen and some few runs execute normally.
This does not seem to be a racing condition as I am able to reproduce this with --hpx:threads=1.
I did some digging when this problem started to occur. HPX 1.9.1 still works fine, of course! The problem seems to start occuring after #6050 was merged: I can reproduce the problem with 962288c , while the commit just before that ( 8edd9b3 ) still works without any issues.
To reproduce:
I used a basic Octo-Tiger (master) build without CUDA / Kokkos / Networking on my laptop
Octo-Tiger currently does not work when using the current HPX master. We get a segfault, usually at the beginning of the application (for instance when loading of options or during the first timestep -- it varies). Note that this does not always happen and some few runs execute normally.
This does not seem to be a racing condition as I am able to reproduce this with --hpx:threads=1.
I did some digging when this problem started to occur. HPX 1.9.1 still works fine, of course! The problem seems to start occuring after #6050 was merged: I can reproduce the problem with 962288c , while the commit just before that ( 8edd9b3 ) still works without any issues.
To reproduce:
./build/octotiger/build/octotiger --hpx:threads=1 --config_file=src/octotiger/test_problems/rotating_star/rotating_star.ini --unigrid=0 --cuda_number_gpus=1 --max_kernels_fused=1 --stop_step=10 --max_gpu_executor_queue_length=5 --max_level=3 --correct_am_hydro=0 --monopole_host_kernel_type=LEGACY --multipole_host_kernel_type=LEGACY --theta=0.5 --monopole_device_kernel_type=OFF --multipole_device_kernel_type=OFF --hydro_device_kernel_type=OFF --hydro_host_kernel_type=LEGACY --amr_boundary_kernel_type=AMR_OPTIMIZED --disable_output=1
The backtrace of the segfault is:
@hkaiser: Do you have any idea what could cause this?
The text was updated successfully, but these errors were encountered: