Geoffroy Lesur
Geoffroy Lesur
The zero tolerance for this test is by design, and it passes the CI on the tested architecture (admittedly disabling some optimisations for nvcc). Which configuration did you test?
This error is most probably because LLVM enable fmad instruction on Apple Mx processors, hence it breaks the precision test. These tests should pass with absolute precision when fmad instructions...
A long-standing issue...! I think the best solution is actually the first one, as it allows us to follow the inevitable evolution of python and sphinx packages. But I'm not...
My point is that readthedocs might just break silently, since the build success (or not) of the doc by readthedocs is not a CI. We can give it a go...
Yes, well, there is no "perfect" boundary conditions for VSI. The one in the test suite imposes a solid body rotation, so that there is no free energy to tap...
We have discussed the issue of boundary conditions for gas+dust mixture during the last idefix days. As of now, the boundary conditions are common to every fluid, so if it's...
can't reproduce the issue: tested with cuda 12.2, and g++10, g++11, g++12 and g++15. Can you be more specific as to which compiler/cuda version/environement/test you're using?
Hi François, So yeah, the timing provided by Idefix between the pushRegion and the popRegion is "all included", meaning that in your example the `EvolveStage` region timing will include the...
ok, I see. You might be facing an out of sync issue since you're doing a very large loop that is not finished on the GPU side, when the CPU...