rkierulf

I think I've found the cause of the Simple and Arbitrary Motion test failures for Metal: in UnitTime.jl, this line is processed incorrectly: ``` return min.(max.((t .- t_start) ./ (t_end...

Interesting! My guess is that the transition from flat to linear growth begins when the simulation becomes large enough to fully saturate the resources of the GPU. After that, due...

The tests are all passing now, except for the oneAPI tests with Julia 1.11. For the 1.11 tests overall, I'm not sure it is using the local versions of the...

Great! I will add some more info to the docs when I have more time. In the meantime, it would be a good idea for someone with a Nvidia GPU...

For now, I would recommend testing with KernelAbstractions v0.9.33 - the 9.34 change made a big difference in performance, since it rearranges the way valid index checks work to ensure...

Ok, this should now be ready for review. I added some info to the docs, and also changed the default groupsizes for both kernels to 256 to keep occupancy high....

Yep, this should be ready to merge!

It looks like the array is just `Float32[1.0f-14, 1.0f-14]`: https://buildkite.com/julialang/komamri-dot-jl/builds/1421#0195c05a-b862-499f-beda-971831f858a9. There is also a warning before about global hostcalls.

Ok, it appears the issue is with a different cumsum here: https://github.com/JuliaHealth/KomaMRI.jl/blob/master/KomaMRIBase/src/timing/TrapezoidalIntegration.jl#L49. This is a 2D cumsum of a matrix across the second dimension. Let me know if you are...

This build has the values printed: https://buildkite.com/julialang/komamri-dot-jl/builds/1428#0195d002-435b-400a-9d1b-1df5624de035. The matrix before the call to cumsum where it crashes has shape 1 x 548 and consists of all zero Float32 values. I...