Degrader-1 regression test - DKS
Degrader-1 test might fail using the DKS version due to a different random number sequence used on the GPU. The impact of the random number sequence on the final energy can be few keW (saw up to 29 keV on the CPU with different SEED). I believe the test if fine, but for it to pass we need either a separate reference for the GPU version, where we can guarantee the same random number sequence or we need to adjust the tolerance for Ekin (its 0.001 keV at the moment)
The CPU tests are also passing/failing even if the seed is kept the same. With the same seed on the CPU I sometimes see up to 0.2 keV difference in Ekin from the reference. I'm not sure what is the reason for this difference since the random number sequence in the CollimatorPhysics should be the same in this case.