The current implementation has commented out a number of areas where particles might be lost, and replaced it with a simple printf statement. It would be better to use the mark_as_lost_short() functions to ensure that the particle weights are set to zero. In adding some monitoring code and testing 100M particle histories on GPU, I'm seeing zero lost particles, but it would be good to add these calls back in and ensure lost particle treatments are working as expected. I'll also note that some of the more common lost particle checks are still in place, but this issue just highlights that a few of the less common ones are missing currently.
The current implementation has commented out a number of areas where particles might be lost, and replaced it with a simple
printfstatement. It would be better to use themark_as_lost_short()functions to ensure that the particle weights are set to zero. In adding some monitoring code and testing 100M particle histories on GPU, I'm seeing zero lost particles, but it would be good to add these calls back in and ensure lost particle treatments are working as expected. I'll also note that some of the more common lost particle checks are still in place, but this issue just highlights that a few of the less common ones are missing currently.