Closed ax3l closed 7 years ago
Sounds promising, but I think the most interesting part of a student project would be how to implement it in Alpaka or at least a fallback of jit compiling is not available for a back end / environment.
Also related to this issue of alpaka: https://github.com/ComputationalRadiationPhysics/alpaka/issues/208
true, related to #24
I added it therein and close this duplicate issue
In order to prevent NxN pre-compiles of functor chains and long startups on CUDA architectures, NVRTC is an option we long thought about.
I just found this library from Nvidia, jitify. It looks like it could significantly easen the process of writing our functor chains in NVRTC. Definitely worth investigating that, maybe as a student project.