On Tue, Dec 18, 2018 at 08:53:54AM -0500, Thor Lancelot Simon wrote: > They did _not_ cause measureable performance problems of any kind, and > though it is theoretically possible to do this sort of thing via a > tightly-protected userspace helper process, I prototyped that too and > it gets very ugly, very fast -- the in-kernel way with a thread is, I > believe, better.
Also note that it is pretty straight-forward to give such a thread the "right" to use the FPU (or similar PCUs) in the kernel. Martin