On 2019-04-04 07:26:37 [-0700], Andy Lutomirski wrote: > I think that David was asking whether we could make kernel_fpu_begin() > regions sometimes be preemptible. The answer is presumably yes, but I > think that should be a separate effort, and it should be justified > with improved performance above and beyond what we get with Jason's > simd_get() stuff. Last time I checked we had no very SIMD loops unless you count $ALG over 1 MiB data or so. And in that case you could use the kernel_fpu_begin()/end combo after a complete SIMD loop. I would like see that at least while going from one page to the next. Sebastian