> If it's any consolation, its userland thread switches between the same > process are much faster than the NPTL kernel version at present (fixes > are in course, if not already mainline). Hm? What is this referring to?
When I pulled out, switches between different threads on ARM NPTL implied the same overhead as a process switch, but somebody had a scheme to fix that by spotting when it was a different thread of the same process and so not shifting so much state. I expect the invalidation or flushing of some cache or other in the process had something to do with it, but you know how little I understand about that :) WIth linuxthreads instead, no kernel context switch happens cos it's all done within one Unix process anyway, no? I can try and dig up the mail thread if that'll help M