Sergey Bugaev, le mar. 07 sept. 2021 16:33:23 +0300, a ecrit:
> This is a new implementation of GSCOPE which largely mirrors its NPTL
> counterpart. Same as in NPTL, instead of a global flag shared between
> threads, there is now a per-thread GSCOPE flag stored in each thread's
> TCB. This makes entering and exiting a GSCOPE faster at the expense of
> making THREAD_GSCOPE_WAIT () slower.
> 
> The largest win is the elimination of many redundant gsync_wake () RPC
> calls; previously, even simplest programs would make dozens of fully
> redundant gsync_wake () calls.

That looks good, thanks!

Samuel

Reply via email to