Le 14/10/2016 à 21:34, Joerg Sonnenberger a écrit :
On Fri, Oct 14, 2016 at 08:56:18PM +0200, Maxime Villard wrote:
We would either be leaking memory or freeing some we allocated for no
reason, the latter meaning we still didn't fix the issue for real. I'm not
sure multiplying critical sections is a good idea.
I don't think we normally migrate running lwps from one cpu to another
for no good reason. It would be quite inefficient to do so.
LAPIC timer? It is a perfectly legitimate reason to stop an lwp and start
executing another one on the same cpu.
Anyway, this is off-topic. If we were still able to implement the cache
properly, we would probably end up adding more branches than I am now in my
patch, so I'm not convinced the performance would be better.