On Thu, Feb 07, 2019 at 06:37:35PM +0000, Stuart Henderson wrote: > On 2019/01/30 12:29, Martin Pieuchot wrote: > > On 28/01/19(Mon) 14:57, Martin Pieuchot wrote: > > > Summer is really warm here. No need to make my machines hotter by > > > spinning a lot. So here's a futex(2) based pthread_rwlock* > > > implementation. I should look familiar to those of you that reviewed > > > the mutex & semaphore implementations. It uses amotic cas & inc/dec. > > > > > > I moved the existing implementation to rthread_rwlock_compat.c to match > > > what has been done for semaphores. > > > > > > Tests, benchmarks and reviews are more than welcome! > > > > Newer version with fewer atomics and some barrier corrections as pointed > > out by visa@. > > This has been through 3 or 4 bulk ports builds on i386 (2 and 4 core > machines) and running on my amd64 workstation (4 core), no problems seen > on either. Has anyone else been running this? The reduction in spinning > is nice.
Yes, I've been running this diff since it was posted on my main laptop (x280). No issues seen and it feels a lot better, especially noticeable when switching back to an install without the diff.
