Subject: Re: Decision time: Intel TSX Lock elision for glibc.
Date: Tuesday 2nd July 2013 16:23:55 UTC (over 4 years ago)
On 07/02/2013 11:53 AM, Andi Kleen wrote: > On Tue, Jul 02, 2013 at 11:10:43AM -0400, Carlos O'Donell wrote: >> On 07/02/2013 02:59 AM, Andreas Jaeger wrote: >>> On 07/02/2013 01:34 AM, Carlos O'Donell wrote: >>>> [...] >>> >>>> In summary: >>>> >>>> * You can check patches 1, 2, 3, 4, 5, and 7 >>> > > I checked in the approved patches. Thanks everyone for your support. > > * Added support for TSX lock elision for pthread mutexes on i386 and x86-64. > This may improve lock scaling of existing programs on TSX capable systems. > When the --enable-lock-elision=yes parameter is specified at configure > time lock elision will be enabled by default for all PTHREAD_MUTEX_DEFAULT > mutexes. Thank you for checking this in. I appreciate the work you've put in to enable this support. I know it has not been easy and that it has consumed a lot of you time, and I appreciate that. There is still more work ahead of us as we look at the top-down API required to directly or indirectly control elision for a lock. > How can a decision be made on the rwlocks now (and later the other patches?) I will reach out the Austin Group immediately to discuss this issue. Cheers, Carlos.