On Friday, August 07, 2015 09:13:11 PM Chris Wilson wrote: > Combining the multiple access checks into a few batches and a single > serialising read can reduce detection times from around 100us to 70us on > a fast Haswell system. > > Signed-off-by: Chris Wilson <ch...@chris-wilson.co.uk> > Cc: Kenneth Graunke <kenn...@whitecape.org> > --- > src/mesa/drivers/dri/i965/intel_screen.c | 165 > +++++++++++++++++++------------ > 1 file changed, 101 insertions(+), 64 deletions(-)
I don't think we should do this. The existing AMD_performance_monitor and INTEL_performance_query code produces bogus data, and is not useful. Robert Bragg's patches from May essentially delete our existing OA backend, and replace it with a new one that uses perf ioctls. At that point, we no longer need to write OACONTROL directly from userspace, so we don't need to check whether it works. This brings us back down to one register write check. I'm not sure about the status of Robert's patches, but we can certainly delete our existing implementation in the meantime. --Ken
signature.asc
Description: This is a digitally signed message part.
_______________________________________________ mesa-dev mailing list mesa-dev@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/mesa-dev
