On Thu, Aug 25, 2016 at 12:51:16PM +0300, Mika Kuoppala wrote: > Chris Wilson <ch...@chris-wilson.co.uk> writes: > > > Similar to the issue with reading from the context status buffer, we > > frequently write to the ELSP register (4 writes per interrupt) and know > > we hold the required spinlock and forcewake throughout. We can > > shortcircuit the I915_WRITE() by precomputing the address of the ELSP > > I915_WRITE_FW are for these situations so the commit message > seems stale in this part. > > > register and avoid all the known checks. > > ...and the checking part is false claim?
I was thinking of WRITE_NOTRACE :| > We get precomputing the loc and removal of the superfluous > posting write so it is an improvement. Yup, tightened up the language to not claim what we already did in 26720ab97fea. -Chris -- Chris Wilson, Intel Open Source Technology Centre _______________________________________________ Intel-gfx mailing list Intel-gfx@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/intel-gfx
