On Thu, Jan 15, 2015 at 11:33:22AM -0800, Kenneth Graunke wrote:
> On Thursday, January 15, 2015 08:56:23 AM Ben Widawsky wrote:
> > On Thu, Jan 15, 2015 at 01:41:14AM -0800, Kenneth Graunke wrote:
> > > There's some debate about whether we should use Meta or BLORP,
> > > but either should run circles around the BLT engine.
> > > 
> > > Improves performance in "copypixrate -blit -back" (from Mesa demos)
> > > by 232.037% +/- 3.15795% (n=10) on Broadwell GT3e.
> > > 
> > > Signed-off-by: Kenneth Graunke <[email protected]>
> > > ---
> > >  src/mesa/drivers/dri/i965/intel_fbo.c | 14 +++++++-------
> > >  1 file changed, 7 insertions(+), 7 deletions(-)
> > > 
> > > Ben: this might be a good candidate for your braswell branch.
> > > 
> > 
> > I'll give it a run through the benchmarks assuming I can get a working 
> > machine.
> > I am always concerned about this kind of switch because it seems very 
> > benchmark
> > dependent to me. If we happen to have benchmarks which are extremely GPU/RCS
> > limited, I could only imagine this would hurt, for example.
> > 
> > Anyway, I'll let the data speak for itself when I have it. Thanks for 
> > giving me
> > new things to try.
> 
> I doubt that it's very benchmark dependent.  In every case that we've looked
> at, the BLT has proven to be slightly slower to massively slower.  I've never
> seen it be faster.
> 
> --Ken

It seems like if you have to stall on the blit to complete, and the 3d ring is
heavily used, you could find a win with blit, no?
_______________________________________________
mesa-dev mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/mesa-dev

Reply via email to