[PATCH 14/20] glamor: stub out lines

Markus Wick markus at selfnet.de
Mon Mar 24 14:11:43 PDT 2014


Am 2014-03-24 20:47, schrieb Keith Packard:
> Even still, uxa is significantly faster than glamor on this one; uxa 
> has
> a simple test for a request that contains only vertical and horizontal
> lines and fills those as rectangles. We could easily do the same thing
> in glamor, which should yield better performance for this case. In 
> fact,
> I suspect we could do the test for all vertical/horizontal with the CPU
> and let a vertex shader compute rectangles from the existing line
> coordinates. Might be a fun exercise.
> 
> 1: intel-glamor-line.perf
> 2: intel-uxa.perf
> 
>        1                 2                 Operation
> ------------   -------------------------   -------------------------
>    4020000.0     10700000.0 (     2.662)   500-pixel horizontal line 
> segment
>     177000.0      9860000.0 (    55.706)   500-pixel vertical line 
> segment

But shouldn't UXA be affected as much as glamor by cache issues?
My test results on ivb glamor vs SNA looks funny as SNA seems to be as 
slow as glamor:

glamor:
3000000 reps @   0.0004 msec (2350000.0/sec): 500-pixel horizontal line 
segment
  600000 reps @   0.0018 msec (565000.0/sec): 500-pixel vertical line 
segment

SNA:
8000000 reps @   0.0001 msec (7290000.0/sec): 500-pixel horizontal line 
segment
  400000 reps @   0.0024 msec (413000.0/sec): 500-pixel vertical line 
segment

As UXA is so much faster, could it be because of line rasterization 
limits of the gpu itself?


More information about the xorg-devel mailing list