[Mesa-dev] [PATCH 00/15] A bunch of shared code and RadeonSI changes

Dieter Nützel Dieter at nuetzel-hh.de
Thu Oct 18 02:05:20 UTC 2018


GREAT work Marek!

Best speed up for months on Polaris 20, at least.
Coming from vacation with injured right ankle joint, so I haven't had 
time for testing before commit. But 'glmark2' numbers are better than 
before all the Spectre shit (~8-9%?!).

In German: 'Da geht noch was...' ;-)

Greetings,
Dieter

=======================================================
     glmark2 2017.07
=======================================================
     OpenGL Information
     GL_VENDOR:     X.Org
     GL_RENDERER:   Radeon RX 580 Series (POLARIS10, DRM 3.26.0, 
4.18.14-1.gce1c446-default, LLVM 8.0.0)
     GL_VERSION:    4.5 (Compatibility Profile) Mesa 18.3.0-devel 
(git-58a51d0a67)
=======================================================
[build] use-vbo=false: FPS: 3382 FrameTime: 0.296 ms
[build] use-vbo=true: FPS: 11679 FrameTime: 0.086 ms
[texture] texture-filter=nearest: FPS: 11607 FrameTime: 0.086 ms
[texture] texture-filter=linear: FPS: 11572 FrameTime: 0.086 ms
[texture] texture-filter=mipmap: FPS: 11676 FrameTime: 0.086 ms
[shading] shading=gouraud: FPS: 12207 FrameTime: 0.082 ms
[shading] shading=blinn-phong-inf: FPS: 11892 FrameTime: 0.084 ms
[shading] shading=phong: FPS: 12073 FrameTime: 0.083 ms
[shading] shading=cel: FPS: 11763 FrameTime: 0.085 ms
[bump] bump-render=high-poly: FPS: 11252 FrameTime: 0.089 ms
[bump] bump-render=normals: FPS: 11366 FrameTime: 0.088 ms
[bump] bump-render=height: FPS: 11226 FrameTime: 0.089 ms
libpng warning: iCCP: known incorrect sRGB profile
[effect2d] kernel=0,1,0;1,-4,1;0,1,0;: FPS: 12171 FrameTime: 0.082 ms
libpng warning: iCCP: known incorrect sRGB profile
[effect2d] kernel=1,1,1,1,1;1,1,1,1,1;1,1,1,1,1;: FPS: 11314 FrameTime: 
0.088 ms
[pulsar] light=false:quads=5:texture=false: FPS: 10452 FrameTime: 0.096 
ms
libpng warning: iCCP: known incorrect sRGB profile
[desktop] blur-radius=5:effect=blur:passes=1:separable=true:windows=4: 
FPS: 5506 FrameTime: 0.182 ms
libpng warning: iCCP: known incorrect sRGB profile
[desktop] effect=shadow:windows=4: FPS: 5864 FrameTime: 0.171 ms
[buffer] 
columns=200:interleave=false:update-dispersion=0.9:update-fraction=0.5:update-method=map: 
FPS: 812 FrameTime: 1.232 ms
[buffer] 
columns=200:interleave=false:update-dispersion=0.9:update-fraction=0.5:update-method=subdata: 
FPS: 1128 FrameTime: 0.887 ms
[buffer] 
columns=200:interleave=true:update-dispersion=0.9:update-fraction=0.5:update-method=map: 
FPS: 893 FrameTime: 1.120 ms
[ideas] speed=duration: FPS: 2999 FrameTime: 0.333 ms
[jellyfish] <default>: FPS: 9422 FrameTime: 0.106 ms
[terrain] <default>: FPS: 1787 FrameTime: 0.560 ms
[shadow] <default>: FPS: 8930 FrameTime: 0.112 ms
[refract] <default>: FPS: 3418 FrameTime: 0.293 ms
[conditionals] fragment-steps=0:vertex-steps=0: FPS: 11901 FrameTime: 
0.084 ms
[conditionals] fragment-steps=5:vertex-steps=0: FPS: 11567 FrameTime: 
0.086 ms
[conditionals] fragment-steps=0:vertex-steps=5: FPS: 11614 FrameTime: 
0.086 ms
[function] fragment-complexity=low:fragment-steps=5: FPS: 11611 
FrameTime: 0.086 ms
[function] fragment-complexity=medium:fragment-steps=5: FPS: 11643 
FrameTime: 0.086 ms
[loop] fragment-loop=false:fragment-steps=5:vertex-steps=5: FPS: 11933 
FrameTime: 0.084 ms
[loop] fragment-steps=5:fragment-uniform=false:vertex-steps=5: FPS: 
11964 FrameTime: 0.084 ms
[loop] fragment-steps=5:fragment-uniform=true:vertex-steps=5: FPS: 11714 
FrameTime: 0.085 ms
=======================================================
                                   glmark2 Score: 9101
=======================================================

Before, even with DRM 3.27.0 (amd-staging-drm-next) I had

glmark2 Score: 8361

Am 03.10.2018 00:35, schrieb Marek Olšák:
> Hi,
> 
> Interesting bits:
> - CP DMA support for GDS (unused but there is a test)
> - switch back to DX sample positions
> - center the viewport in the scanline area for maximizing the guardband
> - optimal PA_SU_PRIM_FILTER_CNTL
> - higher subpixel precision for 4K and lower resolutions
>   (for more precise rendering of T-junctions in geometry)
> 
> Please review.
> 
> Thanks,
> Marek
> _______________________________________________
> mesa-dev mailing list
> mesa-dev at lists.freedesktop.org
> https://lists.freedesktop.org/mailman/listinfo/mesa-dev


More information about the mesa-dev mailing list