AMD Radeon R600g Gallium3D: Mesa 8.0.4 vs. 8.1 Git

[agd5f]

It might also be worth checking that both runs of the applications are using the same rendering paths. 8.1 supports a number of additional GL extensions which may cause the application to choose a more advanced rendering path on 8.1 vs. 8.0.

[Adarion]

Interesting to see how some thing go through the roof performance wise but same card/chip with a different applications sees quite severe regressions. Well as long as mesa devs notice this and get the latter fixed before release it shall be fine. It is a intermediate development "release" anyway so I wouldn't build a house on it.

[gQuigs]

If you have the hardware in question...

you can bisect them too.. manually it's not that difficult. (now fixing the cause on the other hand)

https://wiki.ubuntu.com/X/BisectingMesa

I haven't been able to find any pts bisect docs.. and it doesn't appear as a module for me...

[Hamish Wilson]

Hope some of these issues get fixed before release (typing from a machine with a Radeon HD 4670).

[curaga]

+1 for a Llano re-test, making sure the clocks are up

[jakubo]

i wonder whether there is any news on the APU graphics driver side.
it's been some time, hasnt it?
Besides, i wonder if the results would differ on a bulldozer system. maybe compiled with bdver..
(im considering to buy one so i wouldnt want something "fixed" for intel when it would possibly slow down AMD)

[Michael]

I've thought the test suite supported bisecting to identify performance regressions?

It does, but I don't have the time right now. I'm going to be out of the country for a week so I have other articles to get done prior to that on the Ivy Bridge system, etc.

[darkbasic]

Right, what about the mesa regression-spotter rig you talked about?

[log0]

I've thought the test suite supported bisecting to identify performance regressions?

[curaga]

No love for 4870 it seems.