AMD 2010 Catalyst Driver Year In Review

[smitty3268]

Why is phoronix still peddling this drivel that the new catalyst 2d acceleration is "derived from windows 2d code paths" or derived from direct2d. I recall bridgeman himself saying that the new architecture (ATI 2D Acceleration Architecture) has nothign to do with windows in the slightest.

The switch for enabling it was simply a re-used direct2d development switch, the architecture itself has nothing to do with it.

It has nothing to do with Microsoft's Direct2D API.

It is the same internal driver code that is used in Windows, though, at least a good portion of it. It's just not Direct2D, but rather an internal driver API.

[bwat47]

Why is phoronix still peddling this drivel that the new catalyst 2d acceleration is "derived from windows 2d code paths" or derived from direct2d. I recall bridgeman himself saying that the new architecture (ATI 2D Acceleration Architecture) has nothign to do with windows in the slightest.

The switch for enabling it was simply a re-used direct2d development switch, the architecture itself has nothing to do with it.

[Mr James]

What about varying resolutions?

[Michael]

Perhaps that could be built directly into the test profile, so that it could determine a specified information line to add to all graphs based on that test?

There already is with the SubTitle XML tag.

[smitty3268]

I don't know about that, just look at the article and the questions it has raised here about tessellation levels for example.

I'm with Michael on this, I don't want to have a paragraph of test details on top of every graph, and don't see any simple way of determining exactly which are the important ones to show for each test.

Perhaps that could be built directly into the test profile, so that it could determine a specified information line to add to all graphs based on that test?

[deanjo]

They can easily append the information as a footnote themselves. The test profiles are standardized so it's not like there is some unknown factor or uncertainty to that if comparing the same versions as long as the person knows the test profile/version in use.

I don't know about that, just look at the article and the questions it has raised here about tessellation levels for example.

[Michael]

Well that makes it a bit difficult for those who would like to use the results in PDF's and presentations.

They can easily append the information as a footnote themselves. The test profiles are standardized so it's not like there is some unknown factor or uncertainty to that if comparing the same versions as long as the person knows the test profile/version in use.

[deanjo]

Instead anyone who really cares about those details outside of a PTS environment can click the title of the graph and get to the test profile page on OpenBenchmarking.org and view the test profile information or the raw profile itself.

Well that makes it a bit difficult for those who would like to use the results in PDF's and presentations.

[Michael]

I'm talking about something like this Michael, rough mockup:

Would be tough to generalize all of that needed information in a fully automated manner since pts-core wouldn't know what is important to scan and find in a test profile to report... Especially with all game config files being different, etc. Anything not fully automated I won't do unless someone provides me with all of the necessary patches.

Instead anyone who really cares about those details outside of a PTS environment can click the title of the graph and get to the test profile page on OpenBenchmarking.org and view the test profile information or the raw profile itself.

[deanjo]

Or add the test params after/under the resolution on the top.