dri/ Radeon Companion 3

The irregular Radeon-Development companion

Issue for 2007-04-01

Intro

This page is an accumulation of the events happening in Radeon development, mainly on IRC or on dri-devel. The concept has been shamelessly copied from nouveau and will hopefully be kept up. At least this is the second issue. So didn't immediately stop.

The TiRDC gained a lot of responses lately. So we try to keep our readers up to date with this issue. This issue will be a bit shorter because the author lacks time. But the most important events will be noted.

  • After Nicolai Haehnle (nh) started working on piglit, he detected some bugs and some regressions. Luckyly after he found them he started fixing them. Source

  • Mesa got the GLSL-1 branch merged. That means you can use mesa with GLSL shading programs. Thanks Brian Paul for doing the work of merging and thanks to anyone for writing that code. Source

  • As a sidenote the X.org ATI driver xf86-video-ati got a release candidate. It contains a major re-write of the output mappings from what was known as "Alex's superpatch". Please log all regressions into bugzilla so they can be fixed before the 6.7 release of the driver.

  • Thomas Hellström came up with a patch to allow single-device-mutli-screen DRI. Neat idea. Source

  • Dave Airlie came up with the idea that TTM (the new and way better memory manager in mesa) should be used in wide areas. With his knownlege of TTM (aquried from Thomas Hellström, one of the guys behind TTM) and his great knowledge of the Radeon-Drivers, he ported the work from i915 over to R300. Note: The TTM support is not complete or offical yet. It only supports fencing and buffer management, which is the absolute basic of TTM. Source

  • Jakob Bornecrantz (Wallbraker) and Jesse Barnes (jbarnes) continue working on DRM modesetting. They don't have a public git repository or something yet, but that start a big discussion. Former developers of KGI (the kernel graphics interface) joined #dri-devel and disscussed about the modesetting in DRM. Dave Airlie took part in the discussion and added information about framebuffer drivers. All in all the idea of having a central entity for managing modesetting for framebuffer and X.org drivers was agreed to be a good idea, even if some though it might not be placed right in the DRM. Source

Help wanted and needed

We could really need help in finding regressions that might come from the GLSL-1 branch merge. It would be great if you could run piglit against * latest fglrx * dri drivers from before the GLSL-1 merge (maybe mesa 6.5.2 or better git from about 2 weeks ago) * latest dri drivers from git

It would also be great if you could test Dave Airlie's TTM branch of R300 against latest git of mesa.