R300 Gallium3D Is Now Separate From Classic Driver

[phoronix]

Phoronix: R300 Gallium3D Is Now Separate From Classic Driver

Up to this point the ATI "R300g" driver that provides Gallium3D support for Radeon GPUs up through the Radeon X1000 (R500) series has depended upon files from the "R300c" classic Mesa DRI driver when being compiled. In particular, the R300c shader compiler and its nearly 20,000 lines of code. The R300c compiler has now been copied over directly to live separately within the R300g driver, which means the classic R300 driver can be left to fade off and die...

http://www.phoronix.com/vr.php?view=OTcxNw

[plonoma]

This should have been done a lot earlier.

Is there any chance this becomes part of Mesa 11 release?

[Azpegath]

Wait, you're saying ATI/AMD hasn't killed off UMS, but that their KMS support is solid; by that, do you mean ATI/AMD as in the FOSS team or as in the Catalyst driver? Because if Catalyst supports KMS, it's huge news to me.

[elanthis]

Wait, you're saying ATI/AMD hasn't killed off UMS, but that their KMS support is solid; by that, do you mean ATI/AMD as in the FOSS team or as in the Catalyst driver? Because if Catalyst supports KMS, it's huge news to me.

Obviously he means the FOSS driver. I wouldn't expect any of the binary drivers to support KMS any time soon.

[Azpegath]

Obviously he means the FOSS driver. I wouldn't expect any of the binary drivers to support KMS any time soon.

Yeah, I figured. Somewhat unclear wording though.

[oibaf]

What was the point for this?

I mean, the r300g compiler appears to be no longer actively developed (after the copy only this warning fix get committed and it clearly could be applied also to the r300 classic compiler) and if it was keep compatible with both drivers up to now, it shouldn't too difficult to keep compatibility for future bug fixes.

[agd5f]

What was the point for this?

I mean, the r300g compiler appears to be no longer actively developed (after the copy only this warning fix get committed and it clearly could be applied also to the r300 classic compiler) and if it was keep compatible with both drivers up to now, it shouldn't too difficult to keep compatibility for future bug fixes.

No one develops r300c actively anymore. There have been a number of cases where changes to the compiler broke r300c, but no one noticed for a while. Now we can make complier changes as needed to r300g without worrying about r300c.