Felipe Contreras

I’ve been using radeonhd since the launch. Actually before I was using avivo. I was subscribed to the mailing list and I was filling bug reports. From my point of view I was being an ideal user. I even hacked the code back in the avivo time.

My contributions were apparently irrelevant, I didn’t knew what I was doing with the code and although it seemed to improve things for me, the developers were not interested at all in my patches. That was ok.

Then the mailing list was getting filled with bug reports, I suggested a separate mailing list for that and I was ignored, so I unsubscribed. I kept up to date with the code, recompiling it each time there was significant changes.

Somehow EXA was never working for me, so when I found out there was an IRC channel (thanks to phoronix), I decided to join and ask some questions. Luc Verhaegen was pretty responsive and that was good, I felt my issues could finally be solved. I was wrong.

I thought longstanding issues are more important than new bugs, both as a user and as a developer, but Luc didn’t think so, and I’ll spare the details of what happened in IRC, but the end result was that he wasn’t going to help me, even though he already had the setup to do it.

So I did the only responsible option which is to fill a bug report, it was immediately closed due to wrong reasons by Egbert Eich, apparently I had to prove myself worthy of attention but in the end the bug was closed as invalid.

Apparently as a user I must know that:

• EXA needs DRI
• DRI is disabled by default
• 2560×2560 is too big for 64Mb of ram if you want DRI
• 2560×2560 is the default virtual size

So I had to set a smaller virtual size, manually enable DRI, and then I could use EXA… obviously!

This of course was not explained anywhere, I suggested to add the DRI dependency on the manpage or the wiki, I was ignored, but the default virtual size is something that should just work, I mean, I wasn’t setting that 2560×2560 virtual size, obviously it was radeonhd’s code, so the code is doing something that it can pretty easily see it’s not going to work, why should the user care?! So I filled another bug report for that. Luc closed immediately stating that:

Try setting virtual yourself like a sane human being.

Alex Deucher joined the discussion with a reasonable point of view, but nothing happened.

So when Luc wrote this blog post about the CS stuff I couldn’t help but to write this one.

I review most of the commits of both ati and radeonhd drivers, and from my limited knowledge the radeonhd has a more interesting technical standpoint, but what good is a display driver that you can use properly only with certain hardware? Yeah, there’s r520 support, but only for the cards they test, if you have 64Mb, well, you suck.

There’s more to a software project than beautiful code; it has to work properly. If the main developers are not able to understand that the driver should work on your hardware (just work) then you are better using something that is ‘worse’ technically, but ‘better’ in the end, like the ati driver.

So long radeonhd.