Does A Greedy Intel Driver Improve Performance?

In this test, the Intel Linux driver seems to have regressed a fair amount between the 2.4.1 and 2.6.3 releases. Using UXA had led to a performance slowdown.

Using UXA was slow, again, but the greedy migration heuristics led to a performance boost with EXA. The xf86-video-intel 2.6 driver though was still much slower than the older 2.4 release from Q4'2008.

A greedy Intel driver led to the best performance with the 128x128px mixed composition. The UMA acceleration architecture was also faster than the EXA default, but still comparing EXA to EXA between the 2.4 and 2.6 releases, there is a clear regression.

At simple blitting, the default EXA was the fastest and there was a nice speed-up compared to the 2.4 driver.