I am attempting to get hardware acceleration for the YUV overlay on a Linux box with a Riva TNT2 card. From what I understand, this may be possible using DirectFB and SDL. Is this correct?

Okay - on Linux when I compile SDL with DirectFB then Daphne runs DL really quick (and w/o a problem).

Similar spec machine - when I compile SDL w/o DirectFB, DL runs poorly.

However, the output messages still say that I do not have hardware acceleration. Is this correct?

I suspect that you are getting some form of hardware acceleration, which accounts for your increase in speed, but you are not getting the type of hardware acceleration that uses the hardware to convert and scale the YUV image to RGB. In other words, your CPU may still be doing that work for that. If you are in fact getting this YUV->RGB optimization, then that would mean that there is a bug in SDL as far as reporting the acceleration status.

Well, I do get the different results depending on how SDL is compiled. That is what led me to believe DirectFB is acting up.