On my C1000 I have Cacko 1.23 and libsdl-1.2.5-bvdd-07-2 (piro).
SDL stuff always seemed quite slow to me.
Now I did a comparison between this, a pure Qtopia-SDL on the C1000, and the W100-SDL (piro) on my C760.
"testpalette" from the libsdl source showed these results:
C1000 bvdd: Â ~7 frames/second
C1000 qtopia: ~9 frames/second
C760 Â w100: Â ~14 frames/second

I suppose I'm doing something wrong. My Japanese may be a little rusty or my environment wants some variables ...
Or is it really a driver thing?
What are YOU getting from your SDL ?
For your convenience I attached the testpalette files for ARM.