Current xmame's pretty much a draw (the real barricade now is X11 itself, which isn't designed for gaming like DirectX). When we started, it was 6-8x faster than xmame, so Lawerence and Hans have done some good work. OTOH, we use MMX/SSE/SSE2 assembly blitters for final video output so on games larger than 320x240 or with RGB/scanlines/etc enabled we win by an increasingly large amount.
