Current xaos doesn't enable thread support, and so it can
only utilize a single core on today's multicore machines.
It might also be worth enabling i386 inline assembly (for
i386 only) and 'long double' (on i386 & x86_64).

There is a small penalty for enabling threads, unless you
actually use > 1 threads in which case of course it's a big
win.

Here are some measurements, all measured on a 4 core AMD,
using 'xaos -speedtest [-threads N]':

Current xaos-3.4-2:

Driver speed: 1160 FPS (573.4863 MBPS)
Memcpy speed: 34138.4 FPS (16877.5049 MBPS)
Missaligned memcpy speed: 18535 FPS (9163.4216 MBPS)
Size 6 memcpy speed: 1180 FPS (583.3740 MBPS)
Result:0 Formulaname:Mandelbrot Time:104935 Mloops per sec:190.59
Result:0 Formulaname:Mandelbrot Time:103556 Mloops per sec:193.13
New image caluclation took 0.0050994 seconds (2e+02 fps)
Approximation loop speed: 1381.8 FPS

Added long-double, thread support, using a single thread:

Driver speed: 1202.2 FPS (594.3494 MBPS)
Memcpy speed: 19844.8 FPS (9810.9668 MBPS)
Missaligned memcpy speed: 18606.6 FPS (9198.8196 MBPS)
Size 6 memcpy speed: 1018.2 FPS (503.3826 MBPS)
Result:0 Formulaname:Mandelbrot Time:103142 Mloops per sec:193.91
Result:0 Formulaname:Mandelbrot Time:111102 Mloops per sec:180.01
New image caluclation took 0.0062816 seconds (1.6e+02 fps)
Approximation loop speed: 1347.8 FPS

(It's not clear what happened to 'Memcpy speed' in that test,
but it doesn't seem to affect the performance very much).

Added long-double, thread support, using 4 threads:

Driver speed: 1222.4 FPS (604.3359 MBPS)
Memcpy speed: 34212.4 FPS (16914.0894 MBPS)
Missaligned memcpy speed: 18561.2 FPS (9176.3745 MBPS)
Size 6 memcpy speed: 1274.6 FPS (630.1428 MBPS)
Result:0 Formulaname:Mandelbrot Time:103200 Mloops per sec:193.80
Result:0 Formulaname:Mandelbrot Time:111225 Mloops per sec:179.82
New image caluclation took 0.008562 seconds (1.2e+02 fps)
Approximation loop speed: 4334 FPS