> The best(?) I could try was 4 bit on my 16Mhz IIIx (and I had to bypass the > fastEnough check). The old version was consistantly 0.03 seconds and the > new was 0.01 or 0.02 seconds. Quite a difference!!
(You could put in a for ( i = 0 ; i < 50 ; i++ ) just before the call to RotateBitmap() in rotate.c and get more precision in the timing if you want.) Do you think that now it's fast enough on slower units to make the fullscreen images always rotate? What is interesting is that this seems to have made a particularly large difference on 68K units--on my ARM based NX, the difference is fairly small when running the armlet (20%) but quite large if compiled to 68K code (50-100%). This suggests that when optimizing for 68K units, we should not loop through an array via indexing but should do so with a pointer. Alex _______________________________________________ plucker-dev mailing list [EMAIL PROTECTED] http://lists.rubberchicken.org/mailman/listinfo/plucker-dev
