Short: Crap_c2p (a benchmark for very poor c2ps) Uploader: amlaukka cs helsinki fi Type: game/gag Architecture: m68k-amigaos description: To emphasize the point of using a good c2p I decided to make a benchmark in the style of Jyrki Saarinen's c2p benchmark which was posted to comp.sys.amiga.programmer a while ago. There is one important difference though. The screen size is 320*64 which is due to the reason that the timing routine has 100 ms as maximum. Some of these c2ps don't quite make it so if you want to compare these times with the original benchmark, multiply the times by two. Featured c2ps so far: WritePixelArray8() - the OS function (performance may vary depending of OS version and installed patches etc tkg_c2p - ever wondered why it was so slow? (multiply time by four to get the time to c2p a whole 1x1 screen addx_c2p - maybe the classic case of bad approach. Very simple to make and very slow. Absolutely no effort has been done to optimize this one - not that there's much room for improvement. tabled_c2p - yeah, reorder the bits with tables. What an interesting idea, here you can see why not and besides it needs 2 MB contiguos fast memory. Copy_MOVEMOVE - just as comparison a simple copy loop to get idea what you should expect from a good c2p. For example kalms_CPU3_BLIT1 reaches quite close to this speed. more to come ... Results (or the lack of them) on A1200/040/040: Name Count Mid. time (ms) -------------------------------------- WritePixelArray8: 66 55.79 TKG_c2p: 66 17.83 addx_c2p: 66 22.14 tabled_c2p: 66 11.09 Copy_MOVEMOVE: 66 3.79 credits: Aki Laukkanen - putting things together Teemu Suikki - speed.i macro Jyrki Saarinen - inspiration Andy Clitheroe - TKG c2p (without his permission, sorry) Ville Helin - RRR-yc2p (tabled one) from aminet