dale@amiga.UUCP (05/12/87)
In an article that Tom Rokicki wrote a while back discussion was about clearing memory. I think he was fairly accurate in describing how fast the 68k can clear memory. So using movem.l the processor is able to use every cycle available to it. But since the memory actually has double cycles and the blitter can access each of the cycles it sounds reasonable that the blitter is able to clear at twice the speed. Well sorry, fellas, almost. maybe next time. The blitter is unable to use the same pointer in successive memory slots so each clear cycle (assumeing only DEST dma) is seperated by an unused cycle. So infact it is only as fast as the 68000 is. But this leads into a trick I was playing with for awhile. Maybe someone wants to take it up. Since the blitter is only using everyother cycle. This leaves every other cycle open for the 68000. So to get the fastest clear routine in chip memory, you use both. hint, hint. For copying words from one place to another, the blitter wins hands down. It is using two different pointer (SRC|DEST) so it reads in one cycle and writes in the very next. Well there is a pipeline start up, but the end result is the same. Duck