CompuServe Thread

If you want to clear the mememory once when it's allocated the memory allocator can be told to that. If you are wanting to clear the memory more then once, then this is how I would do it. First I would make sure that the amount of memory is evenly divisuble by four, by rounding up the # of bytes you need to the next highest longword, and then clear by longwords. ;a0-pointer to top of memory to clear, longword aligned ;d0-# of bytes to clear, evenly divisible by 4 add.l a0,d0 ;address of longword to stop at clr.l d1 loop1: move.l d1,(a0)+ cmp.l d0,a0 bne loop1

Jay, The fastest technique I can think of for zeroing a large block of memory would be the following code, where A0 is the start of the block: movem.l d1-d7,a1-a6,-(a7) ; save as many registers as possible (13) ; zero the registers d1-d7,a1-a6 add.l #size,a0 ; start at the top and work down ; I assume here that A0 will be even move.l #(size/(4*13))-1,d0 loop movem.l d1-d7/a1-a6,-(a0) ; zero 4*13 = 52 bytes per loop dbf d0,loop movem.l (a7)+,d1-d7,a1-a6 ; restore the registers Each loop clears 52 bytes and takes 148 cycles, so under 3 clock cycles per byte. Compare this with 3 cycles per byte for MOVE.L D0,(A0)+ plus a further 2.5 cycles per byte for looping. Of course, you have to deal separately with the few bytes left over if the block size is not 52. Personally, I would try to avoid clearing a large block of memory. Do you _really_ need to clear this memory? Surely you will be writing something else into it later, so why zero it now? How about using a flag to indicate whether or not the memory contents are valid? Regards, Shraddhan – via Whap! from Hertfordshire in the UK