Timing graphs for ARM asm memcpy

[TP Diffenbach]

Please examine this page of timings, which compares the time cost of four memcpy implementations:
http://diffenbach.org/rockbox/memcpy/memcpy.comparison.0.256.html

And let me know if you think my memcpy is fast enough to be used instead of the uclinux memcpy.

uclinux is (slightly) faster for some small unaligned copies with copy lengths evenly divisible by powers of two. Mine is faster for any copies of lengths > ~ 56 bytes.

Thanks.

[Bagder]

I think you and your version just rock! ;-)

[TP Diffenbach]

Thanks. What we really need to do is discover how mmecpy is being called, esp. for unaligned copies. But I'm going to let someone else do that.