Copying from Linux

Jay Monkman jtm at
Fri Jul 19 16:36:11 UTC 2002

On Fri, Jul 19, 2002 at 10:42:33AM -0500, Joel Sherrill wrote:
> How does it compare to the *BSD memcpy for the arm? or newlibs?

I just tested the one from NetBSD, and it's faster than the generic
one, but slightly slower than Linux's. I did only test two cases with
each copying 256K from a 4 byte aligned region to a 4 byte aligned
region, and 256 from a 4 byte aligned region to a misaligned region.

For the aligned copies, the Linux routine is about 4X faster and the
NetBSD routine is about 3X faster than the generic one. For the
misaligned copy, they were both about 9X faster.

I couldn't find an ARM specific memcpy in newlib-1.9.0.

I put the function in exec/score/cpu/arm/memcpy.S, and modified the, but when I write a test program, it doesn't get picked
up. Do you know what else I have to change? 

Jay Monkman	    The truth knocks on the door and you say "Go away, I'm 
                    looking for the truth," and so it goes away. Puzzling.
		     - from _Zen_and_the_Art_of_Motorcycle_Maintenance_

More information about the users mailing list