On 05/29/2013 09:51 AM, Bernhard Rosenkränzer wrote:

https://docs.google.com/a/linaro.org/spreadsheet/ccc?key=0AhSH680iSjV_dEpYVn...

Looks interesting! Although there is a fair amount of noise in some of the non-memcpy results, so you might want to do a higher-number of runs then just 3 each.

Also, you might also highlight the regressions in the smaller memcpy, just so folks don't catch it on their own and think you're ignoring it.

But those A15 numbers are really substantial! Very cool!

thanks -john