bzero optimization rarely does

[Roger Sayle]

> But really nobody doesn't use the default value ot CLEAR_RATIO.
> You're still thinking of MOVE_RATIO and memcpy, but this is just a
> test of memset.  (Feel free to write a memcpy test based on this...)

Given that GCC has only had CLEAR_RATIO for less than 24 hours, its
probably not too surprising that most backends have not overridden
the default yet :>

> Care to take a stab at the revision?

Ok. I'll update memset-3.c and write a similar memcpy-3.c.
Give me a day or two to get the MAX values I can.  If we have to
add new testsuite failures on targets that work perfectly, I'd
prefer to avoid any that I have access to or can simulate.

Roger
--