Il 15/04/2014 01:55, Michael R. Hines ha scritto:
Paolo? I see that some of the block-migration code and the qemu-img code is also calling buffer_is_zero() - are you guys depending on the performance of any buffer_is_zero() calls to use the vector-optimized version like migration does?
Yes, definitely. Not to the point that we care about which functions are inlined, but the constant-factor improvement that comes for vectorization can have a noticeable impact.
Paolo
