libvpx

change to avoid IOC

SSE for a 64x64 block with 3 planes can go as high as 3*2^28. So left
shift by 4 may overflow 32 bit int.

Change-Id: I63c84aa56894788bb987299badabbd7cc6fd0be6