Is this issue only happening with 64 bit float processing? Or does it also occur with 32 bit float as well?