No, that's silly. You're not changing the number of I/O operations at all. You're simply moving the location in code where they are done from the kernel's filesystem to the applications's userspace. There's no reason to expect either to be faster by anything other than a constant factor.