[LLVMdev] Memset/memcpy: user control of loop-idiom recognizer

[Joerg Sonnenberger]

On Tue, Dec 02, 2014 at 09:37:42PM +0000, David Chisnall wrote:
> We can often generate a very efficient memcpy loop in the back end if
> we know that the data being copied is strongly aligned.

What surprises me is that a loop should ever result in more efficient
code given that it needs pretty much the same set of constraints.

Joerg