------- Comment #11 from law at redhat dot com 2010-01-14 17:29 ------- Subject: Re: vectorizer created unaligned vector insns
On 01/13/10 02:35, irar at il dot ibm dot com wrote: > ------- Comment #10 from irar at il dot ibm dot com 2010-01-13 09:35 ------- > Yes, I understand that we can't assume that an access is aligned if we can't > prove it's aligned. I don't understand how we can prove that a COMPONENT_REF > is > aligned, i.e., if there is a way to check if a struct is packed, or we'd > better > decide that we always use versioning for COMPONENT_REFs? > Given a raw COMPONENT_REF I don't think you can prove proper alignment. You have to look at the entire effective address computation. Unfortunately, I think this has rather significant implications on how often we can vectorize without needing to perform the runtime check for alignment. Jeff -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=42652
