Re: [PATCH] Cache aligned locks for PPC64
Ulrich Drepper <[email protected]>
| Newsgroups | gmane.comp.lib.phil |
|---|---|
| Organization | Red Hat, Inc. |
| Message-ID | <[email protected]> |
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 Steve Munroe wrote: > I assume this means either not changing the alignment (and leaving big > performance on the table) or using symver and building 2 libraries? This is an issue for all architectures. And it is noting which the implementation has to solve itself. There are situations when you don't want or need the extra alignment and obviously there are some when it is useful. This means ensuring alignment is a higher-level decision. This is what optimizations guides are for. The situation when it is useful to have the extra alignment should be described and then it should show how the *variable* can be aligned. In case of arrays of futexes, explain how to create wrapper types with the appropriate alignment. - -- - --------------. ,-. 444 Castro Street Ulrich Drepper \ ,-----------------' \ Mountain View, CA 94041 USA Red Hat `--' drepper at redhat.com `--------------------------- -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.2.1 (GNU/Linux) iD8DBQE+wWxo2ijCOnn/RHQRAiP4AJ0WNGyGfa91ShjBtVeZLC1+Oah8/ACfUO4T s1twzoSpzidjDohVCr+xJfQ= =dY3a -----END PGP SIGNATURE-----