Date: Wed, 15 Dec 2004 13:32:42 -0600 From: Kirk Strauser <kirk@strauser.com> To: freebsd-alpha@freebsd.org Subject: Re: Safe (but quick) GCC settings on a PC64 with 5.3? Message-ID: <200412151332.45800.kirk@strauser.com> In-Reply-To: <20041213205328.GA1546@dragon.nuxi.com> References: <200412131130.59807.kirk@strauser.com> <83229A4B-4D2E-11D9-9C15-000D93C47836@xcllnt.net> <20041213205328.GA1546@dragon.nuxi.com>
next in thread | previous in thread | raw e-mail | index | archive | help
--nextPart1732427.TU33l7dpZO Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable Content-Disposition: inline On Monday 13 December 2004 14:53, David O'Brien wrote: > I would use "-O2 -fno-strict-aliasing" otherwise there can be problems > with ports. Thanks for the tip. Would that apply to world and kernel as well? I tried= =20 building the OpenSSL library without "-fno-strict-aliasing" and got a few=20 type-punning warnings, but it seemed to run correctly. The GCC manual says that flag disables some optimizations. Using "openssl= =20 speed" as a rough benchmark didn't show a statistically significant=20 difference between using it or not; is anyone aware of any circumstances=20 where using it needlessly would invoke a significant penalty? By the way, my initial benchmark showed about a 50% speed increase in=20 OpenSSL by using "-O2" instead of "-O". Assuming everything continues to=20 work correctly, I'm very pleased about how this is turning out. =2D-=20 Kirk Strauser --nextPart1732427.TU33l7dpZO Content-Type: application/pgp-signature -----BEGIN PGP SIGNATURE----- iD8DBQBBwJFd5sRg+Y0CpvERAtOoAJwOsRscOIj4mT0G08FpyTOJlcVZyACfXUBY /1NofHwEUr+szM3d5Tpm3bw= =ck+6 -----END PGP SIGNATURE----- --nextPart1732427.TU33l7dpZO--
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?200412151332.45800.kirk>