Date: Tue, 10 Sep 2013 14:21:30 +0300 From: Konstantin Belousov <kostikbel@gmail.com> To: John-Mark Gurney <jmg@funkthat.com> Cc: svn-src-head@FreeBSD.org, svn-src-all@FreeBSD.org, src-committers@FreeBSD.org, re@FreeBSD.org, David O'Brien <obrien@FreeBSD.org> Subject: Re: svn commit: r255187 - in head/sys: conf crypto/aesni modules/aesni Message-ID: <20130910112130.GV41229@kib.kiev.ua> In-Reply-To: <20130910064149.GS68682@funkthat.com> References: <201309031831.r83IVNkh026523@svn.freebsd.org> <20130910060132.GA61541@dragon.NUXI.org> <20130910064149.GS68682@funkthat.com>
next in thread | previous in thread | raw e-mail | index | archive | help
[-- Attachment #1 --] On Mon, Sep 09, 2013 at 11:41:49PM -0700, John-Mark Gurney wrote: > David O'Brien wrote this message on Mon, Sep 09, 2013 at 23:01 -0700: > > On Tue, Sep 03, 2013 at 06:31:23PM +0000, John-Mark Gurney wrote: > > > Log: > > > Use the fact that the AES-NI instructions can be pipelined to improve > > > performance... Use SSE2 instructions for calculating the XTS tweek > > > factor... Let the compiler do more work and handle register allocation > > > by using intrinsics, now only the key schedule is in assembly... > > > > Hi John-Mark, > > Unfortunately this does not work with /usr/bin/gcc (which I still use as > > /usr/bin/cc on this old IBM T60 laptop due to clang's enormous pestimation > > on 'make world' and build times. > > > > Please fix this ASAP or back it out out of 10-CURRENT as it does not work > > with /usr/bin/gcc: > > > > 1. /usr/bin/gcc cannot handle the "-maes" option. > > I committed r255440 to address this. > > Please back that out until you have an understand of what the real > problem is... I'm sad that the commit was approved w/o bothering to > attempt to figure out the problem... > > > 2. /usr/src/sys/modules/aesni/../../crypto/aesni/aesencdec.h:30:23: error: wmmintrin.h: No such file or directory > > and then a cascade of errors follow. > > Sounds like you don't have the latest in-tree gcc... I've been > compiling the AES module w/ the in tree gcc for a while.. This is > w/ the changes in r255185... > > > For now, I've sent a patch to re@ for approval to remove the 'aesni' > > module from the build if not using clang. But the LINT build is > > still broken with /usr/bin/gcc. > > I'll strongly object to this change... Sorry for my mistake, I agree with you. IMO this change should be reverted. > > > > Replace .byte hard coded instructions w/ the proper instructions now > > > that both clang and gcc support them... > > > > Is this out-of-tree latest GCC that supports this? > > Nope, in-tree... Sounds like you haven't installed the latest version > of gcc in tree... > > It could be a problem with a bad interaction w/ delete-old as glebius > added the wmmintrin.h header to the delete-old target in r255354... > > I'll take a closer look at this in the morning, but next time please > attempt to contact the author before taking such actions... > > -- > John-Mark Gurney Voice: +1 415 225 5579 > > "All that I will do, has been done, All that I have, has not." [-- Attachment #2 --] -----BEGIN PGP SIGNATURE----- Version: GnuPG v2.0.21 (FreeBSD) iQIcBAEBAgAGBQJSLwC5AAoJEJDCuSvBvK1BXvEP/RCyrvLik2bM441bbCnihGOk Ky3Uwm5wMClbvhtM0rmbr6dlARdI+LxuqRvVILnDpKpyEM3kgI3SG5lhqEemT87B XYWdrvKd+q1FyBWJYoiwwVaWnJDwuZVvgVdW8sFBIsio19s06Dc2DkPm11Mdpir0 tkQ2z49F20dhN6GgViufh6ImZqxdpND/urIATdxu3fOkAg8zPZshShgwq5Nkwv3b 6OP5iYi7IFyQDM8s5QmaiGT2eIR5irxTI07A2qcc4sJvBHzA7ZV7rmzumeY9mRUH a6MAmpseqf+ShaeZLXA2rxNRQz52Ne4G3tdklyN0JOifhAcoKE7Q2mQNsSVqrYPb NeED0i2LgNIeEarEqfyRrDmjaJyBzbGP1yO4spKzBKfDN6fx/o++tPrc5K6b1xPR W3gUOrBqUAZitEWeficafSsyb6VTN5QC4tm9W7js16j/zF0crEV3ThQ6Scdc6RzY R6M6CjEC+5IB6h/9PatW6U2Y5KHNlIYJOb2a/HXBAqsdQbA+6OyWYBS6ETYCiRgZ 23XABhEM+Pq/GAL2Iqwi6ZxakmzyWY7Rch3zveQkTeovE6dOuj5HN7pYKxLygKoZ +rxQ4LxMTHkMkE6bcWT43IJoZOuGre1+6mGS6v/dFLx+576L86qmmnQaUC0IM1nX ID/Lbding0Mg1FS9l6Mm =xcOe -----END PGP SIGNATURE-----
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20130910112130.GV41229>