Re: NetBSD 8.0, OpenSSL and Geode CPU

2018-08-01 Thread Emmanuel Dreyfus
On Wed, Aug 01, 2018 at 10:08:14AM +, Emmanuel Dreyfus wrote: > I would rather add it to CFLAGS, so that SSE is properly tested and > the optimized code not run if not available. For anyone reading this, the fix was commit: http://mail-index.netbsd.org/source-changes/2018/08/01/msg097216.html

Re: NetBSD 8.0, OpenSSL and Geode CPU

2018-08-01 Thread Emmanuel Dreyfus
On Wed, Aug 01, 2018 at 07:25:43AM +, Christos Zoulas wrote: > In article <1nste97.18w2sjvnut8xm%m...@netbsd.org>, > Thanks, if you remove that flag and regen, does it work? I would rather add it to CFLAGS, so that SSE is properly tested and the optimized code not run if not available. -- E

Re: NetBSD 8.0, OpenSSL and Geode CPU

2018-08-01 Thread Christos Zoulas
In article <1nste97.18w2sjvnut8xm%m...@netbsd.org>, Emmanuel Dreyfus wrote: >Christos Zoulas wrote: > >> Please let the OpenSSL folks know. > >Here it is: >https://github.com/openssl/openssl/pull/6828 > >The problem would be a screwed OpenSSL import on our own: the perl >scripts producing the ase

Re: NetBSD 8.0, OpenSSL and Geode CPU

2018-07-31 Thread Emmanuel Dreyfus
Christos Zoulas wrote: > Please let the OpenSSL folks know. Here it is: https://github.com/openssl/openssl/pull/6828 The problem would be a screwed OpenSSL import on our own: the perl scripts producing the asembly files were invoked in src/crypto/external/bsd/openssl/lib/libcrypto/arch/i386/Mak

Re: NetBSD 8.0, OpenSSL and Geode CPU

2018-07-31 Thread Emmanuel Dreyfus
On Tue, Jul 31, 2018 at 08:30:21AM +, Christos Zoulas wrote: > Please let the OpenSSL folks know. Now I thought a bit about it, I am now convinced the MMX optimizaton for GCM should be enabled if CPU has MMX and SSE. Or it should be modified to remove pinsrw, which requires SSE. -- Emmanuel

Re: NetBSD 8.0, OpenSSL and Geode CPU

2018-07-31 Thread Christos Zoulas
In article <1nsrghw.14uf3th1ee92jam%m...@netbsd.org>, Emmanuel Dreyfus wrote: [... stuff deleted ...] > >But looking up pinsrw on a search engine, all pages I find suggest is >should compe with SSE. Here is an example: >http://www.felixcloutier.com/x86/PINSRW.html > >So is this a bug in OpenSSL? I