On Sun, Nov 27, 2016 at 11:28:45PM +0100, Marius Bakke wrote: > Marius Bakke <mba...@fastmail.com> writes: > > > Leo Famulari <l...@famulari.name> writes: > > > >> On Sun, Nov 27, 2016 at 10:02:35PM +0100, Tobias Geerinckx-Rice wrote: > >>> Leo, > >>> > >>> On 27/11/16 21:56, Leo Famulari wrote: > >>> > Will this work on nachines that don't support SSE? My understanding is > >>> > that we don't wish to require it. > >>> > >>> Why wouldn't it? > >>> > >>> I don't often use Icecat, but when I do... I don't see why it would > >>> require SSE by default. ;-) And it's not like we ever patched it to use > >>> IJG's libjpeg to begin with. > >> > >> I'm not sure, but the comment seems to indicate that we don't want to > >> use it: > >> > >> ;; According to > >> ;; http://sourceforge.net/projects/libjpeg-turbo/ , > >> ;; "libjpeg-turbo is a derivative of libjpeg that > >> ;; uses MMX, SSE, SSE2, and NEON SIMD instructions > >> ;; to accelerate baseline JPEG compression/ > >> ;; decompression", so we had better not use it > > > > SSE is pentium III, and SSE2 was introduced in Pentium IV. MMX is even > > older. Are we really committed to supporting 15+ year old hardware? > > Digging a bit further, SSE2 is actually part of the AMD64 > specification[0]. Perhaps we could make it conditional on architecture? > > 0: https://en.wikipedia.org/wiki/SSE2#CPU_support
Sure, if it requires SSE2 to run, and we keep the status quo regarding i686, I don't see why not.
signature.asc
Description: PGP signature
