[IAEP] [Sugar-devel] Planning for the future (Samuel Greenfeld)

[James Cameron]

On Thu, Feb 26, 2015 at 03:31:50PM +1100, James Cameron wrote:
> On Wed, Feb 25, 2015 at 04:20:02PM -0600, Jerry Vonau wrote:
> > > On February 25, 2015 at 3:09 PM James Cameron <quozl at laptop.org> wrote:
> > > On Wed, Feb 25, 2015 at 01:20:19PM -0600, Jerry Vonau wrote:
> > > > I know this is not a sugar issue directly, more of an OLPC issue
> > > > but since Fedora F12 the entire i686 platform's userland is
> > > > being compiled with -mtune=atom which would use sse. This causes
> > > > problems for some parts of sugar now that java is being used
> > > > more and the XO-1 lacks sse.  Fixing one package that uses sse
> > > > might fix one issue but this is really a distro wide setting and
> > > > other issues may float to the top in other areas.
> > >
> > > Thanks, wasn't aware -mtune=atom was being used upstream.  It
> > > explains a lot.  First build after Fedora 11 was 11.2.0 (os874)
> > > using Fedora 14.  So if we rebuild everything there may be an
> > > improvement?  That's probably something that can be set running as
> > > a test.
> > >
> > 
> > Wouldn't all the rpms used need to be recompiled to ensure mtune is
> > set to match throughout the distro?
> 
> Don't think so.  Check my logic:
> 
> The GCC documentation you referenced described -mtune as "Tune to
> cpu-type everything applicable about the generated code, except for
> the ABI and the set of available instructions. "
> 
> -march is more significant, as "Generate instructions for the machine
> type cpu-type. The choices for cpu-type are the same as for
> -mtune. Moreover, specifying -march=cpu-type implies
> -mtune=cpu-type. "
> 
> If the ABI were different between i586 and i686 arch, that would be
> very interesting.
> 
> > Tall order IMHO, good luck
> 
> ;-)
> 
> For the moment, I'm doing a mock --rebuild of webkitgtk3 with
> --arch=i586, and the logs so far show "-march=i586 -mtune=generic"
> instead of "-march=i686 -mtune=atom":

This didn't change the problem, gdb core still showed SSE instructions
used.

Daniel Drake's change to WebKit that fixed this before has since been
lost in the current WebKit sources in git.  Patch is in the history,
but some later patch removed the change.

> 
> $ grep mtune build.log | grep i586 | wc --lines
> 8564
> $ grep mtune build.log | grep atom | wc --lines
> 0
> $ 
> 
> > Jerry
> 
> -- 
> James Cameron
> http://quozl.linux.org.au/

-- 
James Cameron
http://quozl.linux.org.au/