Re: [Libreoffice] "bad value (atom) for -mtune= switch" on boostrap build

On Sun, 2010-12-05 at 08:06 +0900, Takeshi Abe wrote:
> > apparently  -mtune=atom require gcc 4.5. I think that commit need to
> Thanks for that information. I became eager for gcc 4.5 :)

        Lol :-) the background here is (AFAIR) that in the Atom - in order to
get the die area down to some tiny 5x5mm square - and the power right
down, Intel binned some huge chunk of logic to do all but the most
simple instruction re-ordering and dropped speculative execution and
register renaming.

        As such, the same code compiled in a slightly different order can be
substantially faster on Atom than the same instructions in a different
order on the same Atom :-)

        Of course, for those of us with non-Atom CPUs, the change is (most
likely) negligable: we already had all those shuffling transistors
sitting around anyway.

        OTOH, gcc 4.5 is no doubt better in lots of other ways ;-)

        HTH,

                Michael.

-- 
 michael.meeks@novell.com  <><, Pseudo Engineer, itinerant idiot