On Thu, 2013-01-17 at 08:27 +0100, Romain Francoise wrote: > The option text has now also been changed to "If unsure, say N"[1] > with the following rationale: > > | Sadly, -Os keeps on generating some very suboptimal code for certain > | cases, to the point where any I$ miss upside is swamped by the downside. > | The main ones are: > | > | - using "rep movsb" for memcpy, even on CPU's where that is > | horrendously bad for performance. > | > | - not honoring branch prediction information, so any I$ footprint you > | win from smaller code, you lose from less code density in the I$. > | > | - using divide instructions when that is very expensive. > > [1]: http://git.kernel.org/linus/3a55fb0d9fe8e2f4594329edd58c5fd6f35a99dd I'm happy to change this on x86 on trunk; does anyone have opinion about whether it should also be changed for other architectures? (I assume that for the flash-constrained armel flavours we'll have to keep this enabled, regardless of the optimisation possibilites.) Ben. -- Ben Hutchings All the simple programs have been written, and all the good names taken.
Attachment:
signature.asc
Description: This is a digitally signed message part