Bootstrap

[Zack Weinberg]

On Wed, Apr 05, 2000 at 10:26:46AM -0700, Mark Mitchell wrote:
> >>>>> "Jan" == Jan Hubicka <jhub6202@ss1000.ms.mff.cuni.cz> writes:
> 
>     Jan> On Sat, Mar 04, 2000 at 11:29:20AM -0800, Zack Weinberg
>     Jan> wrote:
>     >> On Sat, Mar 04, 2000 at 11:12:43AM -0800, Mark Mitchell wrote:
>     >> > > Jan --
> 
>     Jan> Mark, please don't use this email address, since it comes to
>     Jan> mbox I am not reading regularry and I've discovered your
> 
> OK.
> 
>     >>  I'm not at all convinced that the strlensi expander should
>     >> exist.  rep scasb is dead slow on modern CPUs and the alternate
>     >> is huge.  Under what conditions is it a performance win?
> 
>     Jan> It is probably still worthwhile as -Os optimization and in
>     Jan> case the alignment is known to be 4, otherwise the expander
>     Jan> fails anyway.  Once the code exist and is working, I think we
>     Jan> can keep it.
> 
> The words you qutoe aren't mine -- I never got into the `rep scasb'
> debate, I don't think. :-)

I said them.

Jan, what would you think about removing the code in i386.c to deal
with the unaligned non-scasb case, and never emitting scasb unless -Os
or the terminator is not '\0'?

zw