Re: mainline performance regression (observered on SPARC)

[Andrey Petrov <petrov at netbsd dot org>]

On Fri, May 24, 2002 at 01:53:33PM -0700, Dan Nicolaescu wrote:
> Andrey Petrov <petrov@netbsd.org> writes:
> 
>   > On Fri, May 24, 2002 at 11:36:48AM -0700, Dan Nicolaescu wrote:
>   > > 
>   > > Any idea what is the cause of this? 
>   > > 
>   > > The code is kind of silly, it moves stuff around between the integer
>   > > and fp registers, probably an artifact of the SPARC calling
>   > > conventions (passing doubles in the integer registers). But shouldn't
>   > > inlining get rid of this? 
>   > > 
>   > 
>   > This made me re-read sparc v9 abi, and it says that floating-point
>   > arguments and return values are passed in fpu registers. But some
>   > of the fpu registers are volatile across functions calls and some
>   > are not. I guess that you see saving/restoring of non-volatile fpu
>   > regs.
> 
> -mcpu=ultrasparc is v8plus not v9

Then disregard what I just wrote as off-topic.