Re: FP compares and TARGET_SSE_MATH

[Jan Hubicka <hubicka at ucw dot cz>]

> Hello!
> 
> I would like to ask about FP compares and TARGET_SSE_MATH. Currently, 
> SSE compares are enabled for TARGET_SSE, but user has no choice on which 
> instruction set to use. In pentium4 case, resulting ASM code is a 
> mixture of i387 compares, SSE compares and a lot of register moving to 
> satisfy register constraints of used insns. I suggest to limit SSE 
> compares to TARGET_SSE_MATH.
> 
> By limiting SSE compares to TARGET_SSE_MATH, two insn sets would be 
> separated. If someone still wants to use mixed insn set, one can use 
> TARGET_MIX_SSE_I387 switch. Similar reasoning applies to FP conditional 
> move patterns and min/max patterns.

Do you have any benchmarks that suggest that avoiding the mixture of
both is a lost?  I briefly benchmarked this on SPECfp at the time I was
implementing this and using both sets was a win that time, but times
might've changed.

I also sent patch to teach regclass to discover the dependencies (ie to
avoid putting register X in x87 when it is used in comparsion operator
with register Y that needs to live in SSE).  perhaps we might thing
about sollution in this dirrection. (I originally gave up mostly because
new-RA seemed to make progress)

Honza
> 
> Uros.
>