[tree-ssa] RFC: Never gimplify ASM_EXPRs

[Diego Novillo]

On Thu, 2003-06-05 at 11:35, Andrew Haley wrote:
> Diego Novillo writes:
>  > 
>  > Is this an invalid use of __asm__ then?
> 
> I think so.  The asm is asking for a mem operand for something that it
> never intends actually to access.  Better to pass the pointer in a
> register.
> 
In which case, I guess we'll use the patch you proposed for boehm-gc.

We never do anything significant with ASM_EXPRs.  At most we could
const-prop into ASM_INPUTS, but when we had this enabled, we were
getting ICEs during RTL expansion (execute/20020107-1.c).

I don't really have a strong opinion about this issue.  For me it would
be simpler if we just treat ASM_EXPRs as black boxes much like we treat
MD builtins.  If we don't, then we need a good way of dealing with MD
builtins as well.  I don't know if trying to deal with these nodes would
be worth the effort.


Diego.