Re: PR target/52555: attribute optimize is overriding command line options

[Aldy Hernandez <aldyh at redhat dot com>]

On 02/16/13 05:19, Richard Sandiford wrote:

> Looks good to me otherwise, thanks.

Implemented all your suggestions.

> Sorry that SWITCHABLE_TARGETS has been so much hassle.  TBH, like Jakub
> says in the PR, I was hoping things like the optimize attribute could
> use the target_globals stuff too.  Today we just care about optabs,
> but e.g. arm.c has:
>
>    if (TARGET_THUMB1 && optimize_size)
>      {
>        /* When optimizing for size on Thumb-1, it's better not
>          to use the HI regs, because of the overhead of
>          stacking them.  */
>        for (regno = FIRST_HI_REGNUM;
> 	   regno <= LAST_HI_REGNUM; ++regno)
> 	fixed_regs[regno] = call_used_regs[regno] = 1;
>      }
>
> which affects other cached global state like IRA tables.

...
...

> So in the end I think we'll end up trying solve the same problem that
> the SWITCHABLE_TARGETS stuff was trying to solve, but this time for
> __attribute__((optimize)).

Ughhh... perhaps this can be something for 4.9, reimplementing the 
optimize attribute with the target_globals infrastructure?

Regstrapping attached patch.

OK pending tests?

Attachment: curr
Description: Text document