Re: [Ping][PATCH, Updated] Hookize FRAME_POINTER_REQUIRED

[Richard Henderson <rth at redhat dot com>]

On 06/21/2009 07:52 AM, Anatoly Sokolov wrote:
> Hello.
>
> This patch turns FRAME_POINTER_REQUIRED macro into a hook.
>
> The changes for the moxie target is added and the fr30 target is tested now.
>
> The patch has been bootstrapped on and regression tested on:
> armv5tel-unknown-linux-gnueabi
> i86_64-unknown-linux-gnu
> mipsel-unknown-linux-gnu
> hppa2.0-unknown-linux-gnu
> sparc64-unknown-linux-gnu
>
> Regression tested on:
> avr-unknown-elf
> bfin-unknown-elf
> cris-axis-elf
> fr30-unknown-elf
> frv-unknown-elf
> m32c-unknown-elf
> m32r-unknown-elf
> mmix-knuth-mmixware
> sh-unknown-elf
> xtensa
>
> I don't have the opportunity to test this patch on arc, crx, moxie, mcore,
> score and vax target.
>
> Ok for mainline?

Mostly ok.

(1) All of the functions that are just cfun->calls_alloca aren't needed.
    GCC automatically forces a frame pointer for alloca.

(2) The mcore definition of CAN_ELIMINATE (which you exported to a
    function) uses a version of FRAME_POINTER_REQUIRED that's always
    false.  This can obviously be simplified.

(3) Missed static at definition of avr_frame_pointer_required_p.

Ok with those changes.

r~