On Sat, 5 Nov 2022 23:28:47 +0100
Tobias Burnus <tob...@codesourcery.com> wrote:

[no comment. I wonder why we malloc versus realloc in the first place,
i'd realloc always for starters. We end up calling into libc anyway, we
don't inline any of those calls and we suffer lib-boundary non-IPA
trouble either way, still, no? So why the conditional on our side?
valgrind certainly does not have a problem either way and i'd hope our
analyzer doesn't either. So what's that dance good for anyway? If you
did not remove it altogether anyway that is. ]

> PPS: I lost track of pending patches. Are they any which I should
> review?

You could poke honza in a while (which i will do if there is
interest) WRT
https://gcc.gnu.org/pipermail/gcc-patches/2022-November/605081.html
needed iff we want to have (a cleaned-up version, including
gfortran.texi hunk) what David Love asked for in
https://gcc.gnu.org/pipermail/fortran/2022-October/058395.html
and where Thomas König expressed possible interest as per
https://gcc.gnu.org/pipermail/fortran/2022-November/058441.html
to add support for attribute target_clones.
I tested only c,c++,fortran,lto, don't have cycles for all langs, so if
you do d,ada,m2,..

The other 2 trivial hunks stemming from looking around that
attribute were approved already by Jeff and Richi, i'll push them
sometimes next week or during next weekend.

But David Love raised a question about more of the attributes supported
by the C FE, not just target_clones.

What are your thoughts around those?

blindvt> Tobias___, hi. You might have seen that David Love asked for
blindvt> attribute target_clones. Are there other attributes that you
blindvt> think would be helpful to support?
blindvt> i think flatten would be potentially usable, at least for
blindvt> power-users.
blindvt> not sure about simd. Maybe that? Or something
blindvt> else too? WDYT?
blindvt> i'm willing to spend one or two evenings on
blindvt> these, if you folks thing they are worth adding...

Apart from that, folks will certainly point at other patches pending
review..

thanks and cheers,

Reply via email to