This is the mail archive of the
libc-alpha@sourceware.org
mailing list for the glibc project.
Re: [Patch soft-fp] Add support for various half-precision conversion routines
On Wed, Sep 07, 2016 at 04:35:25PM +0000, Joseph Myers wrote:
> On Wed, 7 Sep 2016, James Greenhalgh wrote:
>
> > 2016-09-07 James Greenhalgh <james.greenhalgh@arm.com>
> >
> > * soft-fp/extendhftf2.c: New.
> > * soft-fp/fixhfti.c: Likewise.
> > * soft-fp/fixunshfti.c: Likewise.
> > * soft-fp/floattihf.c: Likewise.
> > * soft-fp/floatuntihf.c: Likewise.
> > * soft-fp/half.h: Likewise.
>
> This ChangeLog entry is missing trunctfhf2.c.
>
> Your patch description refers to conversions to and from DImode, but they
> aren't in the patch. One might also wonder whether a direct truncation
> from DFmode is relevant to avoid double rounding (and for that matter a
> conversion from SImode, to avoid going via DFmode; going via SFmode would
> involve double rounding).
Hi Joseph,
An initial draft of this patch did implement these conversions. I thought
I had updated the ChangeLog, but clearly didn't. I pulled them out to get
the minimum number of new functions I needed for AArch64 support. After
https://gcc.gnu.org/ml/gcc-patches/2016-09/msg00704.html is approved,
the DI to HF conversion routines will not require a libgcc call (for AArch64).
Likewise DF to HF is handled by GCC proper.
However, as you've observed I will need to add more of these when I get on
to enabling _Float16 for AArch32, so if proposing them now would be prefered,
I'm happy to do that.
Thanks,
James