On Mon, 7 Dec 2020 at 16:34, Kyrylo Tkachov <kyrylo.tkac...@arm.com> wrote:
>
> Hi Prathamesh,
>
> > -----Original Message-----
> > From: Prathamesh Kulkarni <prathamesh.kulka...@linaro.org>
> > Sent: 07 December 2020 11:01
> > To: Kyrylo Tkachov <kyrylo.tkac...@arm.com>
> > Cc: gcc-patches@gcc.gnu.org
> > Subject: Re: [PR66791][ARM] Replace calls to __builtin_neon_vmvn* by ~ for
> > vmvn intrinsics
> >
> > On Thu, 3 Dec 2020 at 16:05, Kyrylo Tkachov <kyrylo.tkac...@arm.com>
> > wrote:
> > >
> > >
> > >
> > > > -----Original Message-----
> > > > From: Gcc-patches <gcc-patches-boun...@gcc.gnu.org> On Behalf Of
> > > > Prathamesh Kulkarni via Gcc-patches
> > > > Sent: 03 December 2020 10:30
> > > > To: gcc Patches <gcc-patches@gcc.gnu.org>; Kyrill Tkachov
> > > > <kyrylo.tkac...@foss.arm.com>
> > > > Subject: Re: [PR66791][ARM] Replace calls to __builtin_neon_vmvn* by ~
> > for
> > > > vmvn intrinsics
> > > >
> > > > On Wed, 25 Nov 2020 at 22:01, Prathamesh Kulkarni
> > > > <prathamesh.kulka...@linaro.org> wrote:
> > > > >
> > > > > Hi,
> > > > > This patch replaces calls to __builtin_neon_vmvnv* builtins with ~
> > > > > operator in arm_neon.h.
> > > > > Cross-tested on arm*-*-*.
> > > > > OK to commit ?
> > >
> > > Ok.
> > Hi Kyrill,
> > I have attached an updated patch that removes entry for vmvn from
> > arm_neon_builtins.def,
> > but I am not sure if that's sufficient ?
> >
> > I tried this:
> > __extension__ extern __inline uint32x2_t
> > __attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
> > vmvn_u32 (uint32x2_t __a)
> > {
> > return (uint32x4_t)__builtin_neon_vmvnv4si ((int32x4_t) __a);
> > // return ~__a;
> > }
> >
> > after removing the entry for vmvn, from arm_neon_builtins.def, but
> > compilation appeared to proceed successfully,
> > so I guess it hasn't stopped the builtin from being created ?
> >
>
> I think you'll want to look at the generated assembly?
> IIRC if the builtin isn't created then the assembly will contain a call to
> __builtin_neon_vmvnv4si, which of course will fail at link-time.
Ah, indeed. When I try to call it, it now results in:
vmvn.c:6:22: warning: implicit declaration of function
‘__builtin_neon_vmvnv2si’; did you mean ‘__builtin_neon_vmlav2si’?
[-Wimplicit-function-declaration]
6 | return (uint32x2_t)__builtin_neon_vmvnv2si ((int32x2_t) __a);
| ^~~~~~~~~~~~~~~~~~~~~~~
| __builtin_neon_vmlav2si
vmvn.c:6:3: error: cannot convert a value of type ‘int’ to vector type
‘__simd64_uint32_t’ which has different size
6 | return (uint32x2_t)__builtin_neon_vmvnv2si ((int32x2_t) __a);
| ^~~~~~
so I suppose it has correctly removed it.
Is it OK to commit the patch ?
Thanks,
Prathamesh
>
> Thanks,
> Kyrill
>
> > Thanks,
> > Prathamesh
> > > Sorry for missing this.
> > > Thanks,
> > > Kyrill
> > >
> > > P.S. please use my @arm.com address when CC'ing me, I don't have access
> > to the foss.arm.com one currently...
> > >
> > > > ping https://gcc.gnu.org/pipermail/gcc-patches/2020-
> > > > November/560223.html
> > > >
> > > > Thanks,
> > > > Prathamesh
> > > > >
> > > > > Thanks,
> > > > > Prathamesh
