The 12/14/2017 20:46, Christophe Lyon wrote:
> On 14 December 2017 at 11:56, Tamar Christina <Tamar.Christina@arm.com> wrote:
> > The 12/13/2017 08:49, Christophe Lyon wrote:
> >> On 12 December 2017 at 18:29, Tamar Christina <tamar.christina@arm.com> wrote:
> >> > Hi All,
> >> >
> >> > The previous test made use of arm_neon.h which made the whole test
> >> > rather fragile and only applicable to some of the arm targets.
> >> >
> >> > So instead I make use of different fpus now to test the generation of
> >> > fmla instructions. The actual instruction itself is not tested as all
> >> > we care about if that the proper .fpu directives are generated.
> >> >
> >> > Regtested on arm-none-eabi and arm-none-linux-gnueabihf
> >> > with no regressions.
> >> >
> >> > Ok for trunk?
> >> >
> >> >
> >> > gcc/testsuite/
> >> > 2017-12-12  Tamar Christina  <tamar.christina@arm.com>
> >> >
> >> >         PR target/82641
> >> >         * gcc.target/arm/pragma_fpu_attribute.c: New.
> >> >         * gcc.target/arm/pragma_fpu_attribute_2.c: New.
> >
> > Hi Christophe,
> >
> > My apologies, I have rebased the patch.
> > New Changelog:
> >
> > gcc/testsuite/
> > 2017-12-14  Tamar Christina  <tamar.christina@arm.com>
> >
> >         PR target/82641
> >         * gcc.target/arm/pragma_fpu_attribute.c: Rewrite to use
> >         no NEON.
> >         * gcc.target/arm/pragma_fpu_attribute_2.c: Likewise.
> >
> 
> Hi,
> 
> Sorry I think there is still something wrong with this patch.
> In pragma_fpu_attribute_2.c, you are not removing
> #include <arm_neon.h>
> as the ChangeLog seems to imply?
>

Sorry that was extremely sloppy of me. I noticed the changelog after sending
but hadn't noticed the #include being left in.
 
> So, with this patch, there are problems on arm-none-linux-gnueabi and
> arm-none-eabi:
> FAIL:    gcc.target/arm/pragma_fpu_attribute.c scan-assembler-times
> \\.fpu\\s+vfpv3-d16 1 (found 0 times)
> FAIL:    gcc.target/arm/pragma_fpu_attribute.c scan-assembler-times
> \\.fpu\\s+vfpv4 1 (found 0 times)
> 
> and pragma_fpu_attribute_2.c still fails to compile:
> In file included from /gcc/testsuite/gcc.target/arm/pragma_fpu_attribute_2.c:6:
> /aci-gcc-fsf/builds/gcc-fsf-gccsrc/obj-arm-none-eabi/gcc3/gcc/include/arm_neon.h:31:2:
> error: #error "NEON intrinsics not available with the soft-float ABI.
> Please use -mfloat-abi=softfp or -mfloat-abi=hard"
> 
> I'm not sure why you don't see this when testing on arm-none-eabi?

It's because I don't have a compiler configured with only -mfloat-abi=soft. So when I run
the tests it's always able to just change the ABI. I resorted to manually testing it.

I've now prevented the tests from running at all on soft float only targets. This should fix
the problem once and for all.

Regtested on arm-none-eabi, arm-none-linux-gnueabi and arm-none-linux-gnueabihf.

Thanks and sorry for the noise,
Tamar

Ok for trunk?

gcc/testsuite/
2017-12-21  Tamar Christina  <tamar.christina@arm.com>

	PR target/82641
	* gcc.target/arm/pragma_fpu_attribute.c: Rewrite to use
	no NEON and require softfp or hard float-abi.
	* gcc.target/arm/pragma_fpu_attribute_2.c: Likewise.

> 
> If you want to see more details:
> http://people.linaro.org/~christophe.lyon/cross-validation/gcc-test-patches/255624-rb8655.patch-2/report-build-info.html
> (ignore the lines with "interrupted", this means there was a problem
> on the host during the build)
> 
> Christophe
> 
> 
> >> >
> >> Sorry, it seems your patch does not apply against ToT, and
> >> the ChangeLog looks incorrect (these are not new files)
> >>
> >> Christophe
> >
> > Thanks,
> > Tamar
> >
> > --

--