Hi~
On Tue, May 19, 2020 at 03:20:37PM -0700, Douglas B Rupp wrote:
> On 5/19/20 3:03 PM, Segher Boessenkool wrote:
> >This is a compile test, so it does not matter at all what the kernel is
> >doing or not doing. And the -mcpu=power8 flag should enable altivec.
> >Apparently it doesn't for you, but why is that?
>
> It's likely the specs file we use for the kernel we are using for
> testing with.
>
> %{mcpu=604:%{!maltivec:%{!mno-altivec:-mno-altivec}}}
> %{mcpu=604:%{!mvsx:%{!mno-vsx:-mno-vsx}}}
> %{!mcpu=*:%{!maltivec:%{!mno-altivec:-mno-altivec}}}
> %{!mcpu=*:%{!mvsx:%{!mno-vsx:-mno-vsx}}}
>
> and
>
> '--with-cpu=604'
>
> >This shouldn't be fixed by whack-a-mole on individual testcases.
> >
> Perhaps a SUBTARGET_OVERRIDE incantation might work? Any ideas or
> advice you might have would be welcome.
[ pasting threads together: ]
> Apologies on this test, the problem is -mdejagnu-cpu doesn't get matched
> by the spec logic
>
> If I pass -mcpu=power8, then indeed the test passes fine.
>
> I'm not sure how to fix this cleanly.
So, which spec are those 4 lines above from?
We have as DRIVER_SELF_SPECS (see rs6000.h)
#define DRIVER_SELF_SPECS \
"%{mdejagnu-cpu=*: %<mcpu=* -mcpu=%*}", \
"%{mdejagnu-tune=*: %<mtune=* -mtune=%*}", \
"%{mdejagnu-*: %<mdejagnu-*}", \
SUBTARGET_DRIVER_SELF_SPECS
so -mdejagnu-cpu= is translated to -mcpu= very early. But yours is
still done earlier?
It can of course be fixed by adding
%{mdejagnu-cpu=604:%{!maltivec:%{!mno-altivec:-mno-altivec}}}
%{mdejagnu-cpu=604:%{!mvsx:%{!mno-vsx:-mno-vsx}}}
%{!mdejagnu-cpu=*:%{!maltivec:%{!mno-altivec:-mno-altivec}}}
%{!mdejagnu-cpu=*:%{!mvsx:%{!mno-vsx:-mno-vsx}}}
to your specs, but that is band-aid on top of band-aid. Why do you want
to add -mno-altivec -mno-vsx whenever possible, anyway?
Segher
