This is the mail archive of the gcc-patches@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: [Patch, testsuite, arm] Skip the arch conflict to enable case neon-thumb2-move pass on more targets.


On 9 Sep 2011, at 14:03, "Richard Guenther" <richard.guenther@gmail.com> wrote:

> On Fri, Sep 9, 2011 at 2:47 PM, Richard Earnshaw <rearnsha@arm.com> wrote:
>> On 09/09/11 11:07, Richard Guenther wrote:
>>> On Fri, Sep 9, 2011 at 12:00 PM, Terry Guo <terry.guo@arm.com> wrote:
>>>> Hello,
>>>> 
>>>> This patch enables the case pass on targets other than armv7-a by skipping
>>>> the architecture conflict message. Is it OK to trunk?
>>> 
>>> I don't like these kind of patches.  Why do the testcases have -march=armv7-a
>>> enabled in the first place?  If they really need it then they shouldn't be
>>> run with another -march and thus instead there should be a
>>> dg-requires-effective-target armv7-a.  If they don't really need it
>>> they shouldn't
>>> add it.
>>> 
>> 
>> I'm not particularly keen on them either, but with the current structure
>> of the testsuite framework the only way to test some unusual option
>> combinations is to force them (we can't test every option set, or even
>> every CPU supported via multilib testing as it would take ~forever to
>> run through all the variants).  So the only alternative is to force some
>> tests to run with specific options.
>> 
>> It might be easier if there were a specific part of the testsuite which
>> was designated to be non-executable, non-multilib.  That part would then
>> be used without all the multilib paraphernalia that leads to these hacks
>> (and would only ever be run once as well, regardless of the number of
>> multilibs being tested).  But until such time, this patch seems like the
>> only way forward.
> 
> I suppose you want a torture that excercises different -march/-mtune
> combinations then.
> 
> But can't you do the pruning somewhere in an .exp file then instead
> of sprinkling it all over the tests itself?
> 

It seems not.  At present the multilib options are all added automatically by the dejagnu infrastructure.

R.

> Richard.
> 
>> R.
>> 
>>> Richard.
>>> 
>>>> BR,
>>>> Terry
>>>> 
>>>> 2011-09-09  Terry Guo  <terry.guo@arm.com>
>>>> 
>>>>          * gcc.target/arm/neon-thumb2-move.c: Skip the
>>>>        architecture conflict to enable the case pass
>>>>        on targets other than armv7-a.
>>>> 
>>>> diff --git a/gcc/testsuite/gcc.target/arm/neon-thumb2-move.c
>>>> b/gcc/testsuite/gcc.target/arm/neon-thumb2-move.c
>>>> index 430a4d5..9cf86dd 100644
>>>> --- a/gcc/testsuite/gcc.target/arm/neon-thumb2-move.c
>>>> +++ b/gcc/testsuite/gcc.target/arm/neon-thumb2-move.c
>>>> @@ -3,6 +3,7 @@
>>>>  /* { dg-require-effective-target arm_thumb2_ok } */
>>>>  /* { dg-options "-O2 -mthumb -march=armv7-a" } */
>>>>  /* { dg-add-options arm_neon } */
>>>> +/* { dg-prune-output "switch .* conflicts with" } */
>>>> 
>>>>  #include <arm_neon.h>
>>>>  #include <stddef.h>
>>>> 
>>>> 
>>>> 
>>>> 
>>>> 
>>>> 
>>> 
>> 
>> 
> 


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]