On 7/7/23 12:08 AM, Kewen.Lin wrote:
> on 2023/7/7 07:00, Peter Bergner wrote:
>> On 7/6/23 5:54 PM, Peter Bergner wrote:
>>> On 6/30/23 7:58 PM, Carl Love via Gcc-patches wrote:
>>>> +++ b/gcc/testsuite/gcc.target/powerpc/test_fpscr_rn_builtin_2.c
>>>> @@ -0,0 +1,153 @@
>>>> +/* { dg-do run { target { powerpc*-*-* } } } */
>>>
>>> powerpc*-*-* is the default for this test directory, so you can drop that,
>>> but you need to disable this test for soft-float systems, so you probably
>>> want:
>>>
>>> /* { dg-do run { target powerpc_fprs } } */
>>
>> We actually want something like powerpc_fprs_hw, but that doesn't exist.
>>
>
> Yeah, good point! I noticed that we have a few test cases which need to
> check soft-float env as well but they don't, I didn't find any related
> issues have been reported, so I would assume that there are very few
> actual testings on this area. Based on this, I'm not sure if it's worthy
> to add a new effective target for it. Personally I'm happy with just using
> powerpc_fprs here to keep it simple. :)
I think powerpc_fprs_hw can be added later by someone if they care.
Using powerpc_fprs is an improvement over powerpc*-*-*, since it will
reduce some FAILs, just not all of them a powerpc_fprs_hw would.
I doubt many people are running the testsuite on real ppc hardware
that doesn't have an FP unit.
Peter
