[llvm-commits] [llvm] r67849 - /llvm/trunk/test/CodeGen/PowerPC/int-fp-conv-0.ll

[Dan Gohman]

On Mar 31, 2009, at 10:52 PM, Chris Lattner wrote:

>
> On Mar 31, 2009, at 8:48 PM, Nick Lewycky wrote:
>
>>>>> =
>>>>> =
>>>>> =
>>>>> ==================================================================
>>>>> --- llvm/trunk/test/CodeGen/PowerPC/int-fp-conv-0.ll (original)
>>>>> +++ llvm/trunk/test/CodeGen/PowerPC/int-fp-conv-0.ll Fri Mar 27
>>>>> 11:17:22 2009
>>>>> @@ -1,4 +1,4 @@
>>>>> -; RUN: llvm-as < %s | llc -march=ppc64 -debug |& not grep store
>>>>> +; RUN: llvm-as < %s | llc -march=ppc64 -debug |& not grep {=
>>>>> store}
>>>> I thought that "store" instructions were never assigned to a
>>>> variable...
>>>
>>> This test is grepping through the -debug output of llc, and the
>>> stores
>>> its
>>> looking for do happen to be printed with =.
>>>
>>> Grepping through -debug output isn't very nice in any case, but at
>>> the
>>> moment there aren't any better ways to express this specific test.
>>
>> The issue with using -debug in a test like this is that the user  
>> might
>> have compiled a release build. I don't know about llc, but opt -
>> debug is
>> disabled and llc -debug might be in the future.
>
> Right, -debug in general doesn't exist in -asserts mode.
>
>> Perhaps it's time to consider a unit test? Is this something you  
>> could
>> test through the API?
>
>
> Does anyone know what this testing is attempting to test?  If not,
> just remove that RUN line.

It's testing that the SelectionDAG phase doesn't introduce a store.  
There
isn't currently a greppable display for the SelectionDAG, though it
ought to be possible to create one, with something like DumpNodes.

Dan