[llvm-commits] [llvm] r65289 - /llvm/trunk/lib/Transforms/Scalar/InstructionCombining.cpp

[Chris Lattner]

On Feb 23, 2009, at 6:18 PM, Dan Gohman wrote:

>
> On Feb 23, 2009, at 5:57 PM, Chris Lattner wrote:
>
>>
>> On Feb 23, 2009, at 4:29 PM, Dan Gohman wrote:
>>
>>>>> assigned a large alignment, which required codegen to perform
>>>>> dynamic stack-pointer re-alignment, which forced the enclosing
>>>>> function to have a frame pointer, which led to increased spilling.
>>>>
>>>> testcase?
>>>
>>>
>>> I can probably construct an artificial testcase if you think it'd  
>>> be useful,
>>> but I don't have anything handy smaller than lencod itself.
>>
>> I'm not interested in an executable testcase, an artificial one is  
>> good.  This is the sort of thing that is non-obvious and could  
>> break in the future, which is why I thought a testcase would be  
>> useful.  Do you agree?
>
> I added a testcase.

Thanks Dan!  To be clear (and not specifically directed at you), most  
regression tests are really best if they are "artificial" ones, that  
exercise only the thing in question.  IMO, checking in huge .ll files  
that were bugpoint reduced is far less useful than a minimal hand  
crafted testcase that exercises the behavior in question.

-Chris