[llvm-commits] [llvm] r48219 - /llvm/trunk/test/CodeGen/PowerPC/Frames-align.ll

[Dale Johannesen]

On Mar 11, 2008, at 10:58 AM, Evan Cheng wrote:

> So now the stack ptr isn't aligned on 16? What if AltiVec is used?
> Does the JIT callback routine makes any assumption about stack
> alignment (as x86)?
>
> Evan

The PPC stack is always 16-byte aligned.  Altivec works.  What doesn't
work is alignments bigger than 16 bytes.  This cannot cause any  
misaligned
loads or stores, and is fairly hard to detect; you'd have to  
explicitly look at
the low bits of an object's address or something like that.
I haven't looked at the JIT but if it follows the standard calling  
convention
there shouldn't be a problem.

> On Mar 11, 2008, at 10:35 AM, Chris Lattner <clattner at apple.com>  
> wrote:
>
>>
>> On Mar 11, 2008, at 10:05 AM, Dale Johannesen wrote:
>>
>>>
>>> On Mar 11, 2008, at 9:30 AM, Chris Lattner wrote:
>>>
>>>> On Mar 11, 2008, at 12:51 AM, Evan Cheng wrote:
>>>>> URL: http://llvm.org/viewvc/llvm-project?rev=48219&view=rev
>>>>> Log:
>>>>> Learn how to xfail a test.
>>>>
>>>> Dale, is this test testing the wrong values?  If so, please update
>>>> and
>>>> unxfail it.
>>>>
>>>> -Chris
>>>
>>> I disabled this feature because it doesn't work in the general case,
>>> isn't trivial to fix, and isn't supported by gcc.
>>> Xfailing the test looks OK to me.
>>
>> An xfailed test says that there is something that was working that  
>> has
>> regressed.  If we don't plan to implement this feature or the test  
>> was
>> wrong initially, please just remove it.
>>
>> -Chris
>> _______________________________________________
>> llvm-commits mailing list
>> llvm-commits at cs.uiuc.edu
>> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits
> _______________________________________________
> llvm-commits mailing list
> llvm-commits at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits