[llvm-commits] [llvm] r91497 - in /llvm/trunk: include/llvm/Metadata.h lib/VMCore/Metadata.cpp
Devang Patel
dpatel at apple.com
Mon Jan 4 08:33:43 PST 2010
On Dec 18, 2009, at 12:15 PM, Victor Hernandez wrote:
>
> On Dec 17, 2009, at 12:06 AM, Chris Lattner wrote:
>
>> On Dec 16, 2009, at 10:56 AM, Devang Patel wrote:
>>
>>>>>
>>>>> Is this new field copied by MetadataContextImpl::copyMD?
>>>>
>>>> The semantic is that if an MDNode is created function-local, then
>>>> it will continue to be function-local even if its operands are
>>>> modified to no longer refer to any function-specific IR.
>>>
>>> That'll break MDNode unique-ness.
>>
>> We don't guarantee MDNode uniqueness, we just endeavor to preserve
>> it when it isn't "too hard".
>
> Devang, you were concerned that 2 MDNodes with identical operands
> would not be the same if one is function-local and the other is not
> (can happen if an originally function-local MDNode has its operands
> modified to match those of a pre-existing non-function-local
> MDNode). Are there any issues with making these MDNodes be distinct?
Again, as Dan asked earlier, what's the use ? IMO, verifier needs are
not at all "too hard" in this case. As Chris agreed, we can have one
utility routine to satisfy verifier's needs. I do not see any need to
1) add Function * in MDNode, 2) or add a bit in MDNode, 3) or extend
MDNode class hierarchy, just for verifier. Let's focus on completing
@llvm.dbg.var support and we can revisit this when there is a real need.
-
Devan
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20100104/4a60ee59/attachment.html>
More information about the llvm-commits
mailing list