[llvm-commits] atomics memoperands patch

[Evan Cheng]

On Jun 20, 2008, at 1:16 PM, Chris Lattner wrote:

>>>>  ID.AddInteger(LD->getAddressingMode());
>>>>  ID.AddInteger(LD->getExtensionType());
>>>>  ID.AddInteger(LD->getMemoryVT().getRawBits());
>>>>  ID.AddInteger(LD->getAlignment());
>>>>  ID.AddInteger(LD->isVolatile());
>>>> } else if (const StoreSDNode *ST = dyn_cast<StoreSDNode>(N)) {
>>>>  ID.AddInteger(ST->getAddressingMode());
>>>>  ID.AddInteger(ST->isTruncatingStore());
>>>>  ID.AddInteger(ST->getMemoryVT().getRawBits());
>>>>  ID.AddInteger(ST->getAlignment());
>>>>  ID.AddInteger(ST->isVolatile());
>>>> }
>>>>
>>>> Do we need the if () else if? AddNodeIDNode has already added those
>>>> fields to the FoldingSetNodeId.
>>>>
>>>
>>> Looking at the code, it does seem wrong to add the fields twice.
>>
>> I think so too. Chris?
>
> Yeah you're right, that is bogus.  Please make sure that all the  
> places that are dealing with loads/stores are putting the same  
> number of values into the folding set though: if this is changed  
> other places may need changes as well.

Actually these are needed. AddNodeIDNode(ID, N->getOpcode(), N- 
 >getVTList(), Ops, NumOps); only add opcode, vtlist and operands to  
CSE id.

Evan

>
>
> Great catch,
>
> -Chris
>