[llvm-commits] [llvm] r63772 - /llvm/trunk/test/Assembler/2007-04-15-BadIntrinsic.ll

Chris Lattner clattner at apple.com
Fri Feb 6 09:04:44 PST 2009


On Feb 6, 2009, at 8:32 AM, Duncan Sands wrote:

>>>> It will not match as an intrinsic in CodeGen and turn into a  
>>>> libcall.
>>>> An assert could presumably be added at that point.
>>>
>>> actually, shouldn't it be possible to catch it in the verifier?
>>> Presumably getIntrinsicID only returns a non-zero value if the
>>> call is to an intrinsic that is known by some part of the system...
>>
>> The problem is that your target may not be linked into (e.g.) "opt".
>
> That means that the intrinsic won't get the right attributes etc.
> I guess that's just life if opt doesn't know about your target?

Sure it will, it will pass it through unmolested.  Opt should not be  
creating target intrinsics that it doesn't know about.

-Chris



More information about the llvm-commits mailing list