[PATCH] D26014: [ConstantFold] Get the correct vector type when folding a getelementptr instruction with vector indices

[Davide Italiano via llvm-commits]

On Wed, Oct 26, 2016 at 3:56 PM, Davide Italiano via llvm-commits
<llvm-commits at lists.llvm.org> wrote:
> davide added inline comments.
>
>
> ================
> Comment at: lib/IR/ConstantFold.cpp:2063-2070
>      if (VectorType *VT = dyn_cast<VectorType>(C->getType()))
>        GEPTy = VectorType::get(GEPTy, VT->getNumElements());
> +    // When the getelementptr has one of more vector indices, it will
> +    // return a vector of pointers. We guarantee that all the vectors
> +    // in this case will have the same width, so we can just look at
> +    // the first one.
> +    else if (VectorType *VT = dyn_cast<VectorType>(Idxs[0]->getType()))
> ----------------
> majnemer wrote:
>> Could this be replaced with `GetElementPtrInst::getGEPReturnType` ?
> hmm, I don't see an overload of `getGEPReturnType` which takes an `ArrayRef` of `Constant` but only `Value` -- do you want me to add one?
>
>
> https://reviews.llvm.org/D26014
>
>
>
> _______________________________________________
> llvm-commits mailing list
> llvm-commits at lists.llvm.org
> http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-commits

Michael found another case where vector GEP break.


+%struct.foo = type { [0 x i64] }
+
+define <8 x i64*> @patatino3(%struct.foo*) {
+  %2 = getelementptr inbounds %struct.foo, %struct.foo* undef, i64 0,
i32 0, <8 x i64> undef
+  ret <8 x i64*> %2
+}

It fails with a different assertion, I'd rather fix it in a separate
patch if there are no objections.

-- 
Davide

"There are no solved problems; there are only problems that are more
or less solved" -- Henri Poincare