[llvm-bugs] [Bug 46922] New: Wrongsided function picked for overload when valid template function exists.
via llvm-bugs
llvm-bugs at lists.llvm.org
Thu Jul 30 15:37:47 PDT 2020
https://bugs.llvm.org/show_bug.cgi?id=46922
Bug ID: 46922
Summary: Wrongsided function picked for overload when valid
template function exists.
Product: clang
Version: unspecified
Hardware: PC
OS: Linux
Status: NEW
Keywords: compile-fail, regression
Severity: normal
Priority: P
Component: CUDA
Assignee: unassignedclangbugs at nondot.org
Reporter: canlewi at sandia.gov
CC: llvm-bugs at lists.llvm.org
The following code doesn’t compile with newer versions of clang:
template<class T>
__device__ __host__ int foo(T *x) {
return 1;
}
__device__ int foo(int *x) {
return 2;
}
__host__ int foo(long *x) {
return 3;
}
__device__ __host__ int bar() {
auto long_val = 1l;
return foo(&long_val);
}
clang++ -O2 -g -x cuda --cuda-gpu-arch=sm_61 -std=c++14 -o main -c main.cpp
give me:
error: reference to __host__ function 'foo' in __host__ __device__ function
return foo(&long_val);
^
main.cpp:10:14: note: 'foo' declared here
__host__ int foo(long *x) {
I believe that the issue is at
https://github.com/llvm/llvm-project/blob/8224c5047e9cef2db4b0e31427cdf90a2568a341/clang/lib/Sema/SemaOverload.cpp#L9860
It’s possible that IdentifyCUDAPreference will return CFP_HostDevice for valid
overloads, but this code doesn’t erase the wrong side candidates in that case.
Then best overload selection picks the wrong side candidate.
If I rewrite those lines as:
bool ContainsSameSideCandidate =
llvm::any_of(Candidates, [&](OverloadCandidate *Cand) {
// Check viable function only.
if (Cand->Viable && Cand->Function) {
auto MatchType = S.IdentifyCUDAPreference(Caller, Cand->Function);
return MatchType == Sema::CFP_HostDevice ||
MatchType == Sema::CFP_SameSide;
}
return false;
});
My code compiles again.
--
You are receiving this mail because:
You are on the CC list for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-bugs/attachments/20200730/953fa7c7/attachment.html>
More information about the llvm-bugs
mailing list