tra added inline comments.
================
Comment at: test/SemaCUDA/call-host-fn-from-device.cu:88
__host__ __device__ void class_specific_delete(T *t, U *u) {
- delete t; // ok, call sized device delete even though host has preferable
non-sized version
+ delete t; // expected-error {{reference to __host__ function 'operator
delete' in __host__ __device__ function}}
delete u; // ok, call non-sized HD delete rather than sized D delete
----------------
The C++ magic is way above my paygrade, but as far as CUDA goes this is a
regression, compared to what nvcc does. This code in NVCC produced a warning
and clang should not error out at this point in time either as
it's not an error to call a host function from HD unless we use HD in a host
function, and we would not know how it's used until later. I think the error
should be postponed until codegen.
Repository:
rC Clang
https://reviews.llvm.org/D47757
_______________________________________________
cfe-commits mailing list
cfe-commits@lists.llvm.org
http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
