[PATCH] D12453: [CUDA] Allow function overloads based on host/device attributes.

[Eli Bendersky via cfe-commits]

eliben added inline comments.

================
Comment at: include/clang/Sema/Sema.h:8599
@@ +8598,3 @@
+    CFP_Never,      // Invalid caller/callee combination.
+    CFP_LastResort, // same as CFP_Never or CFP_Fallback, depending on
+                    // -fcuda-disable-target-call-checks option
----------------
The comment should explain exactly how it depends on the option

================
Comment at: lib/AST/ItaniumMangle.cpp:504
@@ +503,3 @@
+  if (FD->hasAttr<CUDADeviceAttr>() && FD->hasAttr<CUDAHostAttr>())
+    Out << "Ua6deviceUa4host";
+
----------------
Why device before host?

================
Comment at: lib/Sema/SemaCUDA.cpp:98
@@ +97,3 @@
+  CUDAFunctionTarget CallerTarget =
+                         Caller ? IdentifyCUDATarget(Caller) : Sema::CFT_Host,
+                     CalleeTarget = IdentifyCUDATarget(Callee);
----------------
Why allow Caller to be nullptr but not Callee? Also, != nullptr?

================
Comment at: lib/Sema/SemaExprCXX.cpp:2277
@@ +2276,3 @@
+        Matches.set_size(N);
+      } else
+        ++I;
----------------
Nit: wrap the ++I in braces too, otherwise this looks really weird


http://reviews.llvm.org/D12453