[CUDA] When compilation fails, print the compilation mode.

[jlebar]

Summary

That is, instead of "1 error generated", we now say "1 error generated
when compiling for sm_35".

This (partially) solves a usability foogtun wherein e.g. users call a
function that's only defined on sm_60 when compiling for sm_35, and they
get an unhelpful error message.

[tra]

Nice.

cfe/trunk/lib/Frontend/CompilerInstance.cpp

if (getDiagnosticOpts().ShowCarets) {
     unsigned NumErrors = getDiagnostics().getClient()->getNumErrors();
 
     if (NumWarnings)
       OS << NumWarnings << " warning" << (NumWarnings == 1 ? "" : "s");
     if (NumWarnings && NumErrors)
       OS << " and ";
     if (NumErrors)
       OS << NumErrors << " error" << (NumErrors == 1 ? "" : "s");
-    if (NumWarnings || NumErrors)
-      OS << " generated.\n";
+    if (NumWarnings || NumErrors) {
+      OS << " generated";
+      if (getLangOpts().CUDA) {
+        if (!getLangOpts().CUDAIsDevice) {
+          OS << " when compiling for host";
+        } else {
+          OS << " when compiling for " << getTargetOpts().CPU;
+        }
+      }
+      OS << ".\n";
+    }
   }
 
   if (getFrontendOpts().ShowStats) {
     if (hasFileManager()) {
       getFileManager().PrintStats();
       OS << '\n';
     }
     llvm::PrintStatistics(OS);

cfe/trunk/test/SemaCUDA/error-includes-mode.cu

+// RUN: not %clang_cc1 -fsyntax-only %s 2>&1 | FileCheck --check-prefix HOST %s
+// RUN: not %clang_cc1 -triple nvptx-unknown-unknown -target-cpu sm_35 \
+// RUN:   -fcuda-is-device -fsyntax-only %s 2>&1 | FileCheck --check-prefix SM35 %s
+
+// HOST: 1 error generated when compiling for host
+// SM35: 1 error generated when compiling for sm_35
+error;