AMDGPU: Add test for generic builtin behavior

[arsenm]

Summary

This is the same as the default implementation, but
add a comment and make sure the behavior is tested.

[• tstellarAMD]

lib/Basic/Targets.cpp

1854–1857

Why do we need to add this if it's the same as the default?

[arsenm]

lib/Basic/Targets.cpp

1854–1857

Don't really, other than it adds a comment. I can keep just the test

[• tstellarAMD]

lib/Basic/Targets.cpp

1854–1857

Ok, sounds good. LGTM.

[arsenm]

r266383

lib/Basic/Targets.cpp

bool initFeatureMap(llvm::StringMap<bool> &Features,
                       DiagnosticsEngine &Diags, StringRef CPU,
                       const std::vector<std::string> &FeatureVec) const override;
 
   ArrayRef<Builtin::Info> getTargetBuiltins() const override {
     return llvm::makeArrayRef(BuiltinInfo,
                         clang::AMDGPU::LastTSBuiltin - Builtin::FirstTSBuiltin);
   }
 
+  bool isCLZForZeroUndef() const override {
+    // It is -1 instead of expected for intrinsic.
+    return true;
+  }
+
   void getTargetDefines(const LangOptions &Opts,
                         MacroBuilder &Builder) const override {
     if (getTriple().getArch() == llvm::Triple::amdgcn)
       Builder.defineMacro("__AMDGCN__");
     else
       Builder.defineMacro("__R600__");
 
     if (hasFMAF)

test/CodeGenOpenCL/builtins-generic-amdgcn.cl

+// REQUIRES: amdgpu-registered-target
+// RUN: %clang_cc1 -triple amdgcn-unknown-unknown -S -emit-llvm -o - %s | FileCheck %s
+
+// CHECK-LABEL: @test_builtin_clz(
+// CHECK: tail call i32 @llvm.ctlz.i32(i32 %a, i1 true)
+void test_builtin_clz(global int* out, int a)
+{
+  *out = __builtin_clz(a);
+}
+
+// CHECK-LABEL: @test_builtin_clzl(
+// CHECK: tail call i64 @llvm.ctlz.i64(i64 %a, i1 true)
+void test_builtin_clzl(global long* out, long a)
+{
+  *out = __builtin_clzl(a);
+}