[TLS on Darwin] use CXX_FAST_TLS calling convention for tls_init.

This makes sure we don't generate a lot of code to spill/reload
CSRs when calling tls_init from the access functions.

This helps performance when tls_init is not inlined into the access
functions.

llvm-svn: 263854

3 files changed, 22 insertions, 8 deletions

diff --git a/clang/lib/CodeGen/ItaniumCXXABI.cpp b/clang/lib/CodeGen/ItaniumCXXABI.cpp
index 5f72af83b12..ff0e972ba64 100644
--- a/clang/lib/CodeGen/ItaniumCXXABI.cpp
+++ b/clang/lib/CodeGen/ItaniumCXXABI.cpp
@@ -2237,6 +2237,11 @@ void ItaniumCXXABI::EmitThreadLocalInitFuncs(
     CodeGenFunction(CGM)
         .GenerateCXXGlobalInitFunc(InitFunc, CXXThreadLocalInits,
                                    Address(Guard, GuardAlign));
+    // On Darwin platforms, use CXX_FAST_TLS calling convention.
+    if (CGM.getTarget().getTriple().isOSDarwin()) {
+      InitFunc->setCallingConv(llvm::CallingConv::CXX_FAST_TLS);
+      InitFunc->addFnAttr(llvm::Attribute::NoUnwind);
+    }
   }
   for (const VarDecl *VD : CXXThreadLocals) {
     llvm::GlobalVariable *Var =
@@ -2286,8 +2291,11 @@ void ItaniumCXXABI::EmitThreadLocalInitFuncs(
     llvm::BasicBlock *Entry = llvm::BasicBlock::Create(Context, "", Wrapper);
     CGBuilderTy Builder(CGM, Entry);
     if (InitIsInitFunc) {
-      if (Init)
-        Builder.CreateCall(Init);
+      if (Init) {
+        llvm::CallInst *CallVal = Builder.CreateCall(Init);
+        if (isThreadWrapperReplaceable(VD, CGM))
+          CallVal->setCallingConv(llvm::CallingConv::CXX_FAST_TLS);
+      }
     } else {
       // Don't know whether we have an init function. Call it if it exists.
       llvm::Value *Have = Builder.CreateIsNotNull(Init);
diff --git a/clang/test/CodeGenCXX/cxx11-thread-local-reference.cpp b/clang/test/CodeGenCXX/cxx11-thread-local-reference.cpp
index 048a0bd45ad..4c1e5a70cba 100644
--- a/clang/test/CodeGenCXX/cxx11-thread-local-reference.cpp
+++ b/clang/test/CodeGenCXX/cxx11-thread-local-reference.cpp
@@ -23,11 +23,13 @@ int &g() { return r; }
 
 // LINUX: define weak_odr hidden i32* @_ZTW1r() [[ATTR0:#[0-9]+]] {
 // DARWIN: define cxx_fast_tlscc i32* @_ZTW1r() [[ATTR1:#[0-9]+]] {
-// CHECK: call void @_ZTH1r()
+// LINUX: call void @_ZTH1r()
+// DARWIN: call cxx_fast_tlscc void @_ZTH1r()
 // CHECK: load i32*, i32** @r, align 8
 // CHECK: ret i32* %{{.*}}
 
-// CHECK-LABEL: define internal void @__tls_init()
+// LINUX-LABEL: define internal void @__tls_init()
+// DARWIN-LABEL: define internal cxx_fast_tlscc void @__tls_init()
 // CHECK: call void @[[R_INIT]]()
 
 // LINUX: attributes [[ATTR0]] = { {{.*}}"target-features"{{.*}} }
diff --git a/clang/test/CodeGenCXX/cxx11-thread-local.cpp b/clang/test/CodeGenCXX/cxx11-thread-local.cpp
index 40f33e0848b..f465cbdeea8 100644
--- a/clang/test/CodeGenCXX/cxx11-thread-local.cpp
+++ b/clang/test/CodeGenCXX/cxx11-thread-local.cpp
@@ -122,7 +122,8 @@ int f() {
 
 // LINUX-LABEL: define weak_odr hidden i32* @_ZTWN1VIiE1mE()
 // DARWIN-LABEL: define weak_odr hidden cxx_fast_tlscc i32* @_ZTWN1VIiE1mE()
-// CHECK: call void @_ZTHN1VIiE1mE()
+// LINUX: call void @_ZTHN1VIiE1mE()
+// DARWIN: call cxx_fast_tlscc void @_ZTHN1VIiE1mE()
 // CHECK: ret i32* @_ZN1VIiE1mE
 
 
@@ -212,7 +213,8 @@ void set_anon_i() {
 
 // LIUNX: define weak_odr hidden i32* @_ZTW1a() {
 // DARWIN: define cxx_fast_tlscc i32* @_ZTW1a()
-// CHECK:   call void @_ZTH1a()
+// LINUX:   call void @_ZTH1a()
+// DARWIN: call cxx_fast_tlscc void @_ZTH1a()
 // CHECK:   ret i32* @a
 // CHECK: }
 
@@ -222,12 +224,14 @@ void set_anon_i() {
 
 // LINUX-LABEL: define internal i32* @_ZTWL1d()
 // DARWIN-LABEL: define internal cxx_fast_tlscc i32* @_ZTWL1d()
-// CHECK: call void @_ZTHL1d()
+// LINUX: call void @_ZTHL1d()
+// DARWIN: call cxx_fast_tlscc void @_ZTHL1d()
 // CHECK: ret i32* @_ZL1d
 
 // LINUX-LABEL: define weak_odr hidden i32* @_ZTWN1U1mE()
 // DARWIN-LABEL: define cxx_fast_tlscc i32* @_ZTWN1U1mE()
-// CHECK: call void @_ZTHN1U1mE()
+// LINUX: call void @_ZTHN1U1mE()
+// DARWIN: call cxx_fast_tlscc void @_ZTHN1U1mE()
 // CHECK: ret i32* @_ZN1U1mE
 
 // LINUX: attributes [[ATTR]] = { {{.+}} }