Fix and re-enable tail call optimization of expanded libcalls.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@120622 91177308-0d34-0410-b5e6-96231b3b80d8

diff --git a/lib/Target/ARM/ARMISelLowering.cpp b/lib/Target/ARM/ARMISelLowering.cpp
index 3c90704b7e9a4d5fe9e51db7cb7c5402f13a24fd..c162757cbd58b9377f951051bee659dedb6bb2ee 100644 (file)
--- a/lib/Target/ARM/ARMISelLowering.cpp
+++ b/lib/Target/ARM/ARMISelLowering.cpp
@@ -1737,6 +1737,8 @@ bool ARMTargetLowering::isUsedByReturnOnly(SDNode *N) const {
 
   if (NumCopies != 1 && NumCopies != 2)
     return false;
+
+  bool HasRet = false;
   for (unsigned i = 0; i < NumCopies; ++i) {
     SDNode *Copy = Copies[i];
     for (SDNode::use_iterator UI = Copy->use_begin(), UE = Copy->use_end();
@@ -1749,10 +1751,11 @@ bool ARMTargetLowering::isUsedByReturnOnly(SDNode *N) const {
       }
       if (UI->getOpcode() != ARMISD::RET_FLAG)
         return false;
+      HasRet = true;
     }
   }
 
-  return true;
+  return HasRet;
 }
 
 // ConstantPool, JumpTable, GlobalAddress, and ExternalSymbol are lowered as

diff --git a/lib/Target/X86/X86ISelLowering.cpp b/lib/Target/X86/X86ISelLowering.cpp
index 07bd5cf5b86588b1488c822eddff8f11b4312432..f3c1b8537304fe237e1572c60e0b95a6ee47395b 100644 (file)
--- a/lib/Target/X86/X86ISelLowering.cpp
+++ b/lib/Target/X86/X86ISelLowering.cpp
@@ -1339,22 +1339,25 @@ X86TargetLowering::LowerReturn(SDValue Chain,
 }
 
 bool X86TargetLowering::isUsedByReturnOnly(SDNode *N) const {
-  // Temporarily disabled.
-  return false;
   if (N->getNumValues() != 1)
     return false;
   if (!N->hasNUsesOfValue(1, 0))
     return false;
 
   SDNode *Copy = *N->use_begin();
-  if (Copy->getOpcode() != ISD::CopyToReg)
+  if (Copy->getOpcode() != ISD::CopyToReg &&
+      Copy->getOpcode() != ISD::FP_EXTEND)
     return false;
+
+  bool HasRet = false;
   for (SDNode::use_iterator UI = Copy->use_begin(), UE = Copy->use_end();
-       UI != UE; ++UI)
+       UI != UE; ++UI) {
     if (UI->getOpcode() != X86ISD::RET_FLAG)
       return false;
+    HasRet = true;
+  }
 
-  return true;
+  return HasRet;
 }
 
 /// LowerCallResult - Lower the result values of a call into the
@@ -2161,19 +2164,17 @@ X86TargetLowering::LowerCall(SDValue Chain, SDValue Callee,
   } else if (ExternalSymbolSDNode *S = dyn_cast<ExternalSymbolSDNode>(Callee)) {
     unsigned char OpFlags = 0;
 
-    if (!isTailCall) {
-      // On ELF targets, in either X86-64 or X86-32 mode, direct calls to
-      // external symbols should go through the PLT.
-      if (Subtarget->isTargetELF() &&
-          getTargetMachine().getRelocationModel() == Reloc::PIC_) {
-        OpFlags = X86II::MO_PLT;
-      } else if (Subtarget->isPICStyleStubAny() &&
-                 Subtarget->getDarwinVers() < 9) {
-        // PC-relative references to external symbols should go through $stub,
-        // unless we're building with the leopard linker or later, which
-        // automatically synthesizes these stubs.
-        OpFlags = X86II::MO_DARWIN_STUB;
-      }
+    // On ELF targets, in either X86-64 or X86-32 mode, direct calls to
+    // external symbols should go through the PLT.
+    if (Subtarget->isTargetELF() &&
+        getTargetMachine().getRelocationModel() == Reloc::PIC_) {
+      OpFlags = X86II::MO_PLT;
+    } else if (Subtarget->isPICStyleStubAny() &&
+               Subtarget->getDarwinVers() < 9) {
+      // PC-relative references to external symbols should go through $stub,
+      // unless we're building with the leopard linker or later, which
+      // automatically synthesizes these stubs.
+      OpFlags = X86II::MO_DARWIN_STUB;
     }
 
     Callee = DAG.getTargetExternalSymbol(S->getSymbol(), getPointerTy(),

diff --git a/test/CodeGen/X86/sibcall-5.ll b/test/CodeGen/X86/sibcall-5.ll
index 24b88e4cf87667322d7cb181262283259e4456e6..00b835361a671ddd2a782538bc13ef6e4a04ee20 100644 (file)
--- a/test/CodeGen/X86/sibcall-5.ll
+++ b/test/CodeGen/X86/sibcall-5.ll
@@ -1,20 +1,26 @@
-; RUN: llc < %s -march=x86-64 | FileCheck %s
-; XFAIL: *
+; RUN: llc < %s -mtriple=i386-apple-darwin8 | FileCheck %s --check-prefix=X32
+; RUN: llc < %s -mtriple=x86_64-apple-darwin | FileCheck %s --check-prefix=X64
 
 ; Sibcall optimization of expanded libcalls.
 ; rdar://8707777
 
 define double @foo(double %a) nounwind readonly ssp {
 entry:
-; CHECK: foo:
-; CHECK: jmp {{_?}}sin
+; X32: foo:
+; X32: jmp _sin$stub
+
+; X64: foo:
+; X64: jmp _sin
   %0 = tail call double @sin(double %a) nounwind readonly
   ret double %0
 }
 
 define float @bar(float %a) nounwind readonly ssp {
-; CHECK: bar:
-; CHECK: jmp {{_?}}sinf
+; X32: bar:
+; X32: jmp _sinf$stub
+
+; X64: bar:
+; X64: jmp _sinf
 entry:
   %0 = tail call float @sinf(float %a) nounwind readonly
   ret float %0