X86/AArch64/ARM: Factor out common sincos_stret logic; NFCI

Note:
- X86ISelLowering: setLibcallName(SINCOS) was superfluous as
  InitLibcalls() already does it.
- ARMISelLowering: Setting libcallnames for sincos/sincosf seemed
  superfluous as in the darwin case it wouldn't be used while for all
  other cases InitLibcalls already does it.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@321036 91177308-0d34-0410-b5e6-96231b3b80d8

include/llvm/CodeGen/RuntimeLibcalls.def

@@ -165,6 +165,8 @@ HANDLE_LIBCALL(SINCOS_F64, nullptr)
 HANDLE_LIBCALL(SINCOS_F80, nullptr)
 HANDLE_LIBCALL(SINCOS_F128, nullptr)
 HANDLE_LIBCALL(SINCOS_PPCF128, nullptr)
+HANDLE_LIBCALL(SINCOS_STRET_F32, nullptr)
+HANDLE_LIBCALL(SINCOS_STRET_F64, nullptr)
 HANDLE_LIBCALL(POW_F32, "powf")
 HANDLE_LIBCALL(POW_F64, "pow")
 HANDLE_LIBCALL(POW_F80, "powl")

include/llvm/CodeGen/TargetLowering.h

@@ -2438,6 +2438,9 @@ private:
   /// Stores the CallingConv that should be used for each libcall.
   CallingConv::ID LibcallCallingConvs[RTLIB::UNKNOWN_LIBCALL];
 
+  /// Set default libcall names and calling conventions.
+  void InitLibcalls(const Triple &TT);
+
 protected:
   /// Return true if the extension represented by \p I is free.
   /// \pre \p I is a sign, zero, or fp extension and

lib/CodeGen/TargetLoweringBase.cpp

@@ -89,6 +89,17 @@ static cl::opt<unsigned> OptsizeJumpTableDensity(
     cl::desc("Minimum density for building a jump table in "
              "an optsize function"));
 
+static bool darwinHasSinCos(const Triple &TT) {
+  assert(TT.isOSDarwin() && "should be called with darwin triple");
+  // macos < 10.9 has no sincos_stret and we don't bother for 32bit code.
+  if (TT.isMacOSX())
+    return !TT.isMacOSXVersionLT(10, 9) && TT.isArch64Bit();
+  // ios < 7.0 has no sincos_stret (watchos reports version 2.0 but is fine).
+  if (TT.isiOS())
+    return !TT.isOSVersionLT(7, 0) || TT.isWatchOS();
+  return true;
+}
+
 // Although this default value is arbitrary, it is not random. It is assumed
 // that a condition that evaluates the same way by a higher percentage than this
 // is best represented as control flow. Therefore, the default value N should be
@@ -100,10 +111,9 @@ static cl::opt<int> MinPercentageForPredictableBranch(
              "or false to assume that the condition is predictable"),
     cl::Hidden);
 
-/// InitLibcallNames - Set default libcall names.
-static void InitLibcallNames(const char **Names, const Triple &TT) {
+void TargetLoweringBase::InitLibcalls(const Triple &TT) {
 #define HANDLE_LIBCALL(code, name) \
-  Names[RTLIB::code] = name;
+  setLibcallName(RTLIB::code, name);
 #include "llvm/CodeGen/RuntimeLibcalls.def"
 #undef HANDLE_LIBCALL
 
@@ -112,27 +122,38 @@ static void InitLibcallNames(const char **Names, const Triple &TT) {
     // For f16/f32 conversions, Darwin uses the standard naming scheme, instead
     // of the gnueabi-style __gnu_*_ieee.
     // FIXME: What about other targets?
-    Names[RTLIB::FPEXT_F16_F32] = "__extendhfsf2";
-    Names[RTLIB::FPROUND_F32_F16] = "__truncsfhf2";
+    setLibcallName(RTLIB::FPEXT_F16_F32, "__extendhfsf2");
+    setLibcallName(RTLIB::FPROUND_F32_F16, "__truncsfhf2");
 
     // Darwin 10 and higher has an optimized __bzero.
-    if (!TT.isMacOSX() || !TT.isMacOSXVersionLT(10, 6))
-      Names[RTLIB::BZERO] = "__bzero";
+    if (!TT.isMacOSX() || !TT.isMacOSXVersionLT(10, 6) || TT.isArch64Bit())
+      setLibcallName(RTLIB::BZERO, "__bzero");
+
+    if (darwinHasSinCos(TT)) {
+      setLibcallName(RTLIB::SINCOS_STRET_F32, "__sincosf_stret");
+      setLibcallName(RTLIB::SINCOS_STRET_F64, "__sincos_stret");
+      if (TT.isWatchABI()) {
+        setLibcallCallingConv(RTLIB::SINCOS_STRET_F32,
+                              CallingConv::ARM_AAPCS_VFP);
+        setLibcallCallingConv(RTLIB::SINCOS_STRET_F64,
+                              CallingConv::ARM_AAPCS_VFP);
+      }
+    }
   } else {
-    Names[RTLIB::FPEXT_F16_F32] = "__gnu_h2f_ieee";
-    Names[RTLIB::FPROUND_F32_F16] = "__gnu_f2h_ieee";
+    setLibcallName(RTLIB::FPEXT_F16_F32, "__gnu_h2f_ieee");
+    setLibcallName(RTLIB::FPROUND_F32_F16, "__gnu_f2h_ieee");
   }
 
   if (TT.isGNUEnvironment() || TT.isOSFuchsia()) {
-    Names[RTLIB::SINCOS_F32] = "sincosf";
-    Names[RTLIB::SINCOS_F64] = "sincos";
-    Names[RTLIB::SINCOS_F80] = "sincosl";
-    Names[RTLIB::SINCOS_F128] = "sincosl";
-    Names[RTLIB::SINCOS_PPCF128] = "sincosl";
+    setLibcallName(RTLIB::SINCOS_F32, "sincosf");
+    setLibcallName(RTLIB::SINCOS_F64, "sincos");
+    setLibcallName(RTLIB::SINCOS_F80, "sincosl");
+    setLibcallName(RTLIB::SINCOS_F128, "sincosl");
+    setLibcallName(RTLIB::SINCOS_PPCF128, "sincosl");
   }
 
   if (TT.isOSOpenBSD()) {
-    Names[RTLIB::STACKPROTECTOR_CHECK_FAIL] = nullptr;
+    setLibcallName(RTLIB::STACKPROTECTOR_CHECK_FAIL, nullptr);
   }
 }
 
@@ -528,7 +549,7 @@ TargetLoweringBase::TargetLoweringBase(const TargetMachine &tm) : TM(tm) {
 
   std::fill(std::begin(LibcallRoutineNames), std::end(LibcallRoutineNames), nullptr);
 
-  InitLibcallNames(LibcallRoutineNames, TM.getTargetTriple());
+  InitLibcalls(TM.getTargetTriple());
   InitCmpLibcallCCs(CmpLibcallCCs);
   InitLibcallCallingConvs(LibcallCallingConvs);
 }

lib/Target/AArch64/AArch64ISelLowering.cpp

@@ -470,10 +470,9 @@ AArch64TargetLowering::AArch64TargetLowering(const TargetMachine &TM,
   if (Subtarget->hasPerfMon())
     setOperationAction(ISD::READCYCLECOUNTER, MVT::i64, Legal);
 
-  if (Subtarget->isTargetMachO()) {
-    // For iOS, we don't want to the normal expansion of a libcall to
-    // sincos. We want to issue a libcall to __sincos_stret to avoid memory
-    // traffic.
+  if (getLibcallName(RTLIB::SINCOS_STRET_F32) != nullptr &&
+      getLibcallName(RTLIB::SINCOS_STRET_F64) != nullptr) {
+    // Issue __sincos_stret if available.
     setOperationAction(ISD::FSINCOS, MVT::f64, Custom);
     setOperationAction(ISD::FSINCOS, MVT::f32, Custom);
   } else {
@@ -2328,8 +2327,9 @@ SDValue AArch64TargetLowering::LowerFSINCOS(SDValue Op,
   Entry.IsZExt = false;
   Args.push_back(Entry);
 
-  const char *LibcallName =
-      (ArgVT == MVT::f64) ? "__sincos_stret" : "__sincosf_stret";
+  RTLIB::Libcall LC = ArgVT == MVT::f64 ? RTLIB::SINCOS_STRET_F64
+                                        : RTLIB::SINCOS_STRET_F32;
+  const char *LibcallName = getLibcallName(LC);
   SDValue Callee =
       DAG.getExternalSymbol(LibcallName, getPointerTy(DAG.getDataLayout()));
 

lib/Target/ARM/ARMISelLowering.cpp

@@ -1084,20 +1084,11 @@ ARMTargetLowering::ARMTargetLowering(const TargetMachine &TM,
     }
   }
 
-  // Combine sin / cos into one node or libcall if possible.
-  if (Subtarget->hasSinCos()) {
-    setLibcallName(RTLIB::SINCOS_F32, "sincosf");
-    setLibcallName(RTLIB::SINCOS_F64, "sincos");
-    if (Subtarget->isTargetWatchABI()) {
-      setLibcallCallingConv(RTLIB::SINCOS_F32, CallingConv::ARM_AAPCS_VFP);
-      setLibcallCallingConv(RTLIB::SINCOS_F64, CallingConv::ARM_AAPCS_VFP);
-    }
-    if (Subtarget->isTargetIOS() || Subtarget->isTargetWatchOS()) {
-      // For iOS, we don't want to the normal expansion of a libcall to
-      // sincos. We want to issue a libcall to __sincos_stret.
-      setOperationAction(ISD::FSINCOS, MVT::f64, Custom);
-      setOperationAction(ISD::FSINCOS, MVT::f32, Custom);
-    }
+  // Use __sincos_stret if available.
+  if (getLibcallName(RTLIB::SINCOS_STRET_F32) != nullptr &&
+      getLibcallName(RTLIB::SINCOS_STRET_F64) != nullptr) {
+    setOperationAction(ISD::FSINCOS, MVT::f64, Custom);
+    setOperationAction(ISD::FSINCOS, MVT::f32, Custom);
   }
 
   // FP-ARMv8 implements a lot of rounding-like FP operations.
@@ -7523,10 +7514,9 @@ SDValue ARMTargetLowering::LowerFSINCOS(SDValue Op, SelectionDAG &DAG) const {
   Entry.IsZExt = false;
   Args.push_back(Entry);
 
-  const char *LibcallName =
-      (ArgVT == MVT::f64) ? "__sincos_stret" : "__sincosf_stret";
   RTLIB::Libcall LC =
-      (ArgVT == MVT::f64) ? RTLIB::SINCOS_F64 : RTLIB::SINCOS_F32;
+      (ArgVT == MVT::f64) ? RTLIB::SINCOS_STRET_F64 : RTLIB::SINCOS_STRET_F32;
+  const char *LibcallName = getLibcallName(LC);
   CallingConv::ID CC = getLibcallCallingConv(LC);
   SDValue Callee = DAG.getExternalSymbol(LibcallName, getPointerTy(DL));
 

lib/Target/ARM/ARMSubtarget.cpp

@@ -348,11 +348,6 @@ unsigned ARMSubtarget::getMispredictionPenalty() const {
   return SchedModel.MispredictPenalty;
 }
 
-bool ARMSubtarget::hasSinCos() const {
-  return isTargetWatchOS() ||
-    (isTargetIOS() && !getTargetTriple().isOSVersionLT(7, 0));
-}
-
 bool ARMSubtarget::enableMachineScheduler() const {
   // Enable the MachineScheduler before register allocation for subtargets
   // with the use-misched feature.

lib/Target/ARM/ARMSubtarget.h

@@ -712,10 +712,6 @@ public:
 
   unsigned getMispredictionPenalty() const;
 
-  /// This function returns true if the target has sincos() routine in its
-  /// compiler runtime or math libraries.
-  bool hasSinCos() const;
-
   /// Returns true if machine scheduler should be enabled.
   bool enableMachineScheduler() const override;
 

lib/Target/X86/X86ISelLowering.cpp

@@ -1622,16 +1622,11 @@ X86TargetLowering::X86TargetLowering(const X86TargetMachine &TM,
     setLibcallName(RTLIB::MUL_I128, nullptr);
   }
 
-  // Combine sin / cos into one node or libcall if possible.
-  if (Subtarget.hasSinCos()) {
-    setLibcallName(RTLIB::SINCOS_F32, "sincosf");
-    setLibcallName(RTLIB::SINCOS_F64, "sincos");
-    if (Subtarget.isTargetDarwin()) {
-      // For MacOSX, we don't want the normal expansion of a libcall to sincos.
-      // We want to issue a libcall to __sincos_stret to avoid memory traffic.
-      setOperationAction(ISD::FSINCOS, MVT::f64, Custom);
-      setOperationAction(ISD::FSINCOS, MVT::f32, Custom);
-    }
+  // Combine sin / cos into _sincos_stret if it is available.
+  if (getLibcallName(RTLIB::SINCOS_STRET_F32) != nullptr &&
+      getLibcallName(RTLIB::SINCOS_STRET_F64) != nullptr) {
+    setOperationAction(ISD::FSINCOS, MVT::f64, Custom);
+    setOperationAction(ISD::FSINCOS, MVT::f32, Custom);
   }
 
   if (Subtarget.isTargetWin64()) {
@@ -24101,8 +24096,9 @@ static SDValue LowerFSINCOS(SDValue Op, const X86Subtarget &Subtarget,
   // Only optimize x86_64 for now. i386 is a bit messy. For f32,
   // the small struct {f32, f32} is returned in (eax, edx). For f64,
   // the results are returned via SRet in memory.
-  const char *LibcallName =  isF64 ? "__sincos_stret" : "__sincosf_stret";
   const TargetLowering &TLI = DAG.getTargetLoweringInfo();
+  RTLIB::Libcall LC = isF64 ? RTLIB::SINCOS_STRET_F64 : RTLIB::SINCOS_STRET_F32;
+  const char *LibcallName = TLI.getLibcallName(LC);
   SDValue Callee =
       DAG.getExternalSymbol(LibcallName, TLI.getPointerTy(DAG.getDataLayout()));
 

lib/Target/X86/X86Subtarget.cpp

@@ -174,15 +174,6 @@ X86Subtarget::classifyGlobalFunctionReference(const GlobalValue *GV,
   return X86II::MO_NO_FLAG;
 }
 
-bool X86Subtarget::hasSinCos() const {
-  if (getTargetTriple().isMacOSX()) {
-    return !getTargetTriple().isMacOSXVersionLT(10, 9) && is64Bit();
-  } else if (getTargetTriple().isOSFuchsia()) {
-    return true;
-  }
-  return false;
-}
-
 /// Return true if the subtarget allows calls to immediate address.
 bool X86Subtarget::isLegalToCallImmediateAddr() const {
   // FIXME: I386 PE/COFF supports PC relative calls using IMAGE_REL_I386_REL32

lib/Target/X86/X86Subtarget.h

@@ -682,10 +682,6 @@ public:
   /// Return true if the subtarget allows calls to immediate address.
   bool isLegalToCallImmediateAddr() const;
 
-  /// This function returns true if the target has sincos() routine in its
-  /// compiler runtime or math libraries.
-  bool hasSinCos() const;
-
   /// Enable the MachineScheduler pass for all X86 subtargets.
   bool enableMachineScheduler() const override { return true; }