2 files changed, 65 insertions, 20 deletions

diff --git a/llvm/lib/Target/X86/X86FastISel.cpp b/llvm/lib/Target/X86/X86FastISel.cpp
index 484f89e8627..25cd68a6db1 100644
--- a/llvm/lib/Target/X86/X86FastISel.cpp
+++ b/llvm/lib/Target/X86/X86FastISel.cpp
@@ -1948,31 +1948,43 @@ bool X86FastISel::FastLowerArguments() {
     return false;
   
   // Only handle simple cases. i.e. Up to 6 i32/i64 scalar arguments.
-  unsigned Idx = 1;
-  for (Function::const_arg_iterator I = F->arg_begin(), E = F->arg_end();
-       I != E; ++I, ++Idx) {
-    if (Idx > 6)
-      return false;
-
+  unsigned GPRCnt = 0;
+  unsigned FPRCnt = 0;
+  unsigned Idx = 0;
+  for (auto const &Arg : F->args()) {
+    // The first argument is at index 1.
+    ++Idx;
     if (F->getAttributes().hasAttribute(Idx, Attribute::ByVal) ||
         F->getAttributes().hasAttribute(Idx, Attribute::InReg) ||
         F->getAttributes().hasAttribute(Idx, Attribute::StructRet) ||
         F->getAttributes().hasAttribute(Idx, Attribute::Nest))
       return false;
 
-    Type *ArgTy = I->getType();
+    Type *ArgTy = Arg.getType();
     if (ArgTy->isStructTy() || ArgTy->isArrayTy() || ArgTy->isVectorTy())
       return false;
 
     EVT ArgVT = TLI.getValueType(ArgTy);
     if (!ArgVT.isSimple()) return false;
     switch (ArgVT.getSimpleVT().SimpleTy) {
+    default: return false;
     case MVT::i32:
     case MVT::i64:
+      ++GPRCnt;
+      break;
+    case MVT::f32:
+    case MVT::f64:
+      if (!Subtarget->hasSSE1())
+        return false;
+      ++FPRCnt;
       break;
-    default:
-      return false;
     }
+
+    if (GPRCnt > 6)
+      return false;
+
+    if (FPRCnt > 8)
+      return false;
   }
 
   static const MCPhysReg GPR32ArgRegs[] = {
@@ -1981,24 +1993,33 @@ bool X86FastISel::FastLowerArguments() {
   static const MCPhysReg GPR64ArgRegs[] = {
     X86::RDI, X86::RSI, X86::RDX, X86::RCX, X86::R8 , X86::R9
   };
+  static const MCPhysReg XMMArgRegs[] = {
+    X86::XMM0, X86::XMM1, X86::XMM2, X86::XMM3,
+    X86::XMM4, X86::XMM5, X86::XMM6, X86::XMM7
+  };
 
-  Idx = 0;
-  const TargetRegisterClass *RC32 = TLI.getRegClassFor(MVT::i32);
-  const TargetRegisterClass *RC64 = TLI.getRegClassFor(MVT::i64);
-  for (Function::const_arg_iterator I = F->arg_begin(), E = F->arg_end();
-       I != E; ++I, ++Idx) {
-    bool is32Bit = TLI.getValueType(I->getType()) == MVT::i32;
-    const TargetRegisterClass *RC = is32Bit ? RC32 : RC64;
-    unsigned SrcReg = is32Bit ? GPR32ArgRegs[Idx] : GPR64ArgRegs[Idx];
+  unsigned GPRIdx = 0;
+  unsigned FPRIdx = 0;
+  for (auto const &Arg : F->args()) {
+    MVT VT = TLI.getSimpleValueType(Arg.getType());
+    const TargetRegisterClass *RC = TLI.getRegClassFor(VT);
+    unsigned SrcReg;
+    switch (VT.SimpleTy) {
+    default: llvm_unreachable("Unexpected value type.");
+    case MVT::i32: SrcReg = GPR32ArgRegs[GPRIdx++]; break;
+    case MVT::i64: SrcReg = GPR64ArgRegs[GPRIdx++]; break;
+    case MVT::f32: // fall-through
+    case MVT::f64: SrcReg = XMMArgRegs[FPRIdx++]; break;
+    }
     unsigned DstReg = FuncInfo.MF->addLiveIn(SrcReg, RC);
     // FIXME: Unfortunately it's necessary to emit a copy from the livein copy.
     // Without this, EmitLiveInCopies may eliminate the livein if its only
     // use is a bitcast (which isn't turned into an instruction).
     unsigned ResultReg = createResultReg(RC);
     BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DbgLoc,
-            TII.get(TargetOpcode::COPY),
-            ResultReg).addReg(DstReg, getKillRegState(true));
-    UpdateValueMap(I, ResultReg);
+            TII.get(TargetOpcode::COPY), ResultReg)
+      .addReg(DstReg, getKillRegState(true));
+    UpdateValueMap(&Arg, ResultReg);
   }
   return true;
 }
diff --git a/llvm/test/CodeGen/X86/fast-isel-args.ll b/llvm/test/CodeGen/X86/fast-isel-args.ll
index 0f3626565e7..8c86a9cc01d 100644
--- a/llvm/test/CodeGen/X86/fast-isel-args.ll
+++ b/llvm/test/CodeGen/X86/fast-isel-args.ll
@@ -23,3 +23,27 @@ entry:
   %add2 = add nsw i64 %add, %conv1
   ret i64 %add2
 }
+
+define float @t4(float %a, float %b, float %c, float %d, float %e, float %f, float %g, float %h) {
+entry:
+  %add1 = fadd float %a, %b
+  %add2 = fadd float %c, %d
+  %add3 = fadd float %e, %f
+  %add4 = fadd float %g, %h
+  %add5 = fadd float %add1, %add2
+  %add6 = fadd float %add3, %add4
+  %add7 = fadd float %add5, %add6
+  ret float %add7
+}
+
+define double @t5(double %a, double %b, double %c, double %d, double %e, double %f, double %g, double %h) {
+entry:
+  %add1 = fadd double %a, %b
+  %add2 = fadd double %c, %d
+  %add3 = fadd double %e, %f
+  %add4 = fadd double %g, %h
+  %add5 = fadd double %add1, %add2
+  %add6 = fadd double %add3, %add4
+  %add7 = fadd double %add5, %add6
+  ret double %add7
+}