[Power9] Implement float128 parameter passing and return values

[lei]

Summary

This patch just enable parameter passing by value for float128 types.
Passing aggregate/union which contain float128 members will be submitted in subsequent patches.

[nemanjai]

A couple of notes:

• This doesn't just handle parameters, but return values as well
• I think all of these will require predication for P9Vector since we don't want to change the fp128 calling conventions on subtargets that don't support the type

[lei]

Added hadP9Vector to calling convention def for f128.
I also merged the patch D47573 for extload of const into here since it's a 2 line fix and the test case associated depended on this patch.

[nemanjai]

LGTM.

lib/Target/PowerPC/PPCCallingConv.td

def RetCC_PPC_Cold : CallingConv<[
   CCIfType<[i1], CCIfNotSubtarget<"isPPC64()", CCPromoteToType<i32>>>,
 
   CCIfType<[i32], CCAssignToReg<[R3]>>,
   CCIfType<[i64], CCAssignToReg<[X3]>>,
   CCIfType<[i128], CCAssignToReg<[X3]>>,
 
   CCIfType<[f32], CCAssignToReg<[F1]>>,
   CCIfType<[f64], CCAssignToReg<[F1]>>,
+  CCIfType<[f128], CCIfSubtarget<"hasP9Vector()", CCAssignToReg<[V2]>>>,
 
   CCIfType<[v4f64, v4f32, v4i1],
            CCIfSubtarget<"hasQPX()", CCAssignToReg<[QF1]>>>,
 
   CCIfType<[v16i8, v8i16, v4i32, v2i64, v1i128, v4f32, v2f64],
            CCIfSubtarget<"hasAltivec()",
            CCAssignToReg<[V2]>>>
 ]>;
def RetCC_PPC : CallingConv<[
   CCIfType<[i64], CCAssignToReg<[X3, X4, X5, X6]>>,
   CCIfType<[i128], CCAssignToReg<[X3, X4, X5, X6]>>,
 
   // Floating point types returned as "direct" go into F1 .. F8; note that
   // only the ELFv2 ABI fully utilizes all these registers.
   CCIfType<[f32], CCAssignToReg<[F1, F2, F3, F4, F5, F6, F7, F8]>>,
   CCIfType<[f64], CCAssignToReg<[F1, F2, F3, F4, F5, F6, F7, F8]>>,
 
+  // For P9, f128 are passed in vector registers.
+  CCIfType<[f128],
+           CCIfSubtarget<"hasP9Vector()",
+           CCAssignToReg<[V2, V3, V4, V5, V6, V7, V8, V9]>>>,
+
   // QPX vectors are returned in QF1 and QF2. 
   CCIfType<[v4f64, v4f32, v4i1],
            CCIfSubtarget<"hasQPX()", CCAssignToReg<[QF1, QF2]>>>,
  
   // Vector types returned as "direct" go into V2 .. V9; note that only the
   // ELFv2 ABI fully utilizes all these registers.
   CCIfType<[v16i8, v8i16, v4i32, v2i64, v1i128, v4f32, v2f64],
            CCIfSubtarget<"hasAltivec()",
def RetCC_PPC64_ELF_FIS : CallingConv<[
   CCIfType<[i1],   CCPromoteToType<i64>>,
   CCIfType<[i8],   CCPromoteToType<i64>>,
   CCIfType<[i16],  CCPromoteToType<i64>>,
   CCIfType<[i32],  CCPromoteToType<i64>>,
   CCIfType<[i64],  CCAssignToReg<[X3, X4, X5, X6]>>,
   CCIfType<[i128], CCAssignToReg<[X3, X4, X5, X6]>>,
   CCIfType<[f32],  CCAssignToReg<[F1, F2, F3, F4, F5, F6, F7, F8]>>,
   CCIfType<[f64],  CCAssignToReg<[F1, F2, F3, F4, F5, F6, F7, F8]>>,
+  CCIfType<[f128],
+           CCIfSubtarget<"hasP9Vector()",
+           CCAssignToReg<[V2, V3, V4, V5, V6, V7, V8, V9]>>>,
   CCIfType<[v4f64, v4f32, v4i1],
            CCIfSubtarget<"hasQPX()", CCAssignToReg<[QF1, QF2]>>>,
   CCIfType<[v16i8, v8i16, v4i32, v2i64, v1i128, v4f32, v2f64],
            CCIfSubtarget<"hasAltivec()",
            CCAssignToReg<[V2, V3, V4, V5, V6, V7, V8, V9]>>>
 ]>;
 
 //===----------------------------------------------------------------------===//
def CC_PPC32_SVR4_Common : CallingConv<[
   
   // Floats are stored in double precision format, thus they have the same
   // alignment and size as doubles.
   CCIfType<[f32,f64], CCAssignToStack<8, 8>>,  
 
   // QPX vectors that are stored in double precision need 32-byte alignment.
   CCIfType<[v4f64, v4i1], CCAssignToStack<32, 32>>,
 
-  // Vectors get 16-byte stack slots that are 16-byte aligned.
-  CCIfType<[v16i8, v8i16, v4i32, v4f32, v2f64, v2i64], CCAssignToStack<16, 16>>
+  // Vectors and float128 get 16-byte stack slots that are 16-byte aligned.
+  CCIfType<[v16i8, v8i16, v4i32, v4f32, v2f64, v2i64], CCAssignToStack<16, 16>>,
+  CCIfType<[f128], CCIfSubtarget<"hasP9Vector()", CCAssignToStack<16, 16>>>
 ]>;
 
 // This calling convention puts vector arguments always on the stack. It is used
 // to assign vector arguments which belong to the variable portion of the
 // parameter list of a variable argument function.
 def CC_PPC32_SVR4_VarArg : CallingConv<[
   CCDelegateTo<CC_PPC32_SVR4_Common>
 ]>;
 
 // In contrast to CC_PPC32_SVR4_VarArg, this calling convention first tries to
 // put vector arguments in vector registers before putting them on the stack.
 def CC_PPC32_SVR4 : CallingConv<[
   // QPX vectors mirror the scalar FP convention.
   CCIfType<[v4f64, v4f32, v4i1], CCIfSubtarget<"hasQPX()",
     CCAssignToReg<[QF1, QF2, QF3, QF4, QF5, QF6, QF7, QF8]>>>,
 
   // The first 12 Vector arguments are passed in AltiVec registers.
   CCIfType<[v16i8, v8i16, v4i32, v2i64, v1i128, v4f32, v2f64],
            CCIfSubtarget<"hasAltivec()", CCAssignToReg<[V2, V3, V4, V5, V6, V7,
                           V8, V9, V10, V11, V12, V13]>>>,
-           
+  // Float128 types treated as vector arguments.
+  CCIfType<[f128],
+           CCIfSubtarget<"hasP9Vector()", CCAssignToReg<[V2, V3, V4, V5, V6, V7,
+                          V8, V9, V10, V11, V12, V13]>>>,
+           
   CCDelegateTo<CC_PPC32_SVR4_Common>
 ]>;  
 
 // Helper "calling convention" to handle aggregate by value arguments.
 // Aggregate by value arguments are always placed in the local variable space
 // of the caller. This calling convention is only used to assign those stack
 // offsets in the callers stack frame.
 //

lib/Target/PowerPC/PPCISelLowering.cpp

if (Subtarget.hasP9Vector()) {
 
       if (EnableQuadPrecision) {
         addRegisterClass(MVT::f128, &PPC::VRRCRegClass);
         setOperationAction(ISD::FADD, MVT::f128, Legal);
         setOperationAction(ISD::FSUB, MVT::f128, Legal);
         setOperationAction(ISD::FDIV, MVT::f128, Legal);
         setOperationAction(ISD::FMUL, MVT::f128, Legal);
         setOperationAction(ISD::FP_EXTEND, MVT::f128, Legal);
-        setLoadExtAction(ISD::EXTLOAD, MVT::f128, MVT::f64, Expand);
+        // No extending loads to f128 on PPC.
+        for (MVT FPT : MVT::fp_valuetypes())
+          setLoadExtAction(ISD::EXTLOAD, MVT::f128, FPT, Expand);
         setOperationAction(ISD::FMA, MVT::f128, Legal);
         setOperationAction(ISD::FP_ROUND, MVT::f64, Legal);
         setOperationAction(ISD::FP_ROUND, MVT::f32, Legal);
         setTruncStoreAction(MVT::f128, MVT::f64, Expand);
         setTruncStoreAction(MVT::f128, MVT::f32, Expand);
       }
 
     }
static unsigned CalculateStackSlotAlignment(EVT ArgVT, EVT OrigVT,
                                             ISD::ArgFlagsTy Flags,
                                             unsigned PtrByteSize) {
   unsigned Align = PtrByteSize;
 
   // Altivec parameters are padded to a 16 byte boundary.
   if (ArgVT == MVT::v4f32 || ArgVT == MVT::v4i32 ||
       ArgVT == MVT::v8i16 || ArgVT == MVT::v16i8 ||
       ArgVT == MVT::v2f64 || ArgVT == MVT::v2i64 ||
-      ArgVT == MVT::v1i128)
+      ArgVT == MVT::v1i128 || ArgVT == MVT::f128)
     Align = 16;
   // QPX vector types stored in double-precision are padded to a 32 byte
   // boundary.
   else if (ArgVT == MVT::v4f64 || ArgVT == MVT::v4i1)
     Align = 32;
 
   // ByVal parameters are aligned as requested.
   if (Flags.isByVal()) {
if (ArgVT == MVT::f32 || ArgVT == MVT::f64 ||
                     ArgVT == MVT::v4i1)))
       if (AvailableFPRs > 0) {
         --AvailableFPRs;
         return false;
       }
     if (ArgVT == MVT::v4f32 || ArgVT == MVT::v4i32 ||
         ArgVT == MVT::v8i16 || ArgVT == MVT::v16i8 ||
         ArgVT == MVT::v2f64 || ArgVT == MVT::v2i64 ||
-        ArgVT == MVT::v1i128)
+        ArgVT == MVT::v1i128 || ArgVT == MVT::f128)
       if (AvailableVRs > 0) {
         --AvailableVRs;
         return false;
       }
   }
 
   return UseMemory;
 }
case MVT::f64:
       break;
     case MVT::v4f32:
     case MVT::v4i32:
     case MVT::v8i16:
     case MVT::v16i8:
     case MVT::v2f64:
     case MVT::v2i64:
     case MVT::v1i128:
+    case MVT::f128:
       if (!Subtarget.hasQPX()) {
         // These can be scalar arguments or elements of a vector array type
         // passed directly.  The latter are used to implement ELFv2 homogenous
         // vector aggregates.
         if (VR_idx != Num_VR_Regs) {
           unsigned VReg = MF.addLiveIn(VR[VR_idx], &PPC::VRRCRegClass);
           ArgVal = DAG.getCopyFromReg(Chain, dl, VReg, ObjectVT);
           ++VR_idx;
if (CallConv == CallingConv::Fast) {
             continue;
           break;
         case MVT::v4i32:
         case MVT::v8i16:
         case MVT::v16i8:
         case MVT::v2f64:
         case MVT::v2i64:
         case MVT::v1i128:
+        case MVT::f128:
           if (++NumVRsUsed <= NumVRs)
             continue;
           break;
         case MVT::v4f32:
           // When using QPX, this is handled like a FP register, otherwise, it
           // is an Altivec register.
           if (Subtarget.hasQPX()) {
             if (++NumFPRsUsed <= NumFPRs)
if (CallConv == CallingConv::Fast) {
       }
     }
 
     /* Respect alignment of argument on the stack.  */
     unsigned Align =
       CalculateStackSlotAlignment(ArgVT, OrigVT, Flags, PtrByteSize);
     NumBytes = ((NumBytes + Align - 1) / Align) * Align;
 
+    //LEI
     NumBytes += CalculateStackSlotSize(ArgVT, Flags, PtrByteSize);
     if (Flags.isInConsecutiveRegsLast())
       NumBytes = ((NumBytes + PtrByteSize - 1)/PtrByteSize) * PtrByteSize;
   }
 
   unsigned NumBytesActuallyUsed = NumBytes;
 
   // In the old ELFv1 ABI,
for (unsigned i = 0; i != NumOps; ++i) {
     }
     case MVT::v4f32:
     case MVT::v4i32:
     case MVT::v8i16:
     case MVT::v16i8:
     case MVT::v2f64:
     case MVT::v2i64:
     case MVT::v1i128:
+    case MVT::f128:
       if (!Subtarget.hasQPX()) {
       // These can be scalar arguments or elements of a vector array type
       // passed directly.  The latter are used to implement ELFv2 homogenous
       // vector aggregates.
 
       // For a varargs call, named arguments go into VRs or on the stack as
       // usual; unnamed arguments always go to the stack or the corresponding
       // GPRs when within range.  For now, we always put the value in both

test/CodeGen/PowerPC/f128-passByValue.ll

+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc -mcpu=pwr9 -mtriple=powerpc64le-unknown-unknown \
+; RUN:   -enable-ppc-quad-precision -ppc-asm-full-reg-names \
+; RUN:   -verify-machineinstrs < %s | FileCheck %s
+
+; Function Attrs: norecurse nounwind readnone
+define fp128 @loadConstant() {
+  ; CHECK-LABEL: loadConstant:
+  ; CHECK:       # %bb.0: # %entry
+  ; CHECK-NEXT:    addis r3, r2, .LCPI0_0@toc@ha
+  ; CHECK-NEXT:    addi r3, r3, .LCPI0_0@toc@l
+  ; CHECK-NEXT:    lxvx vs34, 0, r3
+  ; CHECK-NEXT:    blr
+  entry:
+    ret fp128 0xL00000000000000004001400000000000
+}
+
+; Function Attrs: norecurse nounwind readnone
+define fp128 @loadConstant2(fp128 %a, fp128 %b) {
+  ; CHECK-LABEL: loadConstant2:
+  ; CHECK:       # %bb.0: # %entry
+  ; CHECK-NEXT:    xsaddqp v2, v2, v3
+  ; CHECK-NEXT:    addis r3, r2, .LCPI1_0@toc@ha
+  ; CHECK-NEXT:    addi r3, r3, .LCPI1_0@toc@l
+  ; CHECK-NEXT:    lxvx vs51, 0, r3
+  ; CHECK-NEXT:    xsaddqp v2, v2, v19
+  ; CHECK-NEXT:    blr
+  entry:
+    %add = fadd fp128 %a, %b
+      %add1 = fadd fp128 %add, 0xL00000000000000004001400000000000
+        ret fp128 %add1
+}
+
+; Test passing float128 by value.
+; Function Attrs: norecurse nounwind readnone
+define signext i32 @fp128Param(fp128 %a) {
+; CHECK-LABEL: fp128Param:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    xscvqpswz v2, v2
+; CHECK-NEXT:    mfvsrwz r3, vs34
+; CHECK-NEXT:    extsw r3, r3
+; CHECK-NEXT:    blr
+entry:
+  %conv = fptosi fp128 %a to i32
+  ret i32 %conv
+}
+
+; Test float128 as return value.
+; Function Attrs: norecurse nounwind readnone
+define fp128 @fp128Return(fp128 %a, fp128 %b) {
+; CHECK-LABEL: fp128Return:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    xsaddqp v2, v2, v3
+; CHECK-NEXT:    blr
+entry:
+  %add = fadd fp128 %a, %b
+  ret fp128 %add
+}
+
+; array of float128 types
+; Function Attrs: norecurse nounwind readonly
+define fp128 @fp128Array(fp128* nocapture readonly %farray,
+                         i32 signext %loopcnt, fp128* nocapture readnone %sum) {
+; CHECK-LABEL: fp128Array:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    sldi r4, r4, 4
+; CHECK-NEXT:    lxv vs34, 0(r3)
+; CHECK-NEXT:    add r4, r3, r4
+; CHECK-NEXT:    lxv vs35, -16(r4)
+; CHECK-NEXT:    xsaddqp v2, v2, v3
+; CHECK-NEXT:    blr
+entry:
+  %0 = load fp128, fp128* %farray, align 16
+  %sub = add nsw i32 %loopcnt, -1
+  %idxprom = sext i32 %sub to i64
+  %arrayidx1 = getelementptr inbounds fp128, fp128* %farray, i64 %idxprom
+  %1 = load fp128, fp128* %arrayidx1, align 16
+  %add = fadd fp128 %0, %1
+  ret fp128 %add
+}
+
+; Up to 12 qualified floating-point arguments can be passed in v2-v13.
+; Function to test passing 13 float128 parameters.
+; Function Attrs: norecurse nounwind readnone
+define fp128 @maxVecParam(fp128 %p1, fp128 %p2, fp128 %p3, fp128 %p4, fp128 %p5,
+                          fp128 %p6, fp128 %p7, fp128 %p8, fp128 %p9, fp128 %p10,
+                          fp128 %p11, fp128 %p12, fp128 %p13) {
+; CHECK-LABEL: maxVecParam:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    xsaddqp v2, v2, v3
+; CHECK-NEXT:    lxv vs51, 224(r1)
+; CHECK-NEXT:    xsaddqp v2, v2, v4
+; CHECK-NEXT:    xsaddqp v2, v2, v5
+; CHECK-NEXT:    xsaddqp v2, v2, v6
+; CHECK-NEXT:    xsaddqp v2, v2, v7
+; CHECK-NEXT:    xsaddqp v2, v2, v8
+; CHECK-NEXT:    xsaddqp v2, v2, v9
+; CHECK-NEXT:    xsaddqp v2, v2, v10
+; CHECK-NEXT:    xsaddqp v2, v2, v11
+; CHECK-NEXT:    xsaddqp v2, v2, v12
+; CHECK-NEXT:    xsaddqp v2, v2, v13
+; CHECK-NEXT:    xssubqp v2, v2, v19
+; CHECK-NEXT:    blr
+entry:
+  %add = fadd fp128 %p1, %p2
+  %add1 = fadd fp128 %add, %p3
+  %add2 = fadd fp128 %add1, %p4
+  %add3 = fadd fp128 %add2, %p5
+  %add4 = fadd fp128 %add3, %p6
+  %add5 = fadd fp128 %add4, %p7
+  %add6 = fadd fp128 %add5, %p8
+  %add7 = fadd fp128 %add6, %p9
+  %add8 = fadd fp128 %add7, %p10
+  %add9 = fadd fp128 %add8, %p11
+  %add10 = fadd fp128 %add9, %p12
+  %sub = fsub fp128 %add10, %p13
+  ret fp128 %sub
+}
+
+; Passing a mix of float128 and other type parameters.
+; Function Attrs: norecurse nounwind readnone
+define fp128 @mixParam_01(fp128 %a, i32 signext %i, fp128 %b) {
+; CHECK-LABEL: mixParam_01:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    mtvsrwa vs36, r5
+; CHECK-NEXT:    xsaddqp v2, v2, v3
+; CHECK-NEXT:    xscvsdqp v19, v4
+; CHECK-NEXT:    xsaddqp v2, v2, v19
+; CHECK-NEXT:    blr
+entry:
+  %add = fadd fp128 %a, %b
+  %conv = sitofp i32 %i to fp128
+  %add1 = fadd fp128 %add, %conv
+  ret fp128 %add1
+}
+; Function Attrs: norecurse nounwind readnone
+define fastcc fp128 @mixParam_01f(fp128 %a, i32 signext %i, fp128 %b) {
+; CHECK-LABEL: mixParam_01f:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    mtvsrwa vs36, r3
+; CHECK-NEXT:    xsaddqp v2, v2, v3
+; CHECK-NEXT:    xscvsdqp v19, v4
+; CHECK-NEXT:    xsaddqp v2, v2, v19
+; CHECK-NEXT:    blr
+entry:
+  %add = fadd fp128 %a, %b
+  %conv = sitofp i32 %i to fp128
+  %add1 = fadd fp128 %add, %conv
+  ret fp128 %add1
+}
+
+; Function Attrs: norecurse nounwind
+define fp128 @mixParam_02(fp128 %p1, double %p2, i64* nocapture %p3,
+                          i16 signext %p4, fp128* nocapture readonly %p5,
+                          i32 signext %p6, i8 zeroext %p7, i32 zeroext %p8) {
+; CHECK-LABEL: mixParam_02:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    add r4, r7, r9
+; CHECK-NEXT:    lwz r3, 96(r1)
+; CHECK-NEXT:    xxlor vs35, f1, f1
+; CHECK-NEXT:    xscvdpqp v3, v3
+; CHECK-NEXT:    add r4, r4, r10
+; CHECK-NEXT:    add r3, r4, r3
+; CHECK-NEXT:    clrldi r3, r3, 32
+; CHECK-NEXT:    std r3, 0(r6)
+; CHECK-NEXT:    lxv vs36, 0(r8)
+; CHECK-NEXT:    xsaddqp v2, v4, v2
+; CHECK-NEXT:    xsaddqp v2, v2, v3
+; CHECK-NEXT:    blr
+entry:
+  %conv = sext i16 %p4 to i32
+  %add = add nsw i32 %conv, %p6
+  %conv1 = zext i8 %p7 to i32
+  %add2 = add nsw i32 %add, %conv1
+  %add3 = add i32 %add2, %p8
+  %conv4 = zext i32 %add3 to i64
+  store i64 %conv4, i64* %p3, align 8
+  %0 = load fp128, fp128* %p5, align 16
+  %add5 = fadd fp128 %0, %p1
+  %conv6 = fpext double %p2 to fp128
+  %add7 = fadd fp128 %add5, %conv6
+  ret fp128 %add7
+}
+
+; Function Attrs: norecurse nounwind
+define fastcc fp128 @mixParam_02f(fp128 %p1, double %p2, i64* nocapture %p3,
+                                  i16 signext %p4, fp128* nocapture readonly %p5,
+                                  i32 signext %p6, i8 zeroext %p7, i32 zeroext %p8) {
+; CHECK-LABEL: mixParam_02f:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    add r4, r4, r6
+; CHECK-NEXT:    xxlor vs35, f1, f1
+; CHECK-NEXT:    add r4, r4, r7
+; CHECK-NEXT:    xscvdpqp v3, v3
+; CHECK-NEXT:    add r4, r4, r8
+; CHECK-NEXT:    clrldi r4, r4, 32
+; CHECK-NEXT:    std r4, 0(r3)
+; CHECK-NEXT:    lxv vs36, 0(r5)
+; CHECK-NEXT:    xsaddqp v2, v4, v2
+; CHECK-NEXT:    xsaddqp v2, v2, v3
+; CHECK-NEXT:    blr
+entry:
+  %conv = sext i16 %p4 to i32
+  %add = add nsw i32 %conv, %p6
+  %conv1 = zext i8 %p7 to i32
+  %add2 = add nsw i32 %add, %conv1
+  %add3 = add i32 %add2, %p8
+  %conv4 = zext i32 %add3 to i64
+  store i64 %conv4, i64* %p3, align 8
+  %0 = load fp128, fp128* %p5, align 16
+  %add5 = fadd fp128 %0, %p1
+  %conv6 = fpext double %p2 to fp128
+  %add7 = fadd fp128 %add5, %conv6
+  ret fp128 %add7
+}
+
+; Passing a mix of float128 and vector parameters.
+; Function Attrs: norecurse nounwind
+define void @mixParam_03(fp128 %f1, double* nocapture %d1, <4 x i32> %vec1,
+                         fp128* nocapture %f2, i32 signext %i1, i8 zeroext %c1,
+                         <4 x i32>* nocapture %vec2) {
+; CHECK-LABEL: mixParam_03:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    ld r3, 104(r1)
+; CHECK-NEXT:    mtvsrwa vs36, r10
+; CHECK-NEXT:    stxv vs34, 0(r9)
+; CHECK-NEXT:    stxvx vs35, 0, r3
+; CHECK-NEXT:    lxv vs34, 0(r9)
+; CHECK-NEXT:    xscvsdqp v3, v4
+; CHECK-NEXT:    xsaddqp v2, v2, v3
+; CHECK-NEXT:    xscvqpdp v2, v2
+; CHECK-NEXT:    stxsd v2, 0(r5)
+; CHECK-NEXT:    blr
+entry:
+  store fp128 %f1, fp128* %f2, align 16
+  store <4 x i32> %vec1, <4 x i32>* %vec2, align 16
+  %0 = load fp128, fp128* %f2, align 16
+  %conv = sitofp i32 %i1 to fp128
+  %add = fadd fp128 %0, %conv
+  %conv1 = fptrunc fp128 %add to double
+  store double %conv1, double* %d1, align 8
+  ret void
+}
+
+; Function Attrs: norecurse nounwind
+define fastcc void @mixParam_03f(fp128 %f1, double* nocapture %d1, <4 x i32> %vec1,
+                                 fp128* nocapture %f2, i32 signext %i1, i8 zeroext %c1,
+                                 <4 x i32>* nocapture %vec2) {
+; CHECK-LABEL: mixParam_03f:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    mtvsrwa vs36, r5
+; CHECK-NEXT:    stxv vs34, 0(r4)
+; CHECK-NEXT:    stxv vs35, 0(r7)
+; CHECK-NEXT:    lxv vs34, 0(r4)
+; CHECK-NEXT:    xscvsdqp v3, v4
+; CHECK-NEXT:    xsaddqp v2, v2, v3
+; CHECK-NEXT:    xscvqpdp v2, v2
+; CHECK-NEXT:    stxsd v2, 0(r3)
+; CHECK-NEXT:    blr
+entry:
+  store fp128 %f1, fp128* %f2, align 16
+  store <4 x i32> %vec1, <4 x i32>* %vec2, align 16
+  %0 = load fp128, fp128* %f2, align 16
+  %conv = sitofp i32 %i1 to fp128
+  %add = fadd fp128 %0, %conv
+  %conv1 = fptrunc fp128 %add to double
+  store double %conv1, double* %d1, align 8
+  ret void
+}