[PowerPC] Add exception constraint to FP arithmetic

[qiucf]

Summary

D55506 introduced concept of strict-fp nodes into LLVM. This patch is to add strict-fp exception support to PowerPC backend, and only contains basic arithmetic operations (+, -, *, /). Patch for the rest of floating point operations will be in following patches.

[• wuzish]

update testcases.

[qiucf]

Rebase and fix conflicts.

[merge_guards_bot]

Unit tests: pass. 61258 tests passed, 0 failed and 736 were skipped.

clang-tidy: pass.

clang-format: pass.

Build artifacts: diff.json, clang-tidy.txt, clang-format.patch, CMakeCache.txt, console-log.txt, test-results.xml

[qiucf]

Fix a test error

[merge_guards_bot]

Unit tests: unknown.

clang-tidy: unknown.

clang-format: unknown.

Build artifacts: console-log.txt

[qiucf]

• Rename and combine some cases together.
• Add some more cases.

[qiucf]

Ping...

[andrew.w.kaylor]

I don't know enough about PowerPC to review the check in the tests, but the basic constrained intrinsic handling looks correct.

[qiucf]

In D63916#1887394, @andrew.w.kaylor wrote:

I don't know enough about PowerPC to review the check in the tests, but the basic constrained intrinsic handling looks correct.

Thanks! @jsji @nemanjai Could you help to review PPC codegen part?

[steven.zhang]

LGTM for the ppc side. But please hold on for a couple of days in case someone has other comments.

llvm/lib/Target/PowerPC/PPCISelLowering.cpp

PPCTargetLowering::PPCTargetLowering(const PPCTargetMachine &TM,
   setOperationAction(ISD::SMUL_LOHI, MVT::i32, Expand);
   setOperationAction(ISD::UMUL_LOHI, MVT::i64, Expand);
   setOperationAction(ISD::SMUL_LOHI, MVT::i64, Expand);
   setOperationAction(ISD::UDIVREM, MVT::i32, Expand);
   setOperationAction(ISD::SDIVREM, MVT::i32, Expand);
   setOperationAction(ISD::UDIVREM, MVT::i64, Expand);
   setOperationAction(ISD::SDIVREM, MVT::i64, Expand);
 
+  // Handle constrained floating-point operations of scalar.
+  // TODO: Handle SPE specific operation.
+  setOperationAction(ISD::STRICT_FADD, MVT::f32, Legal);
+  setOperationAction(ISD::STRICT_FSUB, MVT::f32, Legal);
+  setOperationAction(ISD::STRICT_FMUL, MVT::f32, Legal);
+  setOperationAction(ISD::STRICT_FDIV, MVT::f32, Legal);
+
+  setOperationAction(ISD::STRICT_FADD, MVT::f64, Legal);
+  setOperationAction(ISD::STRICT_FSUB, MVT::f64, Legal);
+  setOperationAction(ISD::STRICT_FMUL, MVT::f64, Legal);
+  setOperationAction(ISD::STRICT_FDIV, MVT::f64, Legal);
+
   // We don't support sin/cos/sqrt/fmod/pow
   setOperationAction(ISD::FSIN , MVT::f64, Expand);
   setOperationAction(ISD::FCOS , MVT::f64, Expand);
   setOperationAction(ISD::FSINCOS, MVT::f64, Expand);
   setOperationAction(ISD::FREM , MVT::f64, Expand);
   setOperationAction(ISD::FPOW , MVT::f64, Expand);
   setOperationAction(ISD::FSIN , MVT::f32, Expand);
   setOperationAction(ISD::FCOS , MVT::f32, Expand);
if (Subtarget.hasVSX()) {
       setOperationAction(ISD::FABS, MVT::v2f64, Legal);
       setOperationAction(ISD::FCOPYSIGN, MVT::v4f32, Legal);
       setOperationAction(ISD::FCOPYSIGN, MVT::v2f64, Legal);
 
       if (Subtarget.hasDirectMove())
         setOperationAction(ISD::BUILD_VECTOR, MVT::v2i64, Custom);
       setOperationAction(ISD::BUILD_VECTOR, MVT::v2f64, Custom);
 
+      // Handle constrained floating-point operations of vector.
+      // The predictor is `hasVSX` because altivec instruction has
+      // no exception but VSX vector instruction has.
+      setOperationAction(ISD::STRICT_FADD, MVT::v4f32, Legal);
+      setOperationAction(ISD::STRICT_FSUB, MVT::v4f32, Legal);
+      setOperationAction(ISD::STRICT_FMUL, MVT::v4f32, Legal);
+      setOperationAction(ISD::STRICT_FDIV, MVT::v4f32, Legal);
+
+      setOperationAction(ISD::STRICT_FADD, MVT::v2f64, Legal);
+      setOperationAction(ISD::STRICT_FSUB, MVT::v2f64, Legal);
+      setOperationAction(ISD::STRICT_FMUL, MVT::v2f64, Legal);
+      setOperationAction(ISD::STRICT_FDIV, MVT::v2f64, Legal);
+
       addRegisterClass(MVT::v2i64, &PPC::VSRCRegClass);
     }
 
     if (Subtarget.hasP8Altivec()) {
       addRegisterClass(MVT::v2i64, &PPC::VRRCRegClass);
       addRegisterClass(MVT::v1i128, &PPC::VRRCRegClass);
     }
 
if (Subtarget.hasP9Vector()) {
         setTruncStoreAction(MVT::f128, MVT::f32, Expand);
         setOperationAction(ISD::BITCAST, MVT::i128, Custom);
         // No implementation for these ops for PowerPC.
         setOperationAction(ISD::FSIN , MVT::f128, Expand);
         setOperationAction(ISD::FCOS , MVT::f128, Expand);
         setOperationAction(ISD::FPOW, MVT::f128, Expand);
         setOperationAction(ISD::FPOWI, MVT::f128, Expand);
         setOperationAction(ISD::FREM, MVT::f128, Expand);
+
+        // Handle constrained floating-point operations of fp128
+        setOperationAction(ISD::STRICT_FADD, MVT::f128, Legal);
+        setOperationAction(ISD::STRICT_FSUB, MVT::f128, Legal);
+        setOperationAction(ISD::STRICT_FMUL, MVT::f128, Legal);
+        setOperationAction(ISD::STRICT_FDIV, MVT::f128, Legal);
       }
       setOperationAction(ISD::FP_EXTEND, MVT::v2f32, Custom);
       setOperationAction(ISD::BSWAP, MVT::v8i16, Legal);
       setOperationAction(ISD::BSWAP, MVT::v4i32, Legal);
       setOperationAction(ISD::BSWAP, MVT::v2i64, Legal);
       setOperationAction(ISD::BSWAP, MVT::v1i128, Legal);
     }
 
if (TM.Options.UnsafeFPMath) {
       setOperationAction(ISD::FSQRT, MVT::v4f32, Legal);
     } else {
       setOperationAction(ISD::FDIV, MVT::v4f64, Expand);
       setOperationAction(ISD::FSQRT, MVT::v4f64, Expand);
 
       setOperationAction(ISD::FDIV, MVT::v4f32, Expand);
       setOperationAction(ISD::FSQRT, MVT::v4f32, Expand);
     }
+
+    // TODO: Handle constrained floating-point operations of v4f64
   }
 
   if (Subtarget.has64BitSupport())
     setOperationAction(ISD::PREFETCH, MVT::Other, Legal);
 
   setOperationAction(ISD::READCYCLECOUNTER, MVT::i64, isPPC64 ? Legal : Custom);
 
   if (!isPPC64) {

llvm/lib/Target/PowerPC/PPCInstrInfo.td

 defm FSELD : AForm_1r<63, 23,
                       (outs f8rc:$FRT), (ins f8rc:$FRA, f8rc:$FRC, f8rc:$FRB),
                       "fsel", "$FRT, $FRA, $FRC, $FRB", IIC_FPGeneral,
                       [(set f64:$FRT, (PPCfsel f64:$FRA, f64:$FRC, f64:$FRB))]>;
 defm FSELS : AForm_1r<63, 23,
                       (outs f4rc:$FRT), (ins f8rc:$FRA, f4rc:$FRC, f4rc:$FRB),
                       "fsel", "$FRT, $FRA, $FRC, $FRB", IIC_FPGeneral,
                       [(set f32:$FRT, (PPCfsel f64:$FRA, f32:$FRC, f32:$FRB))]>;
-let Uses = [RM] in {
+let Uses = [RM], mayRaiseFPException = 1 in {
   let isCommutable = 1 in {
   defm FADD  : AForm_2r<63, 21,
                         (outs f8rc:$FRT), (ins f8rc:$FRA, f8rc:$FRB),
                         "fadd", "$FRT, $FRA, $FRB", IIC_FPAddSub,
-                        [(set f64:$FRT, (fadd f64:$FRA, f64:$FRB))]>;
+                        [(set f64:$FRT, (any_fadd f64:$FRA, f64:$FRB))]>;
   defm FADDS : AForm_2r<59, 21,
                         (outs f4rc:$FRT), (ins f4rc:$FRA, f4rc:$FRB),
                         "fadds", "$FRT, $FRA, $FRB", IIC_FPGeneral,
-                        [(set f32:$FRT, (fadd f32:$FRA, f32:$FRB))]>;
+                        [(set f32:$FRT, (any_fadd f32:$FRA, f32:$FRB))]>;
   } // isCommutable
   defm FDIV  : AForm_2r<63, 18,
                         (outs f8rc:$FRT), (ins f8rc:$FRA, f8rc:$FRB),
                         "fdiv", "$FRT, $FRA, $FRB", IIC_FPDivD,
-                        [(set f64:$FRT, (fdiv f64:$FRA, f64:$FRB))]>;
+                        [(set f64:$FRT, (any_fdiv f64:$FRA, f64:$FRB))]>;
   defm FDIVS : AForm_2r<59, 18,
                         (outs f4rc:$FRT), (ins f4rc:$FRA, f4rc:$FRB),
                         "fdivs", "$FRT, $FRA, $FRB", IIC_FPDivS,
-                        [(set f32:$FRT, (fdiv f32:$FRA, f32:$FRB))]>;
+                        [(set f32:$FRT, (any_fdiv f32:$FRA, f32:$FRB))]>;
   let isCommutable = 1 in {
   defm FMUL  : AForm_3r<63, 25,
                         (outs f8rc:$FRT), (ins f8rc:$FRA, f8rc:$FRC),
                         "fmul", "$FRT, $FRA, $FRC", IIC_FPFused,
-                        [(set f64:$FRT, (fmul f64:$FRA, f64:$FRC))]>;
+                        [(set f64:$FRT, (any_fmul f64:$FRA, f64:$FRC))]>;
   defm FMULS : AForm_3r<59, 25,
                         (outs f4rc:$FRT), (ins f4rc:$FRA, f4rc:$FRC),
                         "fmuls", "$FRT, $FRA, $FRC", IIC_FPGeneral,
-                        [(set f32:$FRT, (fmul f32:$FRA, f32:$FRC))]>;
+                        [(set f32:$FRT, (any_fmul f32:$FRA, f32:$FRC))]>;
   } // isCommutable
   defm FSUB  : AForm_2r<63, 20,
                         (outs f8rc:$FRT), (ins f8rc:$FRA, f8rc:$FRB),
                         "fsub", "$FRT, $FRA, $FRB", IIC_FPAddSub,
-                        [(set f64:$FRT, (fsub f64:$FRA, f64:$FRB))]>;
+                        [(set f64:$FRT, (any_fsub f64:$FRA, f64:$FRB))]>;
   defm FSUBS : AForm_2r<59, 20,
                         (outs f4rc:$FRT), (ins f4rc:$FRA, f4rc:$FRB),
                         "fsubs", "$FRT, $FRA, $FRB", IIC_FPGeneral,
-                        [(set f32:$FRT, (fsub f32:$FRA, f32:$FRB))]>;
+                        [(set f32:$FRT, (any_fsub f32:$FRA, f32:$FRB))]>;
   }
 }
 
 let hasSideEffects = 0 in {
 let PPC970_Unit = 1 in {  // FXU Operations.
   let isSelect = 1 in
   def ISEL  : AForm_4<31, 15,
                      (outs gprc:$rT), (ins gprc_nor0:$rA, gprc:$rB, crbitrc:$cond),

llvm/lib/Target/PowerPC/PPCInstrVSX.td

let mayStore = 1, mayLoad = 0 in {
 
     def STXVW4X : XX1Form_memOp<31, 908,
                          (outs), (ins vsrc:$XT, memrr:$dst),
                          "stxvw4x $XT, $dst", IIC_LdStSTFD,
                          []>;
     }
   } // mayStore
 
-  let Uses = [RM] in {
+  let Uses = [RM], mayRaiseFPException = 1 in {
   // Add/Mul Instructions
   let isCommutable = 1 in {
     def XSADDDP : XX3Form<60, 32,
                           (outs vsfrc:$XT), (ins vsfrc:$XA, vsfrc:$XB),
                           "xsadddp $XT, $XA, $XB", IIC_VecFP,
-                          [(set f64:$XT, (fadd f64:$XA, f64:$XB))]>;
+                          [(set f64:$XT, (any_fadd f64:$XA, f64:$XB))]>;
     def XSMULDP : XX3Form<60, 48,
                           (outs vsfrc:$XT), (ins vsfrc:$XA, vsfrc:$XB),
                           "xsmuldp $XT, $XA, $XB", IIC_VecFP,
-                          [(set f64:$XT, (fmul f64:$XA, f64:$XB))]>;
+                          [(set f64:$XT, (any_fmul f64:$XA, f64:$XB))]>;
 
     def XVADDDP : XX3Form<60, 96,
                           (outs vsrc:$XT), (ins vsrc:$XA, vsrc:$XB),
                           "xvadddp $XT, $XA, $XB", IIC_VecFP,
-                          [(set v2f64:$XT, (fadd v2f64:$XA, v2f64:$XB))]>;
+                          [(set v2f64:$XT, (any_fadd v2f64:$XA, v2f64:$XB))]>;
 
     def XVADDSP : XX3Form<60, 64,
                           (outs vsrc:$XT), (ins vsrc:$XA, vsrc:$XB),
                           "xvaddsp $XT, $XA, $XB", IIC_VecFP,
-                          [(set v4f32:$XT, (fadd v4f32:$XA, v4f32:$XB))]>;
+                          [(set v4f32:$XT, (any_fadd v4f32:$XA, v4f32:$XB))]>;
 
     def XVMULDP : XX3Form<60, 112,
                           (outs vsrc:$XT), (ins vsrc:$XA, vsrc:$XB),
                           "xvmuldp $XT, $XA, $XB", IIC_VecFP,
-                          [(set v2f64:$XT, (fmul v2f64:$XA, v2f64:$XB))]>;
+                          [(set v2f64:$XT, (any_fmul v2f64:$XA, v2f64:$XB))]>;
 
     def XVMULSP : XX3Form<60, 80,
                           (outs vsrc:$XT), (ins vsrc:$XA, vsrc:$XB),
                           "xvmulsp $XT, $XA, $XB", IIC_VecFP,
-                          [(set v4f32:$XT, (fmul v4f32:$XA, v4f32:$XB))]>;
+                          [(set v4f32:$XT, (any_fmul v4f32:$XA, v4f32:$XB))]>;
   }
 
   // Subtract Instructions
   def XSSUBDP : XX3Form<60, 40,
                         (outs vsfrc:$XT), (ins vsfrc:$XA, vsfrc:$XB),
                         "xssubdp $XT, $XA, $XB", IIC_VecFP,
-                        [(set f64:$XT, (fsub f64:$XA, f64:$XB))]>;
+                        [(set f64:$XT, (any_fsub f64:$XA, f64:$XB))]>;
 
   def XVSUBDP : XX3Form<60, 104,
                         (outs vsrc:$XT), (ins vsrc:$XA, vsrc:$XB),
                         "xvsubdp $XT, $XA, $XB", IIC_VecFP,
-                        [(set v2f64:$XT, (fsub v2f64:$XA, v2f64:$XB))]>;
+                        [(set v2f64:$XT, (any_fsub v2f64:$XA, v2f64:$XB))]>;
   def XVSUBSP : XX3Form<60, 72,
                         (outs vsrc:$XT), (ins vsrc:$XA, vsrc:$XB),
                         "xvsubsp $XT, $XA, $XB", IIC_VecFP,
-                        [(set v4f32:$XT, (fsub v4f32:$XA, v4f32:$XB))]>;
+                        [(set v4f32:$XT, (any_fsub v4f32:$XA, v4f32:$XB))]>;
 
   // FMA Instructions
   let BaseName = "XSMADDADP" in {
   let isCommutable = 1 in
   def XSMADDADP : XX3Form<60, 33,
                           (outs vsfrc:$XT), (ins vsfrc:$XTi, vsfrc:$XA, vsfrc:$XB),
                           "xsmaddadp $XT, $XA, $XB", IIC_VecFP,
                           [(set f64:$XT, (fma f64:$XA, f64:$XB, f64:$XTi))]>,
def XVNMSUBMSP : XX3Form<60, 217,
                           RegConstraint<"$XTi = $XT">, NoEncode<"$XTi">,
                           AltVSXFMARel;
   }
 
   // Division Instructions
   def XSDIVDP : XX3Form<60, 56,
                         (outs vsfrc:$XT), (ins vsfrc:$XA, vsfrc:$XB),
                         "xsdivdp $XT, $XA, $XB", IIC_FPDivD,
-                        [(set f64:$XT, (fdiv f64:$XA, f64:$XB))]>;
+                        [(set f64:$XT, (any_fdiv f64:$XA, f64:$XB))]>;
   def XSSQRTDP : XX2Form<60, 75,
                         (outs vsfrc:$XT), (ins vsfrc:$XB),
                         "xssqrtdp $XT, $XB", IIC_FPSqrtD,
                         [(set f64:$XT, (fsqrt f64:$XB))]>;
 
   def XSREDP : XX2Form<60, 90,
                         (outs vsfrc:$XT), (ins vsfrc:$XB),
                         "xsredp $XT, $XB", IIC_VecFP,
                         [(set f64:$XT, (PPCfre f64:$XB))]>;
   def XSRSQRTEDP : XX2Form<60, 74,
                            (outs vsfrc:$XT), (ins vsfrc:$XB),
                            "xsrsqrtedp $XT, $XB", IIC_VecFP,
                            [(set f64:$XT, (PPCfrsqrte f64:$XB))]>;
 
   def XSTDIVDP : XX3Form_1<60, 61,
                          (outs crrc:$crD), (ins vsfrc:$XA, vsfrc:$XB),
                          "xstdivdp $crD, $XA, $XB", IIC_FPCompare, []>;
   def XSTSQRTDP : XX2Form_1<60, 106,
                           (outs crrc:$crD), (ins vsfrc:$XB),
                           "xstsqrtdp $crD, $XB", IIC_FPCompare, []>;
 
   def XVDIVDP : XX3Form<60, 120,
                         (outs vsrc:$XT), (ins vsrc:$XA, vsrc:$XB),
                         "xvdivdp $XT, $XA, $XB", IIC_FPDivD,
-                        [(set v2f64:$XT, (fdiv v2f64:$XA, v2f64:$XB))]>;
+                        [(set v2f64:$XT, (any_fdiv v2f64:$XA, v2f64:$XB))]>;
   def XVDIVSP : XX3Form<60, 88,
                         (outs vsrc:$XT), (ins vsrc:$XA, vsrc:$XB),
                         "xvdivsp $XT, $XA, $XB", IIC_FPDivS,
-                        [(set v4f32:$XT, (fdiv v4f32:$XA, v4f32:$XB))]>;
+                        [(set v4f32:$XT, (any_fdiv v4f32:$XA, v4f32:$XB))]>;
 
   def XVSQRTDP : XX2Form<60, 203,
                         (outs vsrc:$XT), (ins vsrc:$XB),
                         "xvsqrtdp $XT, $XB", IIC_FPSqrtD,
                         [(set v2f64:$XT, (fsqrt v2f64:$XB))]>;
   def XVSQRTSP : XX2Form<60, 139,
                         (outs vsrc:$XT), (ins vsrc:$XB),
                         "xvsqrtsp $XT, $XB", IIC_FPSqrtS,
def XVMAXSP : XX3Form<60, 192,
                         [(set vsrc:$XT,
                               (int_ppc_vsx_xvmaxsp vsrc:$XA, vsrc:$XB))]>;
   def XVMINSP : XX3Form<60, 200,
                         (outs vsrc:$XT), (ins vsrc:$XA, vsrc:$XB),
                         "xvminsp $XT, $XA, $XB", IIC_VecFP,
                         [(set vsrc:$XT,
                               (int_ppc_vsx_xvminsp vsrc:$XA, vsrc:$XB))]>;
   } // isCommutable
-} // Uses = [RM]
+} // Uses = [RM], mayRaiseFPException
 
   // Logical Instructions
   let isCommutable = 1 in
   def XXLAND : XX3Form<60, 130,
                        (outs vsrc:$XT), (ins vsrc:$XA, vsrc:$XB),
                        "xxland $XT, $XA, $XB", IIC_VecGeneral,
                        [(set v4i32:$XT, (and v4i32:$XA, v4i32:$XB))]>;
   def XXLANDC : XX3Form<60, 138,
let AddedComplexity = 400 in { // Prefer VSX patterns over non-VSX patterns.
   def : Pat<(f32 (selectcc i1:$lhs, i1:$rhs, f32:$tval, f32:$fval, SETGT)),
             (SELECT_VSSRC (CRANDC $rhs, $lhs), $tval, $fval)>;
   def : Pat<(f32 (selectcc i1:$lhs, i1:$rhs, f32:$tval, f32:$fval, SETUGT)),
             (SELECT_VSSRC (CRANDC $lhs, $rhs), $tval, $fval)>;
   def : Pat<(f32 (selectcc i1:$lhs, i1:$rhs, f32:$tval, f32:$fval, SETNE)),
             (SELECT_VSSRC (CRXOR $lhs, $rhs), $tval, $fval)>;
 
   // VSX Elementary Scalar FP arithmetic (SP)
+  let mayRaiseFPException = 1 in {
   let isCommutable = 1 in {
     def XSADDSP : XX3Form<60, 0,
                           (outs vssrc:$XT), (ins vssrc:$XA, vssrc:$XB),
                           "xsaddsp $XT, $XA, $XB", IIC_VecFP,
-                          [(set f32:$XT, (fadd f32:$XA, f32:$XB))]>;
+                          [(set f32:$XT, (any_fadd f32:$XA, f32:$XB))]>;
     def XSMULSP : XX3Form<60, 16,
                           (outs vssrc:$XT), (ins vssrc:$XA, vssrc:$XB),
                           "xsmulsp $XT, $XA, $XB", IIC_VecFP,
-                          [(set f32:$XT, (fmul f32:$XA, f32:$XB))]>;
+                          [(set f32:$XT, (any_fmul f32:$XA, f32:$XB))]>;
   } // isCommutable
   def XSSUBSP : XX3Form<60, 8,
                         (outs vssrc:$XT), (ins vssrc:$XA, vssrc:$XB),
                         "xssubsp $XT, $XA, $XB", IIC_VecFP,
-                        [(set f32:$XT, (fsub f32:$XA, f32:$XB))]>;
+                        [(set f32:$XT, (any_fsub f32:$XA, f32:$XB))]>;
   def XSDIVSP : XX3Form<60, 24,
                         (outs vssrc:$XT), (ins vssrc:$XA, vssrc:$XB),
                         "xsdivsp $XT, $XA, $XB", IIC_FPDivS,
-                        [(set f32:$XT, (fdiv f32:$XA, f32:$XB))]>;
+                        [(set f32:$XT, (any_fdiv f32:$XA, f32:$XB))]>;
+  } // mayRaiseFPException
   def XSRESP : XX2Form<60, 26,
                         (outs vssrc:$XT), (ins vssrc:$XB),
                         "xsresp $XT, $XB", IIC_VecFP,
                         [(set f32:$XT, (PPCfre f32:$XB))]>;
   def XSRSP : XX2Form<60, 281,
                         (outs vssrc:$XT), (ins vsfrc:$XB),
                         "xsrsp $XT, $XB", IIC_VecFP, []>;
   def XSSQRTSP : XX2Form<60, 11,
def XSNABSQP  : X_VT5_XO5_VB5<63,  8, 804, "xsnabsqp",
                                 [(set f128:$vT, (fneg (fabs f128:$vB)))]>;
   def XSNEGQP   : X_VT5_XO5_VB5<63, 16, 804, "xsnegqp",
                                 [(set f128:$vT, (fneg f128:$vB))]>;
 
   //===--------------------------------------------------------------------===//
   // Quad-Precision Scalar Floating-Point Arithmetic Instructions:
 
   // Add/Divide/Multiply/Subtract
+  let mayRaiseFPException = 1 in {
   let isCommutable = 1 in {
   def XSADDQP   : X_VT5_VA5_VB5   <63,   4, "xsaddqp",
-                                   [(set f128:$vT, (fadd f128:$vA, f128:$vB))]>;
+                                   [(set f128:$vT, (any_fadd f128:$vA, f128:$vB))]>;
   def XSMULQP   : X_VT5_VA5_VB5   <63,  36, "xsmulqp",
-                                   [(set f128:$vT, (fmul f128:$vA, f128:$vB))]>;
+                                   [(set f128:$vT, (any_fmul f128:$vA, f128:$vB))]>;
   }
   def XSSUBQP   : X_VT5_VA5_VB5   <63, 516, "xssubqp" ,
-                                   [(set f128:$vT, (fsub f128:$vA, f128:$vB))]>;
+                                   [(set f128:$vT, (any_fsub f128:$vA, f128:$vB))]>;
   def XSDIVQP   : X_VT5_VA5_VB5   <63, 548, "xsdivqp",
-                                   [(set f128:$vT, (fdiv f128:$vA, f128:$vB))]>;
+                                   [(set f128:$vT, (any_fdiv f128:$vA, f128:$vB))]>;
+  } // mayRaiseFPException
   // Square-Root
   def XSSQRTQP  : X_VT5_XO5_VB5   <63, 27, 804, "xssqrtqp",
                                    [(set f128:$vT, (fsqrt f128:$vB))]>;
   // (Negative) Multiply-{Add/Subtract}
   def XSMADDQP : X_VT5_VA5_VB5_FMA <63, 388, "xsmaddqp",
                                     [(set f128:$vT,
                                           (fma f128:$vA, f128:$vB,
                                                f128:$vTi))]>;

llvm/test/CodeGen/PowerPC/fp-strict-f128.ll

+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc -verify-machineinstrs -ppc-asm-full-reg-names -ppc-vsr-nums-as-vr < %s -mtriple=powerpc64le-unknown-linux -mcpu=pwr9 -enable-ppc-quad-precision=true | FileCheck %s
+
+declare fp128 @llvm.experimental.constrained.fadd.f128(fp128, fp128, metadata, metadata)
+declare fp128 @llvm.experimental.constrained.fsub.f128(fp128, fp128, metadata, metadata)
+declare fp128 @llvm.experimental.constrained.fmul.f128(fp128, fp128, metadata, metadata)
+declare fp128 @llvm.experimental.constrained.fdiv.f128(fp128, fp128, metadata, metadata)
+
+define fp128 @fadd_f128(fp128 %f1, fp128 %f2) {
+; CHECK-LABEL: fadd_f128:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    xsaddqp v2, v2, v3
+; CHECK-NEXT:    blr
+  %res = call fp128 @llvm.experimental.constrained.fadd.f128(
+                        fp128 %f1, fp128 %f2,
+                        metadata !"round.dynamic",
+                        metadata !"fpexcept.strict")
+  ret fp128 %res
+}
+
+define fp128 @fsub_f128(fp128 %f1, fp128 %f2) {
+; CHECK-LABEL: fsub_f128:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    xssubqp v2, v2, v3
+; CHECK-NEXT:    blr
+  %res = call fp128 @llvm.experimental.constrained.fsub.f128(
+                        fp128 %f1, fp128 %f2,
+                        metadata !"round.dynamic",
+                        metadata !"fpexcept.strict")
+  ret fp128 %res
+}
+
+define fp128 @fmul_f128(fp128 %f1, fp128 %f2) {
+; CHECK-LABEL: fmul_f128:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    xsmulqp v2, v2, v3
+; CHECK-NEXT:    blr
+  %res = call fp128 @llvm.experimental.constrained.fmul.f128(
+                        fp128 %f1, fp128 %f2,
+                        metadata !"round.dynamic",
+                        metadata !"fpexcept.strict")
+  ret fp128 %res
+}
+
+define fp128 @fdiv_f128(fp128 %f1, fp128 %f2) {
+; CHECK-LABEL: fdiv_f128:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    xsdivqp v2, v2, v3
+; CHECK-NEXT:    blr
+  %res = call fp128 @llvm.experimental.constrained.fdiv.f128(
+                        fp128 %f1, fp128 %f2,
+                        metadata !"round.dynamic",
+                        metadata !"fpexcept.strict")
+  ret fp128 %res
+}

llvm/test/CodeGen/PowerPC/fp-strict.ll

+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc -verify-machineinstrs -ppc-asm-full-reg-names -ppc-vsr-nums-as-vr < %s -mtriple=powerpc64-unknown-linux -mcpu=pwr8 | FileCheck %s
+; RUN: llc -verify-machineinstrs -ppc-asm-full-reg-names -ppc-vsr-nums-as-vr < %s -mtriple=powerpc64le-unknown-linux -mcpu=pwr9 | FileCheck %s
+; RUN: llc -verify-machineinstrs -ppc-asm-full-reg-names -ppc-vsr-nums-as-vr < %s -mtriple=powerpc64le-unknown-linux -mcpu=pwr8 -mattr=-vsx | FileCheck %s -check-prefix=NOVSX
+
+declare float @llvm.experimental.constrained.fadd.f32(float, float, metadata, metadata)
+declare double @llvm.experimental.constrained.fadd.f64(double, double, metadata, metadata)
+declare <4 x float> @llvm.experimental.constrained.fadd.v4f32(<4 x float>, <4 x float>, metadata, metadata)
+declare <2 x double> @llvm.experimental.constrained.fadd.v2f64(<2 x double>, <2 x double>, metadata, metadata)
+
+declare float @llvm.experimental.constrained.fsub.f32(float, float, metadata, metadata)
+declare double @llvm.experimental.constrained.fsub.f64(double, double, metadata, metadata)
+declare <4 x float> @llvm.experimental.constrained.fsub.v4f32(<4 x float>, <4 x float>, metadata, metadata)
+declare <2 x double> @llvm.experimental.constrained.fsub.v2f64(<2 x double>, <2 x double>, metadata, metadata)
+
+declare float @llvm.experimental.constrained.fmul.f32(float, float, metadata, metadata)
+declare double @llvm.experimental.constrained.fmul.f64(double, double, metadata, metadata)
+declare <4 x float> @llvm.experimental.constrained.fmul.v4f32(<4 x float>, <4 x float>, metadata, metadata)
+declare <2 x double> @llvm.experimental.constrained.fmul.v2f64(<2 x double>, <2 x double>, metadata, metadata)
+
+declare float @llvm.experimental.constrained.fdiv.f32(float, float, metadata, metadata)
+declare double @llvm.experimental.constrained.fdiv.f64(double, double, metadata, metadata)
+declare <4 x float> @llvm.experimental.constrained.fdiv.v4f32(<4 x float>, <4 x float>, metadata, metadata)
+declare <2 x double> @llvm.experimental.constrained.fdiv.v2f64(<2 x double>, <2 x double>, metadata, metadata)
+
+define float @fadd_f32(float %f1, float %f2) {
+; CHECK-LABEL: fadd_f32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    xsaddsp f1, f1, f2
+; CHECK-NEXT:    blr
+;
+; NOVSX-LABEL: fadd_f32:
+; NOVSX:       # %bb.0:
+; NOVSX-NEXT:    fadds f1, f1, f2
+; NOVSX-NEXT:    blr
+  %res = call float @llvm.experimental.constrained.fadd.f32(
+                        float %f1, float %f2,
+                        metadata !"round.dynamic",
+                        metadata !"fpexcept.strict")
+  ret float %res
+}
+
+define double @fadd_f64(double %f1, double %f2) {
+; CHECK-LABEL: fadd_f64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    xsadddp f1, f1, f2
+; CHECK-NEXT:    blr
+;
+; NOVSX-LABEL: fadd_f64:
+; NOVSX:       # %bb.0:
+; NOVSX-NEXT:    fadd f1, f1, f2
+; NOVSX-NEXT:    blr
+  %res = call double @llvm.experimental.constrained.fadd.f64(
+                        double %f1, double %f2,
+                        metadata !"round.dynamic",
+                        metadata !"fpexcept.strict")
+  ret double %res
+}
+
+define <4 x float> @fadd_v4f32(<4 x float> %vf1, <4 x float> %vf2) {
+; CHECK-LABEL: fadd_v4f32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    xvaddsp v2, v2, v3
+; CHECK-NEXT:    blr
+;
+; NOVSX-LABEL: fadd_v4f32:
+; NOVSX:       # %bb.0:
+; NOVSX-NEXT:    addi r3, r1, -32
+; NOVSX-NEXT:    addi r4, r1, -48
+; NOVSX-NEXT:    stvx v3, 0, r3
+; NOVSX-NEXT:    stvx v2, 0, r4
+; NOVSX-NEXT:    addi r3, r1, -16
+; NOVSX-NEXT:    lfs f0, -20(r1)
+; NOVSX-NEXT:    lfs f1, -36(r1)
+; NOVSX-NEXT:    fadds f0, f1, f0
+; NOVSX-NEXT:    lfs f1, -40(r1)
+; NOVSX-NEXT:    stfs f0, -4(r1)
+; NOVSX-NEXT:    lfs f0, -24(r1)
+; NOVSX-NEXT:    fadds f0, f1, f0
+; NOVSX-NEXT:    lfs f1, -44(r1)
+; NOVSX-NEXT:    stfs f0, -8(r1)
+; NOVSX-NEXT:    lfs f0, -28(r1)
+; NOVSX-NEXT:    fadds f0, f1, f0
+; NOVSX-NEXT:    lfs f1, -48(r1)
+; NOVSX-NEXT:    stfs f0, -12(r1)
+; NOVSX-NEXT:    lfs f0, -32(r1)
+; NOVSX-NEXT:    fadds f0, f1, f0
+; NOVSX-NEXT:    stfs f0, -16(r1)
+; NOVSX-NEXT:    lvx v2, 0, r3
+; NOVSX-NEXT:    blr
+  %res = call <4 x float> @llvm.experimental.constrained.fadd.v4f32(
+                        <4 x float> %vf1, <4 x float> %vf2,
+                        metadata !"round.dynamic",
+                        metadata !"fpexcept.strict")
+  ret <4 x float> %res
+}
+
+define <2 x double> @fadd_v2f64(<2 x double> %vf1, <2 x double> %vf2) {
+; CHECK-LABEL: fadd_v2f64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    xvadddp v2, v2, v3
+; CHECK-NEXT:    blr
+;
+; NOVSX-LABEL: fadd_v2f64:
+; NOVSX:       # %bb.0:
+; NOVSX-NEXT:    fadd f2, f2, f4
+; NOVSX-NEXT:    fadd f1, f1, f3
+; NOVSX-NEXT:    blr
+  %res = call <2 x double> @llvm.experimental.constrained.fadd.v2f64(
+                        <2 x double> %vf1, <2 x double> %vf2,
+                        metadata !"round.dynamic",
+                        metadata !"fpexcept.strict")
+  ret <2 x double> %res
+}
+
+define float @fsub_f32(float %f1, float %f2) {
+; CHECK-LABEL: fsub_f32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    xssubsp f1, f1, f2
+; CHECK-NEXT:    blr
+;
+; NOVSX-LABEL: fsub_f32:
+; NOVSX:       # %bb.0:
+; NOVSX-NEXT:    fsubs f1, f1, f2
+; NOVSX-NEXT:    blr
+
+  %res = call float @llvm.experimental.constrained.fsub.f32(
+                        float %f1, float %f2,
+                        metadata !"round.dynamic",
+                        metadata !"fpexcept.strict")
+  ret float %res;
+}
+
+define double @fsub_f64(double %f1, double %f2) {
+; CHECK-LABEL: fsub_f64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    xssubdp f1, f1, f2
+; CHECK-NEXT:    blr
+;
+; NOVSX-LABEL: fsub_f64:
+; NOVSX:       # %bb.0:
+; NOVSX-NEXT:    fsub f1, f1, f2
+; NOVSX-NEXT:    blr
+
+  %res = call double @llvm.experimental.constrained.fsub.f64(
+                        double %f1, double %f2,
+                        metadata !"round.dynamic",
+                        metadata !"fpexcept.strict")
+  ret double %res;
+}
+
+define <4 x float> @fsub_v4f32(<4 x float> %vf1, <4 x float> %vf2) {
+; CHECK-LABEL: fsub_v4f32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    xvsubsp v2, v2, v3
+; CHECK-NEXT:    blr
+;
+; NOVSX-LABEL: fsub_v4f32:
+; NOVSX:       # %bb.0:
+; NOVSX-NEXT:    addi r3, r1, -32
+; NOVSX-NEXT:    addi r4, r1, -48
+; NOVSX-NEXT:    stvx v3, 0, r3
+; NOVSX-NEXT:    stvx v2, 0, r4
+; NOVSX-NEXT:    addi r3, r1, -16
+; NOVSX-NEXT:    lfs f0, -20(r1)
+; NOVSX-NEXT:    lfs f1, -36(r1)
+; NOVSX-NEXT:    fsubs f0, f1, f0
+; NOVSX-NEXT:    lfs f1, -40(r1)
+; NOVSX-NEXT:    stfs f0, -4(r1)
+; NOVSX-NEXT:    lfs f0, -24(r1)
+; NOVSX-NEXT:    fsubs f0, f1, f0
+; NOVSX-NEXT:    lfs f1, -44(r1)
+; NOVSX-NEXT:    stfs f0, -8(r1)
+; NOVSX-NEXT:    lfs f0, -28(r1)
+; NOVSX-NEXT:    fsubs f0, f1, f0
+; NOVSX-NEXT:    lfs f1, -48(r1)
+; NOVSX-NEXT:    stfs f0, -12(r1)
+; NOVSX-NEXT:    lfs f0, -32(r1)
+; NOVSX-NEXT:    fsubs f0, f1, f0
+; NOVSX-NEXT:    stfs f0, -16(r1)
+; NOVSX-NEXT:    lvx v2, 0, r3
+; NOVSX-NEXT:    blr
+  %res = call <4 x float> @llvm.experimental.constrained.fsub.v4f32(
+                        <4 x float> %vf1, <4 x float> %vf2,
+                        metadata !"round.dynamic",
+                        metadata !"fpexcept.strict")
+  ret <4 x float> %res;
+}
+
+define <2 x double> @fsub_v2f64(<2 x double> %vf1, <2 x double> %vf2) {
+; CHECK-LABEL: fsub_v2f64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    xvsubdp v2, v2, v3
+; CHECK-NEXT:    blr
+;
+; NOVSX-LABEL: fsub_v2f64:
+; NOVSX:       # %bb.0:
+; NOVSX-NEXT:    fsub f2, f2, f4
+; NOVSX-NEXT:    fsub f1, f1, f3
+; NOVSX-NEXT:    blr
+  %res = call <2 x double> @llvm.experimental.constrained.fsub.v2f64(
+                        <2 x double> %vf1, <2 x double> %vf2,
+                        metadata !"round.dynamic",
+                        metadata !"fpexcept.strict")
+  ret <2 x double> %res;
+}
+
+define float @fmul_f32(float %f1, float %f2) {
+; CHECK-LABEL: fmul_f32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    xsmulsp f1, f1, f2
+; CHECK-NEXT:    blr
+;
+; NOVSX-LABEL: fmul_f32:
+; NOVSX:       # %bb.0:
+; NOVSX-NEXT:    fmuls f1, f1, f2
+; NOVSX-NEXT:    blr
+
+  %res = call float @llvm.experimental.constrained.fmul.f32(
+                        float %f1, float %f2,
+                        metadata !"round.dynamic",
+                        metadata !"fpexcept.strict")
+  ret float %res;
+}
+
+define double @fmul_f64(double %f1, double %f2) {
+; CHECK-LABEL: fmul_f64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    xsmuldp f1, f1, f2
+; CHECK-NEXT:    blr
+;
+; NOVSX-LABEL: fmul_f64:
+; NOVSX:       # %bb.0:
+; NOVSX-NEXT:    fmul f1, f1, f2
+; NOVSX-NEXT:    blr
+
+  %res = call double @llvm.experimental.constrained.fmul.f64(
+                        double %f1, double %f2,
+                        metadata !"round.dynamic",
+                        metadata !"fpexcept.strict")
+  ret double %res;
+}
+
+define <4 x float> @fmul_v4f32(<4 x float> %vf1, <4 x float> %vf2) {
+; CHECK-LABEL: fmul_v4f32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    xvmulsp v2, v2, v3
+; CHECK-NEXT:    blr
+;
+; NOVSX-LABEL: fmul_v4f32:
+; NOVSX:       # %bb.0:
+; NOVSX-NEXT:    addi r3, r1, -32
+; NOVSX-NEXT:    addi r4, r1, -48
+; NOVSX-NEXT:    stvx v3, 0, r3
+; NOVSX-NEXT:    stvx v2, 0, r4
+; NOVSX-NEXT:    addi r3, r1, -16
+; NOVSX-NEXT:    lfs f0, -20(r1)
+; NOVSX-NEXT:    lfs f1, -36(r1)
+; NOVSX-NEXT:    fmuls f0, f1, f0
+; NOVSX-NEXT:    lfs f1, -40(r1)
+; NOVSX-NEXT:    stfs f0, -4(r1)
+; NOVSX-NEXT:    lfs f0, -24(r1)
+; NOVSX-NEXT:    fmuls f0, f1, f0
+; NOVSX-NEXT:    lfs f1, -44(r1)
+; NOVSX-NEXT:    stfs f0, -8(r1)
+; NOVSX-NEXT:    lfs f0, -28(r1)
+; NOVSX-NEXT:    fmuls f0, f1, f0
+; NOVSX-NEXT:    lfs f1, -48(r1)
+; NOVSX-NEXT:    stfs f0, -12(r1)
+; NOVSX-NEXT:    lfs f0, -32(r1)
+; NOVSX-NEXT:    fmuls f0, f1, f0
+; NOVSX-NEXT:    stfs f0, -16(r1)
+; NOVSX-NEXT:    lvx v2, 0, r3
+; NOVSX-NEXT:    blr
+  %res = call <4 x float> @llvm.experimental.constrained.fmul.v4f32(
+                        <4 x float> %vf1, <4 x float> %vf2,
+                        metadata !"round.dynamic",
+                        metadata !"fpexcept.strict")
+  ret <4 x float> %res;
+}
+
+define <2 x double> @fmul_v2f64(<2 x double> %vf1, <2 x double> %vf2) {
+; CHECK-LABEL: fmul_v2f64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    xvmuldp v2, v2, v3
+; CHECK-NEXT:    blr
+;
+; NOVSX-LABEL: fmul_v2f64:
+; NOVSX:       # %bb.0:
+; NOVSX-NEXT:    fmul f2, f2, f4
+; NOVSX-NEXT:    fmul f1, f1, f3
+; NOVSX-NEXT:    blr
+  %res = call <2 x double> @llvm.experimental.constrained.fmul.v2f64(
+                        <2 x double> %vf1, <2 x double> %vf2,
+                        metadata !"round.dynamic",
+                        metadata !"fpexcept.strict")
+  ret <2 x double> %res;
+}
+
+define float @fdiv_f32(float %f1, float %f2) {
+; CHECK-LABEL: fdiv_f32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    xsdivsp f1, f1, f2
+; CHECK-NEXT:    blr
+;
+; NOVSX-LABEL: fdiv_f32:
+; NOVSX:       # %bb.0:
+; NOVSX-NEXT:    fdivs f1, f1, f2
+; NOVSX-NEXT:    blr
+
+  %res = call float @llvm.experimental.constrained.fdiv.f32(
+                        float %f1, float %f2,
+                        metadata !"round.dynamic",
+                        metadata !"fpexcept.strict")
+  ret float %res;
+}
+
+define double @fdiv_f64(double %f1, double %f2) {
+; CHECK-LABEL: fdiv_f64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    xsdivdp f1, f1, f2
+; CHECK-NEXT:    blr
+;
+; NOVSX-LABEL: fdiv_f64:
+; NOVSX:       # %bb.0:
+; NOVSX-NEXT:    fdiv f1, f1, f2
+; NOVSX-NEXT:    blr
+
+  %res = call double @llvm.experimental.constrained.fdiv.f64(
+                        double %f1, double %f2,
+                        metadata !"round.dynamic",
+                        metadata !"fpexcept.strict")
+  ret double %res;
+}
+
+define <4 x float> @fdiv_v4f32(<4 x float> %vf1, <4 x float> %vf2) {
+; CHECK-LABEL: fdiv_v4f32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    xvdivsp v2, v2, v3
+; CHECK-NEXT:    blr
+;
+; NOVSX-LABEL: fdiv_v4f32:
+; NOVSX:       # %bb.0:
+; NOVSX-NEXT:    addi r3, r1, -32
+; NOVSX-NEXT:    addi r4, r1, -48
+; NOVSX-NEXT:    stvx v3, 0, r3
+; NOVSX-NEXT:    stvx v2, 0, r4
+; NOVSX-NEXT:    addi r3, r1, -16
+; NOVSX-NEXT:    lfs f0, -20(r1)
+; NOVSX-NEXT:    lfs f1, -36(r1)
+; NOVSX-NEXT:    fdivs f0, f1, f0
+; NOVSX-NEXT:    lfs f1, -40(r1)
+; NOVSX-NEXT:    stfs f0, -4(r1)
+; NOVSX-NEXT:    lfs f0, -24(r1)
+; NOVSX-NEXT:    fdivs f0, f1, f0
+; NOVSX-NEXT:    lfs f1, -44(r1)
+; NOVSX-NEXT:    stfs f0, -8(r1)
+; NOVSX-NEXT:    lfs f0, -28(r1)
+; NOVSX-NEXT:    fdivs f0, f1, f0
+; NOVSX-NEXT:    lfs f1, -48(r1)
+; NOVSX-NEXT:    stfs f0, -12(r1)
+; NOVSX-NEXT:    lfs f0, -32(r1)
+; NOVSX-NEXT:    fdivs f0, f1, f0
+; NOVSX-NEXT:    stfs f0, -16(r1)
+; NOVSX-NEXT:    lvx v2, 0, r3
+; NOVSX-NEXT:    blr
+  %res = call <4 x float> @llvm.experimental.constrained.fdiv.v4f32(
+                        <4 x float> %vf1, <4 x float> %vf2,
+                        metadata !"round.dynamic",
+                        metadata !"fpexcept.strict")
+  ret <4 x float> %res
+}
+
+define <2 x double> @fdiv_v2f64(<2 x double> %vf1, <2 x double> %vf2) {
+; CHECK-LABEL: fdiv_v2f64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    xvdivdp v2, v2, v3
+; CHECK-NEXT:    blr
+;
+; NOVSX-LABEL: fdiv_v2f64:
+; NOVSX:       # %bb.0:
+; NOVSX-NEXT:    fdiv f2, f2, f4
+; NOVSX-NEXT:    fdiv f1, f1, f3
+; NOVSX-NEXT:    blr
+  %res = call <2 x double> @llvm.experimental.constrained.fdiv.v2f64(
+                        <2 x double> %vf1, <2 x double> %vf2,
+                        metadata !"round.dynamic",
+                        metadata !"fpexcept.strict")
+  ret <2 x double> %res
+}
+
+define double @no_fma_fold(double %f1, double %f2, double %f3) {
+; CHECK-LABEL: no_fma_fold:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    xsmuldp f0, f1, f2
+; CHECK-NEXT:    xsadddp f1, f0, f3
+; CHECK-NEXT:    blr
+;
+; NOVSX-LABEL: no_fma_fold:
+; NOVSX:       # %bb.0:
+; NOVSX-NEXT:    fmul f0, f1, f2
+; NOVSX-NEXT:    fadd f1, f0, f3
+; NOVSX-NEXT:    blr
+  %mul = call double @llvm.experimental.constrained.fmul.f64(
+                        double %f1, double %f2,
+                        metadata !"round.dynamic",
+                        metadata !"fpexcept.strict")
+  %add = call double @llvm.experimental.constrained.fadd.f64(
+                        double %mul, double %f3,
+                        metadata !"round.dynamic",
+                        metadata !"fpexcept.strict")
+  ret double %add
+}

llvm/test/CodeGen/PowerPC/vector-constrained-fp-intrinsics.ll

 ; PC64LE-NEXT:    lxvd2x 0, 0, 3
 ; PC64LE-NEXT:    addi 3, 4, .LCPI4_1@toc@l
 ; PC64LE-NEXT:    addi 4, 5, .LCPI4_2@toc@l
 ; PC64LE-NEXT:    lxvd2x 1, 0, 3
 ; PC64LE-NEXT:    lxvd2x 2, 0, 4
 ; PC64LE-NEXT:    xxswapd 0, 0
 ; PC64LE-NEXT:    xxswapd 1, 1
 ; PC64LE-NEXT:    xxswapd 2, 2
-; PC64LE-NEXT:    xvdivdp 34, 1, 0
-; PC64LE-NEXT:    xvdivdp 35, 2, 0
+; PC64LE-NEXT:    xvdivdp 35, 1, 0
+; PC64LE-NEXT:    xvdivdp 34, 2, 0
 ; PC64LE-NEXT:    blr
 ;
 ; PC64LE9-LABEL: constrained_vector_fdiv_v4f64:
 ; PC64LE9:       # %bb.0: # %entry
 ; PC64LE9-NEXT:    addis 3, 2, .LCPI4_0@toc@ha
 ; PC64LE9-NEXT:    addi 3, 3, .LCPI4_0@toc@l
 ; PC64LE9-NEXT:    lxvx 0, 0, 3
 ; PC64LE9-NEXT:    addis 3, 2, .LCPI4_1@toc@ha
 ; PC64LE9-NEXT:    addi 3, 3, .LCPI4_1@toc@l
 ; PC64LE9-NEXT:    lxvx 1, 0, 3
 ; PC64LE9-NEXT:    addis 3, 2, .LCPI4_2@toc@ha
 ; PC64LE9-NEXT:    addi 3, 3, .LCPI4_2@toc@l
-; PC64LE9-NEXT:    xvdivdp 34, 1, 0
-; PC64LE9-NEXT:    lxvx 1, 0, 3
 ; PC64LE9-NEXT:    xvdivdp 35, 1, 0
+; PC64LE9-NEXT:    lxvx 1, 0, 3
+; PC64LE9-NEXT:    xvdivdp 34, 1, 0
 ; PC64LE9-NEXT:    blr
 entry:
   %div = call <4 x double> @llvm.experimental.constrained.fdiv.v4f64(
            <4 x double> <double 1.000000e+00, double 2.000000e+00,
                          double 3.000000e+00, double 4.000000e+00>,
            <4 x double> <double 1.000000e+01, double 1.000000e+01,
                          double 1.000000e+01, double 1.000000e+01>,
            metadata !"round.dynamic",

 ; PC64LE-NEXT:    addi 4, 4, .LCPI14_1@toc@l
 ; PC64LE-NEXT:    lxvd2x 0, 0, 3
 ; PC64LE-NEXT:    addi 3, 5, .LCPI14_2@toc@l
 ; PC64LE-NEXT:    lxvd2x 1, 0, 4
 ; PC64LE-NEXT:    lxvd2x 2, 0, 3
 ; PC64LE-NEXT:    xxswapd 0, 0
 ; PC64LE-NEXT:    xxswapd 1, 1
 ; PC64LE-NEXT:    xxswapd 2, 2
-; PC64LE-NEXT:    xvmuldp 34, 1, 0
-; PC64LE-NEXT:    xvmuldp 35, 1, 2
+; PC64LE-NEXT:    xvmuldp 35, 1, 0
+; PC64LE-NEXT:    xvmuldp 34, 1, 2
 ; PC64LE-NEXT:    blr
 ;
 ; PC64LE9-LABEL: constrained_vector_fmul_v4f64:
 ; PC64LE9:       # %bb.0: # %entry
 ; PC64LE9-NEXT:    addis 3, 2, .LCPI14_0@toc@ha
 ; PC64LE9-NEXT:    addi 3, 3, .LCPI14_0@toc@l
 ; PC64LE9-NEXT:    lxvx 0, 0, 3
 ; PC64LE9-NEXT:    addis 3, 2, .LCPI14_1@toc@ha
 ; PC64LE9-NEXT:    addi 3, 3, .LCPI14_1@toc@l
 ; PC64LE9-NEXT:    lxvx 1, 0, 3
 ; PC64LE9-NEXT:    addis 3, 2, .LCPI14_2@toc@ha
 ; PC64LE9-NEXT:    addi 3, 3, .LCPI14_2@toc@l
-; PC64LE9-NEXT:    xvmuldp 34, 1, 0
-; PC64LE9-NEXT:    lxvx 0, 0, 3
 ; PC64LE9-NEXT:    xvmuldp 35, 1, 0
+; PC64LE9-NEXT:    lxvx 0, 0, 3
+; PC64LE9-NEXT:    xvmuldp 34, 1, 0
 ; PC64LE9-NEXT:    blr
 entry:
   %mul = call <4 x double> @llvm.experimental.constrained.fmul.v4f64(
            <4 x double> <double 0x7FEFFFFFFFFFFFFF, double 0x7FEFFFFFFFFFFFFF,
                          double 0x7FEFFFFFFFFFFFFF, double 0x7FEFFFFFFFFFFFFF>,
            <4 x double> <double 2.000000e+00, double 3.000000e+00,
                          double 4.000000e+00, double 5.000000e+00>,
            metadata !"round.dynamic",

 ; PC64LE-NEXT:    addi 4, 4, .LCPI19_1@toc@l
 ; PC64LE-NEXT:    lxvd2x 0, 0, 3
 ; PC64LE-NEXT:    addi 3, 5, .LCPI19_2@toc@l
 ; PC64LE-NEXT:    lxvd2x 1, 0, 4
 ; PC64LE-NEXT:    lxvd2x 2, 0, 3
 ; PC64LE-NEXT:    xxswapd 0, 0
 ; PC64LE-NEXT:    xxswapd 1, 1
 ; PC64LE-NEXT:    xxswapd 2, 2
-; PC64LE-NEXT:    xvadddp 34, 1, 0
-; PC64LE-NEXT:    xvadddp 35, 1, 2
+; PC64LE-NEXT:    xvadddp 35, 1, 0
+; PC64LE-NEXT:    xvadddp 34, 1, 2
 ; PC64LE-NEXT:    blr
 ;
 ; PC64LE9-LABEL: constrained_vector_fadd_v4f64:
 ; PC64LE9:       # %bb.0: # %entry
 ; PC64LE9-NEXT:    addis 3, 2, .LCPI19_0@toc@ha
 ; PC64LE9-NEXT:    addi 3, 3, .LCPI19_0@toc@l
 ; PC64LE9-NEXT:    lxvx 0, 0, 3
 ; PC64LE9-NEXT:    addis 3, 2, .LCPI19_1@toc@ha
 ; PC64LE9-NEXT:    addi 3, 3, .LCPI19_1@toc@l
 ; PC64LE9-NEXT:    lxvx 1, 0, 3
 ; PC64LE9-NEXT:    addis 3, 2, .LCPI19_2@toc@ha
 ; PC64LE9-NEXT:    addi 3, 3, .LCPI19_2@toc@l
-; PC64LE9-NEXT:    xvadddp 34, 1, 0
-; PC64LE9-NEXT:    lxvx 0, 0, 3
 ; PC64LE9-NEXT:    xvadddp 35, 1, 0
+; PC64LE9-NEXT:    lxvx 0, 0, 3
+; PC64LE9-NEXT:    xvadddp 34, 1, 0
 ; PC64LE9-NEXT:    blr
 entry:
   %add = call <4 x double> @llvm.experimental.constrained.fadd.v4f64(
            <4 x double> <double 0x7FEFFFFFFFFFFFFF, double 0x7FEFFFFFFFFFFFFF,
                          double 0x7FEFFFFFFFFFFFFF, double 0x7FEFFFFFFFFFFFFF>,
            <4 x double> <double 1.000000e+00, double 1.000000e-01,
                          double 2.000000e+00, double 2.000000e-01>,
            metadata !"round.dynamic",

 ; PC64LE-NEXT:    addi 4, 4, .LCPI24_1@toc@l
 ; PC64LE-NEXT:    lxvd2x 0, 0, 3
 ; PC64LE-NEXT:    addi 3, 5, .LCPI24_2@toc@l
 ; PC64LE-NEXT:    lxvd2x 1, 0, 4
 ; PC64LE-NEXT:    lxvd2x 2, 0, 3
 ; PC64LE-NEXT:    xxswapd 0, 0
 ; PC64LE-NEXT:    xxswapd 1, 1
 ; PC64LE-NEXT:    xxswapd 2, 2
-; PC64LE-NEXT:    xvsubdp 34, 1, 0
-; PC64LE-NEXT:    xvsubdp 35, 1, 2
+; PC64LE-NEXT:    xvsubdp 35, 1, 0
+; PC64LE-NEXT:    xvsubdp 34, 1, 2
 ; PC64LE-NEXT:    blr
 ;
 ; PC64LE9-LABEL: constrained_vector_fsub_v4f64:
 ; PC64LE9:       # %bb.0: # %entry
 ; PC64LE9-NEXT:    addis 3, 2, .LCPI24_0@toc@ha
 ; PC64LE9-NEXT:    addi 3, 3, .LCPI24_0@toc@l
 ; PC64LE9-NEXT:    lxvx 0, 0, 3
 ; PC64LE9-NEXT:    addis 3, 2, .LCPI24_1@toc@ha
 ; PC64LE9-NEXT:    addi 3, 3, .LCPI24_1@toc@l
 ; PC64LE9-NEXT:    lxvx 1, 0, 3
 ; PC64LE9-NEXT:    addis 3, 2, .LCPI24_2@toc@ha
 ; PC64LE9-NEXT:    addi 3, 3, .LCPI24_2@toc@l
-; PC64LE9-NEXT:    xvsubdp 34, 1, 0
-; PC64LE9-NEXT:    lxvx 0, 0, 3
 ; PC64LE9-NEXT:    xvsubdp 35, 1, 0
+; PC64LE9-NEXT:    lxvx 0, 0, 3
+; PC64LE9-NEXT:    xvsubdp 34, 1, 0
 ; PC64LE9-NEXT:    blr
 entry:
   %sub = call <4 x double> @llvm.experimental.constrained.fsub.v4f64(
            <4 x double> <double 0xFFEFFFFFFFFFFFFF, double 0xFFEFFFFFFFFFFFFF,
                          double 0xFFEFFFFFFFFFFFFF, double 0xFFEFFFFFFFFFFFFF>,
            <4 x double> <double 1.000000e+00, double 1.000000e-01,
                          double 2.000000e+00, double 2.000000e-01>,
            metadata !"round.dynamic",