[X86] Remove X86 specific scalar FMA intrinsics and upgrade to tart independent FMA and extractelement/insertelement.

llvm-svn: 336315

4 files changed, 33 insertions, 58 deletions

diff --git a/llvm/lib/IR/AutoUpgrade.cpp b/llvm/lib/IR/AutoUpgrade.cpp
index 2f111e5a587..4ef514d7b73 100644
--- a/llvm/lib/IR/AutoUpgrade.cpp
+++ b/llvm/lib/IR/AutoUpgrade.cpp
@@ -74,7 +74,7 @@ static bool ShouldUpgradeX86Intrinsic(Function *F, StringRef Name) {
   if (Name=="ssse3.pabs.b.128" || // Added in 6.0
       Name=="ssse3.pabs.w.128" || // Added in 6.0
       Name=="ssse3.pabs.d.128" || // Added in 6.0
-      Name.startswith("fma.vfmadd.p") || // Added in 7.0
+      Name.startswith("fma.vfmadd.") || // Added in 7.0
       Name.startswith("fma.vfmsub.") || // Added in 7.0
       Name.startswith("fma.vfmsubadd.") || // Added in 7.0
       Name.startswith("fma.vfnmadd.") || // Added in 7.0
@@ -2746,74 +2746,55 @@ void llvm::UpgradeIntrinsicCall(CallInst *CI, Function *NewFn) {
         Rep = EmitX86Select(Builder, CI->getArgOperand(3), Rep,
                             CI->getArgOperand(2));
       }
-    } else if (IsX86 && (Name.startswith("fma.vfmadd.p") ||
-                         Name.startswith("fma.vfmsub.p") ||
-                         Name.startswith("fma.vfnmadd.p") ||
-                         Name.startswith("fma.vfnmsub.p"))) {
+    } else if (IsX86 && (Name.startswith("fma.vfmadd.") ||
+                         Name.startswith("fma.vfmsub.") ||
+                         Name.startswith("fma.vfnmadd.") ||
+                         Name.startswith("fma.vfnmsub."))) {
       bool NegMul = Name[6] == 'n';
       bool NegAcc = NegMul ? Name[8] == 's' : Name[7] == 's';
+      bool IsScalar = NegMul ? Name[12] == 's' : Name[11] == 's';
 
       Value *Ops[] = { CI->getArgOperand(0), CI->getArgOperand(1),
                        CI->getArgOperand(2) };
 
-      if (NegMul)
+      if (IsScalar) {
+        Ops[0] = Builder.CreateExtractElement(Ops[0], (uint64_t)0);
+        Ops[1] = Builder.CreateExtractElement(Ops[1], (uint64_t)0);
+        Ops[2] = Builder.CreateExtractElement(Ops[2], (uint64_t)0);
+      }
+
+      if (NegMul && !IsScalar)
         Ops[0] = Builder.CreateFNeg(Ops[0]);
+      if (NegMul && IsScalar)
+        Ops[1] = Builder.CreateFNeg(Ops[1]);
       if (NegAcc)
         Ops[2] = Builder.CreateFNeg(Ops[2]);
 
       Rep = Builder.CreateCall(Intrinsic::getDeclaration(CI->getModule(),
                                                          Intrinsic::fma,
-                                                         CI->getType()), Ops);
-    } else if (IsX86 && (Name.startswith("fma.vfmsub.s") ||
-                         Name.startswith("fma.vfmsubadd.p"))) {
-      // Handle FMSUB and FSUBADD.
+                                                         Ops[0]->getType()),
+                               Ops);
+
+      if (IsScalar)
+        Rep = Builder.CreateInsertElement(CI->getArgOperand(0), Rep,
+                                          (uint64_t)0);
+    } else if (IsX86 && Name.startswith("fma.vfmsubadd.p")) {
+      // Handle FSUBADD.
       unsigned VecWidth = CI->getType()->getPrimitiveSizeInBits();
       unsigned EltWidth = CI->getType()->getScalarSizeInBits();
       Intrinsic::ID IID;
-      if (Name[10] == '.' && Name[11] == 's') {
-        // Scalar FMSUB
-        if (EltWidth == 32)
-          IID = Intrinsic::x86_fma_vfmadd_ss;
-        else if (EltWidth == 64)
-          IID = Intrinsic::x86_fma_vfmadd_sd;
-        else
-          llvm_unreachable("Unexpected intrinsic");
-      } else {
-        // FMSUBADD
-        if (VecWidth == 128 && EltWidth == 32)
-          IID = Intrinsic::x86_fma_vfmaddsub_ps;
-        else if (VecWidth == 128 && EltWidth == 64)
-          IID = Intrinsic::x86_fma_vfmaddsub_pd;
-        else if (VecWidth == 256 && EltWidth == 32)
-          IID = Intrinsic::x86_fma_vfmaddsub_ps_256;
-        else if (VecWidth == 256 && EltWidth == 64)
-          IID = Intrinsic::x86_fma_vfmaddsub_pd_256;
-        else
-          llvm_unreachable("Unexpected intrinsic");
-      }
-      Value *Arg2 = Builder.CreateFNeg(CI->getArgOperand(2));
-      Value *Ops[] = { CI->getArgOperand(0), CI->getArgOperand(1), Arg2 };
-      Rep = Builder.CreateCall(Intrinsic::getDeclaration(CI->getModule(), IID),
-                                                         Ops);
-    } else if (IsX86 && (Name.startswith("fma.vfnmadd.s") ||
-                         Name.startswith("fma.vfnmsub.s"))) {
-      Value *Arg0 = CI->getArgOperand(0);
-      Value *Arg1 = CI->getArgOperand(1);
-      Value *Arg2 = CI->getArgOperand(2);
-      unsigned EltWidth = CI->getType()->getScalarSizeInBits();
-      // Scalar FNMADD/FNMSUB
-      Arg1 = Builder.CreateFNeg(Arg1); // Arg0 is passthru so invert Arg1.
-      Intrinsic::ID IID;
-      if (EltWidth == 32)
-        IID = Intrinsic::x86_fma_vfmadd_ss;
-      else if (EltWidth == 64)
-        IID = Intrinsic::x86_fma_vfmadd_sd;
+      if (VecWidth == 128 && EltWidth == 32)
+        IID = Intrinsic::x86_fma_vfmaddsub_ps;
+      else if (VecWidth == 128 && EltWidth == 64)
+        IID = Intrinsic::x86_fma_vfmaddsub_pd;
+      else if (VecWidth == 256 && EltWidth == 32)
+        IID = Intrinsic::x86_fma_vfmaddsub_ps_256;
+      else if (VecWidth == 256 && EltWidth == 64)
+        IID = Intrinsic::x86_fma_vfmaddsub_pd_256;
       else
         llvm_unreachable("Unexpected intrinsic");
-      // Invert for FNMSUB.
-      if (Name[8] == 's')
-        Arg2 = Builder.CreateFNeg(Arg2);
-      Value *Ops[] = { Arg0, Arg1, Arg2 };
+      Value *Arg2 = Builder.CreateFNeg(CI->getArgOperand(2));
+      Value *Ops[] = { CI->getArgOperand(0), CI->getArgOperand(1), Arg2 };
       Rep = Builder.CreateCall(Intrinsic::getDeclaration(CI->getModule(), IID),
                                                          Ops);
     } else if (IsX86 && (Name.startswith("avx512.mask.pternlog.") ||
diff --git a/llvm/lib/Target/X86/X86IntrinsicsInfo.h b/llvm/lib/Target/X86/X86IntrinsicsInfo.h
index ed0c69cc16d..af26fe3a6c8 100644
--- a/llvm/lib/Target/X86/X86IntrinsicsInfo.h
+++ b/llvm/lib/Target/X86/X86IntrinsicsInfo.h
@@ -1249,8 +1249,6 @@ static const IntrinsicData  IntrinsicsWithoutChain[] = {
   X86_INTRINSIC_DATA(avx512_vpshrd_w_512, INTR_TYPE_3OP_IMM8, X86ISD::VSHRD, 0),
   X86_INTRINSIC_DATA(bmi_bextr_32,         INTR_TYPE_2OP, X86ISD::BEXTR, 0),
   X86_INTRINSIC_DATA(bmi_bextr_64,         INTR_TYPE_2OP, X86ISD::BEXTR, 0),
-  X86_INTRINSIC_DATA(fma_vfmadd_sd,        INTR_TYPE_3OP, X86ISD::FMADDS1, 0),
-  X86_INTRINSIC_DATA(fma_vfmadd_ss,        INTR_TYPE_3OP, X86ISD::FMADDS1, 0),
   X86_INTRINSIC_DATA(fma_vfmaddsub_pd,     INTR_TYPE_3OP, X86ISD::FMADDSUB, 0),
   X86_INTRINSIC_DATA(fma_vfmaddsub_pd_256, INTR_TYPE_3OP, X86ISD::FMADDSUB, 0),
   X86_INTRINSIC_DATA(fma_vfmaddsub_ps,     INTR_TYPE_3OP, X86ISD::FMADDSUB, 0),
diff --git a/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp b/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp
index 3f9856d240b..92533ffdb4b 100644
--- a/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp
+++ b/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp
@@ -2545,8 +2545,6 @@ Instruction *InstCombiner::visitCallInst(CallInst &CI) {
   case Intrinsic::x86_avx512_mask3_vfmsub_sd:
   case Intrinsic::x86_avx512_mask3_vfnmsub_ss:
   case Intrinsic::x86_avx512_mask3_vfnmsub_sd:
-  case Intrinsic::x86_fma_vfmadd_ss:
-  case Intrinsic::x86_fma_vfmadd_sd:
   case Intrinsic::x86_sse_cmp_ss:
   case Intrinsic::x86_sse_min_ss:
   case Intrinsic::x86_sse_max_ss:
diff --git a/llvm/lib/Transforms/InstCombine/InstCombineSimplifyDemanded.cpp b/llvm/lib/Transforms/InstCombine/InstCombineSimplifyDemanded.cpp
index 2a31ac9540b..97d24019eb6 100644
--- a/llvm/lib/Transforms/InstCombine/InstCombineSimplifyDemanded.cpp
+++ b/llvm/lib/Transforms/InstCombine/InstCombineSimplifyDemanded.cpp
@@ -1497,8 +1497,6 @@ Value *InstCombiner::SimplifyDemandedVectorElts(Value *V, APInt DemandedElts,
     case Intrinsic::x86_avx512_mask_sub_sd_round:
     case Intrinsic::x86_avx512_mask_max_sd_round:
     case Intrinsic::x86_avx512_mask_min_sd_round:
-    case Intrinsic::x86_fma_vfmadd_ss:
-    case Intrinsic::x86_fma_vfmadd_sd:
     case Intrinsic::x86_avx512_mask_vfmadd_ss:
     case Intrinsic::x86_avx512_mask_vfmadd_sd:
     case Intrinsic::x86_avx512_maskz_vfmadd_ss: