[X86][SSE] (Reapplied) Replace (V)PMOVSX and (V)PMOVZX integer extension intrinsics with generic IR (llvm)

This patch removes the llvm intrinsics VPMOVSX and (V)PMOVZX sign/zero extension intrinsics and auto-upgrades to SEXT/ZEXT calls instead. We already did this for SSE41 PMOVSX sometime ago so much of that implementation can be reused.

Reapplied now that the the companion patch (D20684) removes/auto-upgrade the clang intrinsics has been committed.

Differential Revision: http://reviews.llvm.org/D20686

llvm-svn: 271131

3 files changed, 12 insertions, 65 deletions

diff --git a/llvm/lib/IR/AutoUpgrade.cpp b/llvm/lib/IR/AutoUpgrade.cpp
index d05bb47a68c..fd2873ed01f 100644
--- a/llvm/lib/IR/AutoUpgrade.cpp
+++ b/llvm/lib/IR/AutoUpgrade.cpp
@@ -178,6 +178,9 @@ static bool UpgradeIntrinsicFunction1(Function *F, Function *&NewFn) {
         Name.startswith("x86.avx2.pbroadcast") ||
         Name.startswith("x86.avx.vpermil.") ||
         Name.startswith("x86.sse41.pmovsx") ||
+        Name.startswith("x86.sse41.pmovzx") ||
+        Name.startswith("x86.avx2.pmovsx") ||
+        Name.startswith("x86.avx2.pmovzx") ||
         Name == "x86.sse2.cvtdq2pd" ||
         Name == "x86.sse2.cvtps2pd" ||
         Name == "x86.avx.cvtdq2.pd.256" ||
@@ -544,19 +547,25 @@ void llvm::UpgradeIntrinsicCall(CallInst *CI, Function *NewFn) {
       for (unsigned I = 0; I < EltNum; ++I)
         Rep = Builder.CreateInsertElement(Rep, Load,
                                           ConstantInt::get(I32Ty, I));
-    } else if (Name.startswith("llvm.x86.sse41.pmovsx")) {
+    } else if (Name.startswith("llvm.x86.sse41.pmovsx") ||
+               Name.startswith("llvm.x86.sse41.pmovzx") ||
+               Name.startswith("llvm.x86.avx2.pmovsx") ||
+               Name.startswith("llvm.x86.avx2.pmovzx")) {
       VectorType *SrcTy = cast<VectorType>(CI->getArgOperand(0)->getType());
       VectorType *DstTy = cast<VectorType>(CI->getType());
       unsigned NumDstElts = DstTy->getNumElements();
 
-      // Extract a subvector of the first NumDstElts lanes and sign extend.
+      // Extract a subvector of the first NumDstElts lanes and sign/zero extend.
       SmallVector<int, 8> ShuffleMask;
       for (int i = 0; i != (int)NumDstElts; ++i)
         ShuffleMask.push_back(i);
 
       Value *SV = Builder.CreateShuffleVector(
           CI->getArgOperand(0), UndefValue::get(SrcTy), ShuffleMask);
-      Rep = Builder.CreateSExt(SV, DstTy);
+
+      bool DoSext = (StringRef::npos != Name.find("pmovsx"));
+      Rep = DoSext ? Builder.CreateSExt(SV, DstTy)
+                   : Builder.CreateZExt(SV, DstTy);
     } else if (Name == "llvm.x86.avx2.vbroadcasti128") {
       // Replace vbroadcasts with a vector shuffle.
       Type *VT = VectorType::get(Type::getInt64Ty(C), 2);
diff --git a/llvm/lib/Target/X86/X86IntrinsicsInfo.h b/llvm/lib/Target/X86/X86IntrinsicsInfo.h
index 93d1e94d1ca..6d8f254b5c4 100644
--- a/llvm/lib/Target/X86/X86IntrinsicsInfo.h
+++ b/llvm/lib/Target/X86/X86IntrinsicsInfo.h
@@ -361,18 +361,6 @@ static const IntrinsicData  IntrinsicsWithoutChain[] = {
   X86_INTRINSIC_DATA(avx2_pminu_d, INTR_TYPE_2OP, ISD::UMIN, 0),
   X86_INTRINSIC_DATA(avx2_pminu_w, INTR_TYPE_2OP, ISD::UMIN, 0),
   X86_INTRINSIC_DATA(avx2_pmovmskb, INTR_TYPE_1OP, X86ISD::MOVMSK, 0),
-  X86_INTRINSIC_DATA(avx2_pmovsxbd, INTR_TYPE_1OP, X86ISD::VSEXT, 0),
-  X86_INTRINSIC_DATA(avx2_pmovsxbq, INTR_TYPE_1OP, X86ISD::VSEXT, 0),
-  X86_INTRINSIC_DATA(avx2_pmovsxbw, INTR_TYPE_1OP, X86ISD::VSEXT, 0),
-  X86_INTRINSIC_DATA(avx2_pmovsxdq, INTR_TYPE_1OP, X86ISD::VSEXT, 0),
-  X86_INTRINSIC_DATA(avx2_pmovsxwd, INTR_TYPE_1OP, X86ISD::VSEXT, 0),
-  X86_INTRINSIC_DATA(avx2_pmovsxwq, INTR_TYPE_1OP, X86ISD::VSEXT, 0),
-  X86_INTRINSIC_DATA(avx2_pmovzxbd, INTR_TYPE_1OP, X86ISD::VZEXT, 0),
-  X86_INTRINSIC_DATA(avx2_pmovzxbq, INTR_TYPE_1OP, X86ISD::VZEXT, 0),
-  X86_INTRINSIC_DATA(avx2_pmovzxbw, INTR_TYPE_1OP, X86ISD::VZEXT, 0),
-  X86_INTRINSIC_DATA(avx2_pmovzxdq, INTR_TYPE_1OP, X86ISD::VZEXT, 0),
-  X86_INTRINSIC_DATA(avx2_pmovzxwd, INTR_TYPE_1OP, X86ISD::VZEXT, 0),
-  X86_INTRINSIC_DATA(avx2_pmovzxwq, INTR_TYPE_1OP, X86ISD::VZEXT, 0),
   X86_INTRINSIC_DATA(avx2_pmul_dq, INTR_TYPE_2OP, X86ISD::PMULDQ, 0),
   X86_INTRINSIC_DATA(avx2_pmulh_w, INTR_TYPE_2OP, ISD::MULHS, 0),
   X86_INTRINSIC_DATA(avx2_pmulhu_w, INTR_TYPE_2OP, ISD::MULHU, 0),
@@ -2288,12 +2276,6 @@ static const IntrinsicData  IntrinsicsWithoutChain[] = {
   X86_INTRINSIC_DATA(sse41_pminsd,      INTR_TYPE_2OP, ISD::SMIN, 0),
   X86_INTRINSIC_DATA(sse41_pminud,      INTR_TYPE_2OP, ISD::UMIN, 0),
   X86_INTRINSIC_DATA(sse41_pminuw,      INTR_TYPE_2OP, ISD::UMIN, 0),
-  X86_INTRINSIC_DATA(sse41_pmovzxbd,    INTR_TYPE_1OP, X86ISD::VZEXT, 0),
-  X86_INTRINSIC_DATA(sse41_pmovzxbq,    INTR_TYPE_1OP, X86ISD::VZEXT, 0),
-  X86_INTRINSIC_DATA(sse41_pmovzxbw,    INTR_TYPE_1OP, X86ISD::VZEXT, 0),
-  X86_INTRINSIC_DATA(sse41_pmovzxdq,    INTR_TYPE_1OP, X86ISD::VZEXT, 0),
-  X86_INTRINSIC_DATA(sse41_pmovzxwd,    INTR_TYPE_1OP, X86ISD::VZEXT, 0),
-  X86_INTRINSIC_DATA(sse41_pmovzxwq,    INTR_TYPE_1OP, X86ISD::VZEXT, 0),
   X86_INTRINSIC_DATA(sse41_pmuldq,      INTR_TYPE_2OP, X86ISD::PMULDQ, 0),
   X86_INTRINSIC_DATA(sse4a_extrqi,      INTR_TYPE_3OP, X86ISD::EXTRQI, 0),
   X86_INTRINSIC_DATA(sse4a_insertqi,    INTR_TYPE_4OP, X86ISD::INSERTQI, 0),
diff --git a/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp b/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp
index c07ca85d3a4..c87d0466a9d 100644
--- a/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp
+++ b/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp
@@ -325,24 +325,6 @@ static Value *simplifyX86immShift(const IntrinsicInst &II,
   return Builder.CreateAShr(Vec, ShiftVec);
 }
 
-static Value *simplifyX86extend(const IntrinsicInst &II,
-                                InstCombiner::BuilderTy &Builder,
-                                bool SignExtend) {
-  VectorType *SrcTy = cast<VectorType>(II.getArgOperand(0)->getType());
-  VectorType *DstTy = cast<VectorType>(II.getType());
-  unsigned NumDstElts = DstTy->getNumElements();
-
-  // Extract a subvector of the first NumDstElts lanes and sign/zero extend.
-  SmallVector<int, 8> ShuffleMask;
-  for (int i = 0; i != (int)NumDstElts; ++i)
-    ShuffleMask.push_back(i);
-
-  Value *SV = Builder.CreateShuffleVector(II.getArgOperand(0),
-                                          UndefValue::get(SrcTy), ShuffleMask);
-  return SignExtend ? Builder.CreateSExt(SV, DstTy)
-                    : Builder.CreateZExt(SV, DstTy);
-}
-
 static Value *simplifyX86insertps(const IntrinsicInst &II,
                                   InstCombiner::BuilderTy &Builder) {
   auto *CInt = dyn_cast<ConstantInt>(II.getArgOperand(2));
@@ -1649,32 +1631,6 @@ Instruction *InstCombiner::visitCallInst(CallInst &CI) {
     break;
   }
 
-  case Intrinsic::x86_avx2_pmovsxbd:
-  case Intrinsic::x86_avx2_pmovsxbq:
-  case Intrinsic::x86_avx2_pmovsxbw:
-  case Intrinsic::x86_avx2_pmovsxdq:
-  case Intrinsic::x86_avx2_pmovsxwd:
-  case Intrinsic::x86_avx2_pmovsxwq:
-    if (Value *V = simplifyX86extend(*II, *Builder, true))
-      return replaceInstUsesWith(*II, V);
-    break;
-
-  case Intrinsic::x86_sse41_pmovzxbd:
-  case Intrinsic::x86_sse41_pmovzxbq:
-  case Intrinsic::x86_sse41_pmovzxbw:
-  case Intrinsic::x86_sse41_pmovzxdq:
-  case Intrinsic::x86_sse41_pmovzxwd:
-  case Intrinsic::x86_sse41_pmovzxwq:
-  case Intrinsic::x86_avx2_pmovzxbd:
-  case Intrinsic::x86_avx2_pmovzxbq:
-  case Intrinsic::x86_avx2_pmovzxbw:
-  case Intrinsic::x86_avx2_pmovzxdq:
-  case Intrinsic::x86_avx2_pmovzxwd:
-  case Intrinsic::x86_avx2_pmovzxwq:
-    if (Value *V = simplifyX86extend(*II, *Builder, false))
-      return replaceInstUsesWith(*II, V);
-    break;
-
   case Intrinsic::x86_sse41_insertps:
     if (Value *V = simplifyX86insertps(*II, *Builder))
       return replaceInstUsesWith(*II, V);