Index: lib/Transforms/InstCombine/InstCombineCalls.cpp
===================================================================
--- lib/Transforms/InstCombine/InstCombineCalls.cpp
+++ lib/Transforms/InstCombine/InstCombineCalls.cpp
@@ -199,34 +199,65 @@
 
 static Value *SimplifyX86immshift(const IntrinsicInst &II,
                                   InstCombiner::BuilderTy &Builder,
-                                  bool ShiftLeft) {
-  // Simplify if count is constant. To 0 if >= BitWidth,
-  // otherwise to shl/lshr.
-  auto CDV = dyn_cast<ConstantDataVector>(II.getArgOperand(1));
-  auto CInt = dyn_cast<ConstantInt>(II.getArgOperand(1));
-  if (!CDV && !CInt)
+                                  bool LogicalShift, bool ShiftLeft) {
+  // Simplify if count is constant.
+  auto Arg1 = II.getArgOperand(1);
+  auto CAZ = dyn_cast<ConstantAggregateZero>(Arg1);
+  auto CDV = dyn_cast<ConstantDataVector>(Arg1);
+  auto CInt = dyn_cast<ConstantInt>(Arg1);
+  if (!CAZ && !CDV && !CInt)
     return nullptr;
-  ConstantInt *Count;
-  if (CDV)
-    Count = cast<ConstantInt>(CDV->getElementAsConstant(0));
-  else
-    Count = CInt;
+
+  APInt Count(64, 0);
+  if (CDV) {
+    // Shift amount vectors use the entire lower 64-bits integer.
+    auto VT = cast<VectorType>(CDV->getType());
+    unsigned BitWidth = VT->getElementType()->getPrimitiveSizeInBits();
+    assert((64 % BitWidth) == 0 && "Unexpected packed shift size");
+    unsigned NumSubElts = 64 / BitWidth;
+
+    // Concatenate the sub-elements to create the 64-bit value.
+    for (unsigned i = 0; i != NumSubElts; ++i) {
+      unsigned SubEltIdx = (NumSubElts - 1) - i;
+      auto SubElt = cast<ConstantInt>(CDV->getElementAsConstant(SubEltIdx));
+      Count = Count.shl(BitWidth);
+      Count |= SubElt->getValue().zextOrTrunc(64);
+    }
+  }
+  else if (CInt)
+    Count = CInt->getValue();
 
   auto Vec = II.getArgOperand(0);
   auto VT = cast<VectorType>(Vec->getType());
   auto SVT = VT->getElementType();
-  if (Count->getZExtValue() > (SVT->getPrimitiveSizeInBits() - 1))
-    return ConstantAggregateZero::get(VT);
-
   unsigned VWidth = VT->getNumElements();
+  unsigned BitWidth = SVT->getPrimitiveSizeInBits();
+
+  // If shift-by-zero then just return the original value.
+  if (Count == 0)
+    return Vec;
+
+  // Handle cases when Shift >= BitWidth.
+  if (Count.uge(BitWidth)) {
+    // If LogicalShift - just return zero.
+    if (LogicalShift)
+      return ConstantAggregateZero::get(VT);
+
+    // If ArithmeticShift - clamp Shift to (BitWidth - 1).
+    Count = APInt(64, BitWidth - 1);
+  }
 
   // Get a constant vector of the same type as the first operand.
-  auto VTCI = ConstantInt::get(VT->getElementType(), Count->getZExtValue());
+  auto ShiftAmt = ConstantInt::get(SVT, Count.zextOrTrunc(BitWidth));
+  auto ShiftVec = Builder.CreateVectorSplat(VWidth, ShiftAmt);
 
   if (ShiftLeft)
-    return Builder.CreateShl(Vec, Builder.CreateVectorSplat(VWidth, VTCI));
+    return Builder.CreateShl(Vec, ShiftVec);
+
+  if (LogicalShift)
+    return Builder.CreateLShr(Vec, ShiftVec);
 
-  return Builder.CreateLShr(Vec, Builder.CreateVectorSplat(VWidth, VTCI));
+  return Builder.CreateAShr(Vec, ShiftVec);
 }
 
 static Value *SimplifyX86extend(const IntrinsicInst &II,
@@ -753,6 +784,19 @@
     break;
   }
 
+  // Constant fold ashr( <A x Bi>, Ci ).
+  case Intrinsic::x86_sse2_psra_d:
+  case Intrinsic::x86_sse2_psra_w:
+  case Intrinsic::x86_sse2_psrai_d:
+  case Intrinsic::x86_sse2_psrai_w:
+  case Intrinsic::x86_avx2_psra_d:
+  case Intrinsic::x86_avx2_psra_w:
+  case Intrinsic::x86_avx2_psrai_d:
+  case Intrinsic::x86_avx2_psrai_w:
+    if (Value *V = SimplifyX86immshift(*II, *Builder, false, false))
+      return ReplaceInstUsesWith(*II, V);
+    break;
+
   // Constant fold lshr( <A x Bi>, Ci ).
   case Intrinsic::x86_sse2_psrl_d:
   case Intrinsic::x86_sse2_psrl_q:
@@ -766,7 +810,7 @@
   case Intrinsic::x86_avx2_psrli_d:
   case Intrinsic::x86_avx2_psrli_q:
   case Intrinsic::x86_avx2_psrli_w:
-    if (Value *V = SimplifyX86immshift(*II, *Builder, false))
+    if (Value *V = SimplifyX86immshift(*II, *Builder, true, false))
       return ReplaceInstUsesWith(*II, V);
     break;
 
@@ -783,7 +827,7 @@
   case Intrinsic::x86_avx2_pslli_d:
   case Intrinsic::x86_avx2_pslli_q:
   case Intrinsic::x86_avx2_pslli_w:
-    if (Value *V = SimplifyX86immshift(*II, *Builder, true))
+    if (Value *V = SimplifyX86immshift(*II, *Builder, true, true))
       return ReplaceInstUsesWith(*II, V);
     break;
 
Index: test/Transforms/InstCombine/x86-vector-shifts.ll
===================================================================
--- test/Transforms/InstCombine/x86-vector-shifts.ll
+++ test/Transforms/InstCombine/x86-vector-shifts.ll
@@ -2,6 +2,102 @@
 target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
 
 ;
+; ASHR - Immediate
+;
+
+define <8 x i16> @sse2_psrai_w_0(<8 x i16> %v) nounwind readnone uwtable {
+; CHECK-LABEL: @sse2_psrai_w_0
+; CHECK: ret <8 x i16> %v
+  %1 = tail call <8 x i16> @llvm.x86.sse2.psrai.w(<8 x i16> %v, i32 0)
+  ret <8 x i16> %1
+}
+
+define <8 x i16> @sse2_psrai_w_15(<8 x i16> %v) nounwind readnone uwtable {
+; CHECK-LABEL: @sse2_psrai_w_15
+; CHECK: %1 = ashr <8 x i16> %v, <i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15>
+; CHECK: ret <8 x i16> %1
+  %1 = tail call <8 x i16> @llvm.x86.sse2.psrai.w(<8 x i16> %v, i32 15)
+  ret <8 x i16> %1
+}
+
+define <8 x i16> @sse2_psrai_w_64(<8 x i16> %v) nounwind readnone uwtable {
+; CHECK-LABEL: @sse2_psrai_w_64
+; CHECK: %1 = ashr <8 x i16> %v, <i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15>
+; CHECK: ret <8 x i16> %1
+  %1 = tail call <8 x i16> @llvm.x86.sse2.psrai.w(<8 x i16> %v, i32 64)
+  ret <8 x i16> %1
+}
+
+define <4 x i32> @sse2_psrai_d_0(<4 x i32> %v) nounwind readnone uwtable {
+; CHECK-LABEL: @sse2_psrai_d_0
+; CHECK: ret <4 x i32> %v
+  %1 = tail call <4 x i32> @llvm.x86.sse2.psrai.d(<4 x i32> %v, i32 0)
+  ret <4 x i32> %1
+}
+
+define <4 x i32> @sse2_psrai_d_15(<4 x i32> %v) nounwind readnone uwtable {
+; CHECK-LABEL: @sse2_psrai_d_15
+; CHECK: %1 = ashr <4 x i32> %v, <i32 15, i32 15, i32 15, i32 15>
+; CHECK: ret <4 x i32> %1
+  %1 = tail call <4 x i32> @llvm.x86.sse2.psrai.d(<4 x i32> %v, i32 15)
+  ret <4 x i32> %1
+}
+
+define <4 x i32> @sse2_psrai_d_64(<4 x i32> %v) nounwind readnone uwtable {
+; CHECK-LABEL: @sse2_psrai_d_64
+; CHECK: %1 = ashr <4 x i32> %v, <i32 31, i32 31, i32 31, i32 31>
+; CHECK: ret <4 x i32> %1
+  %1 = tail call <4 x i32> @llvm.x86.sse2.psrai.d(<4 x i32> %v, i32 64)
+  ret <4 x i32> %1
+}
+
+define <16 x i16> @avx2_psrai_w_0(<16 x i16> %v) nounwind readnone uwtable {
+; CHECK-LABEL: @avx2_psrai_w_0
+; CHECK: ret <16 x i16> %v
+  %1 = tail call <16 x i16> @llvm.x86.avx2.psrai.w(<16 x i16> %v, i32 0)
+  ret <16 x i16> %1
+}
+
+define <16 x i16> @avx2_psrai_w_15(<16 x i16> %v) nounwind readnone uwtable {
+; CHECK-LABEL: @avx2_psrai_w_15
+; CHECK: %1 = ashr <16 x i16> %v, <i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15>
+; CHECK: ret <16 x i16> %1
+  %1 = tail call <16 x i16> @llvm.x86.avx2.psrai.w(<16 x i16> %v, i32 15)
+  ret <16 x i16> %1
+}
+
+define <16 x i16> @avx2_psrai_w_64(<16 x i16> %v) nounwind readnone uwtable {
+; CHECK-LABEL: @avx2_psrai_w_64
+; CHECK: %1 = ashr <16 x i16> %v, <i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15>
+; CHECK: ret <16 x i16> %1
+  %1 = tail call <16 x i16> @llvm.x86.avx2.psrai.w(<16 x i16> %v, i32 64)
+  ret <16 x i16> %1
+}
+
+define <8 x i32> @avx2_psrai_d_0(<8 x i32> %v) nounwind readnone uwtable {
+; CHECK-LABEL: @avx2_psrai_d_0
+; CHECK: ret <8 x i32> %v
+  %1 = tail call <8 x i32> @llvm.x86.avx2.psrai.d(<8 x i32> %v, i32 0)
+  ret <8 x i32> %1
+}
+
+define <8 x i32> @avx2_psrai_d_15(<8 x i32> %v) nounwind readnone uwtable {
+; CHECK-LABEL: @avx2_psrai_d_15
+; CHECK: %1 = ashr <8 x i32> %v, <i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15>
+; CHECK: ret <8 x i32> %1
+  %1 = tail call <8 x i32> @llvm.x86.avx2.psrai.d(<8 x i32> %v, i32 15)
+  ret <8 x i32> %1
+}
+
+define <8 x i32> @avx2_psrai_d_64(<8 x i32> %v) nounwind readnone uwtable {
+; CHECK-LABEL: @avx2_psrai_d_64
+; CHECK: %1 = ashr <8 x i32> %v, <i32 31, i32 31, i32 31, i32 31, i32 31, i32 31, i32 31, i32 31>
+; CHECK: ret <8 x i32> %1
+  %1 = tail call <8 x i32> @llvm.x86.avx2.psrai.d(<8 x i32> %v, i32 64)
+  ret <8 x i32> %1
+}
+
+;
 ; LSHR - Immediate
 ;
 
@@ -274,9 +370,144 @@
 }
 
 ;
+; ASHR - Constant Vector
+;
+
+define <8 x i16> @sse2_psra_w_0(<8 x i16> %v) nounwind readnone uwtable {
+; CHECK-LABEL: @sse2_psra_w_0
+; CHECK: ret <8 x i16> %v
+  %1 = tail call <8 x i16> @llvm.x86.sse2.psra.w(<8 x i16> %v, <8 x i16> zeroinitializer)
+  ret <8 x i16> %1
+}
+
+define <8 x i16> @sse2_psra_w_15(<8 x i16> %v) nounwind readnone uwtable {
+; CHECK-LABEL: @sse2_psra_w_15
+; CHECK: %1 = ashr <8 x i16> %v, <i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15>
+; CHECK: ret <8 x i16> %1
+  %1 = tail call <8 x i16> @llvm.x86.sse2.psra.w(<8 x i16> %v, <8 x i16> <i16 15, i16 0, i16 0, i16 0, i16 9999, i16 9999, i16 9999, i16 9999>)
+  ret <8 x i16> %1
+}
+
+define <8 x i16> @sse2_psra_w_15_splat(<8 x i16> %v) nounwind readnone uwtable {
+; CHECK-LABEL: @sse2_psra_w_15_splat
+; CHECK: %1 = ashr <8 x i16> %v, <i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15>
+; CHECK: ret <8 x i16> %1
+  %1 = tail call <8 x i16> @llvm.x86.sse2.psra.w(<8 x i16> %v, <8 x i16> <i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15>)
+  ret <8 x i16> %1
+}
+
+define <8 x i16> @sse2_psra_w_64(<8 x i16> %v) nounwind readnone uwtable {
+; CHECK-LABEL: @sse2_psra_w_64
+; CHECK: %1 = ashr <8 x i16> %v, <i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15>
+; CHECK: ret <8 x i16> %1
+  %1 = tail call <8 x i16> @llvm.x86.sse2.psra.w(<8 x i16> %v, <8 x i16> <i16 64, i16 0, i16 0, i16 0, i16 9999, i16 9999, i16 9999, i16 9999>)
+  ret <8 x i16> %1
+}
+
+define <4 x i32> @sse2_psra_d_0(<4 x i32> %v) nounwind readnone uwtable {
+; CHECK-LABEL: @sse2_psra_d_0
+; CHECK: ret <4 x i32> %v
+  %1 = tail call <4 x i32> @llvm.x86.sse2.psra.d(<4 x i32> %v, <4 x i32> zeroinitializer)
+  ret <4 x i32> %1
+}
+
+define <4 x i32> @sse2_psra_d_15(<4 x i32> %v) nounwind readnone uwtable {
+; CHECK-LABEL: @sse2_psra_d_15
+; CHECK: %1 = ashr <4 x i32> %v, <i32 15, i32 15, i32 15, i32 15>
+; CHECK: ret <4 x i32> %1
+  %1 = tail call <4 x i32> @llvm.x86.sse2.psra.d(<4 x i32> %v, <4 x i32> <i32 15, i32 0, i32 9999, i32 9999>)
+  ret <4 x i32> %1
+}
+
+define <4 x i32> @sse2_psra_d_15_splat(<4 x i32> %v) nounwind readnone uwtable {
+; CHECK-LABEL: @sse2_psra_d_15_splat
+; CHECK: %1 = ashr <4 x i32> %v, <i32 31, i32 31, i32 31, i32 31>
+; CHECK: ret <4 x i32> %1
+  %1 = tail call <4 x i32> @llvm.x86.sse2.psra.d(<4 x i32> %v, <4 x i32> <i32 15, i32 15, i32 15, i32 15>)
+  ret <4 x i32> %1
+}
+
+define <4 x i32> @sse2_psra_d_64(<4 x i32> %v) nounwind readnone uwtable {
+; CHECK-LABEL: @sse2_psra_d_64
+; CHECK: %1 = ashr <4 x i32> %v, <i32 31, i32 31, i32 31, i32 31>
+; CHECK: ret <4 x i32> %1
+  %1 = tail call <4 x i32> @llvm.x86.sse2.psra.d(<4 x i32> %v, <4 x i32> <i32 64, i32 0, i32 9999, i32 9999>)
+  ret <4 x i32> %1
+}
+
+define <16 x i16> @avx2_psra_w_0(<16 x i16> %v) nounwind readnone uwtable {
+; CHECK-LABEL: @avx2_psra_w_0
+; CHECK: ret <16 x i16> %v
+  %1 = tail call <16 x i16> @llvm.x86.avx2.psra.w(<16 x i16> %v, <8 x i16> zeroinitializer)
+  ret <16 x i16> %1
+}
+
+define <16 x i16> @avx2_psra_w_15(<16 x i16> %v) nounwind readnone uwtable {
+; CHECK-LABEL: @avx2_psra_w_15
+; CHECK: %1 = ashr <16 x i16> %v, <i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15>
+; CHECK: ret <16 x i16> %1
+  %1 = tail call <16 x i16> @llvm.x86.avx2.psra.w(<16 x i16> %v, <8 x i16> <i16 15, i16 0, i16 0, i16 0, i16 9999, i16 9999, i16 9999, i16 9999>)
+  ret <16 x i16> %1
+}
+
+define <16 x i16> @avx2_psra_w_15_splat(<16 x i16> %v) nounwind readnone uwtable {
+; CHECK-LABEL: @avx2_psra_w_15_splat
+; CHECK: %1 = ashr <16 x i16> %v, <i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15>
+; CHECK: ret <16 x i16> %1
+  %1 = tail call <16 x i16> @llvm.x86.avx2.psra.w(<16 x i16> %v, <8 x i16> <i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15>)
+  ret <16 x i16> %1
+}
+
+define <16 x i16> @avx2_psra_w_64(<16 x i16> %v) nounwind readnone uwtable {
+; CHECK-LABEL: @avx2_psra_w_64
+; CHECK: %1 = ashr <16 x i16> %v, <i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15>
+; CHECK: ret <16 x i16> %1
+  %1 = tail call <16 x i16> @llvm.x86.avx2.psra.w(<16 x i16> %v, <8 x i16> <i16 64, i16 0, i16 0, i16 0, i16 9999, i16 9999, i16 9999, i16 9999>)
+  ret <16 x i16> %1
+}
+
+define <8 x i32> @avx2_psra_d_0(<8 x i32> %v) nounwind readnone uwtable {
+; CHECK-LABEL: @avx2_psra_d_0
+; CHECK: ret <8 x i32> %v
+  %1 = tail call <8 x i32> @llvm.x86.avx2.psra.d(<8 x i32> %v, <4 x i32> zeroinitializer)
+  ret <8 x i32> %1
+}
+
+define <8 x i32> @avx2_psra_d_15(<8 x i32> %v) nounwind readnone uwtable {
+; CHECK-LABEL: @avx2_psra_d_15
+; CHECK: %1 = ashr <8 x i32> %v, <i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15>
+; CHECK: ret <8 x i32> %1
+  %1 = tail call <8 x i32> @llvm.x86.avx2.psra.d(<8 x i32> %v, <4 x i32> <i32 15, i32 0, i32 9999, i32 9999>)
+  ret <8 x i32> %1
+}
+
+define <8 x i32> @avx2_psra_d_15_splat(<8 x i32> %v) nounwind readnone uwtable {
+; CHECK-LABEL: @avx2_psra_d_15_splat
+; CHECK: %1 = ashr <8 x i32> %v, <i32 31, i32 31, i32 31, i32 31, i32 31, i32 31, i32 31, i32 31>
+; CHECK: ret <8 x i32> %1
+  %1 = tail call <8 x i32> @llvm.x86.avx2.psra.d(<8 x i32> %v, <4 x i32> <i32 15, i32 15, i32 15, i32 15>)
+  ret <8 x i32> %1
+}
+
+define <8 x i32> @avx2_psra_d_64(<8 x i32> %v) nounwind readnone uwtable {
+; CHECK-LABEL: @avx2_psra_d_64
+; CHECK: %1 = ashr <8 x i32> %v, <i32 31, i32 31, i32 31, i32 31, i32 31, i32 31, i32 31, i32 31>
+; CHECK: ret <8 x i32> %1
+  %1 = tail call <8 x i32> @llvm.x86.avx2.psra.d(<8 x i32> %v, <4 x i32> <i32 64, i32 0, i32 9999, i32 9999>)
+  ret <8 x i32> %1
+}
+
+;
 ; LSHR - Constant Vector
 ;
 
+define <8 x i16> @sse2_psrl_w_0(<8 x i16> %v) nounwind readnone uwtable {
+; CHECK-LABEL: @sse2_psrl_w_0
+; CHECK: ret <8 x i16> %v
+  %1 = tail call <8 x i16> @llvm.x86.sse2.psrl.w(<8 x i16> %v, <8 x i16> zeroinitializer)
+  ret <8 x i16> %1
+}
+
 define <8 x i16> @sse2_psrl_w_15(<8 x i16> %v) nounwind readnone uwtable {
 ; CHECK-LABEL: @sse2_psrl_w_15
 ; CHECK: %1 = lshr <8 x i16> %v, <i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15>
@@ -285,6 +516,13 @@
   ret <8 x i16> %1
 }
 
+define <8 x i16> @sse2_psrl_w_15_splat(<8 x i16> %v) nounwind readnone uwtable {
+; CHECK-LABEL: @sse2_psrl_w_15_splat
+; CHECK: ret <8 x i16> zeroinitializer
+  %1 = tail call <8 x i16> @llvm.x86.sse2.psrl.w(<8 x i16> %v, <8 x i16> <i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15>)
+  ret <8 x i16> %1
+}
+
 define <8 x i16> @sse2_psrl_w_64(<8 x i16> %v) nounwind readnone uwtable {
 ; CHECK-LABEL: @sse2_psrl_w_64
 ; CHECK: ret <8 x i16> zeroinitializer
@@ -292,6 +530,13 @@
   ret <8 x i16> %1
 }
 
+define <4 x i32> @sse2_psrl_d_0(<4 x i32> %v) nounwind readnone uwtable {
+; CHECK-LABEL: @sse2_psrl_d_0
+; CHECK: ret <4 x i32> %v
+  %1 = tail call <4 x i32> @llvm.x86.sse2.psrl.d(<4 x i32> %v, <4 x i32> zeroinitializer)
+  ret <4 x i32> %1
+}
+
 define <4 x i32> @sse2_psrl_d_15(<4 x i32> %v) nounwind readnone uwtable {
 ; CHECK-LABEL: @sse2_psrl_d_15
 ; CHECK: %1 = lshr <4 x i32> %v, <i32 15, i32 15, i32 15, i32 15>
@@ -300,6 +545,13 @@
   ret <4 x i32> %1
 }
 
+define <4 x i32> @sse2_psrl_d_15_splat(<4 x i32> %v) nounwind readnone uwtable {
+; CHECK-LABEL: @sse2_psrl_d_15_splat
+; CHECK: ret <4 x i32> zeroinitializer
+  %1 = tail call <4 x i32> @llvm.x86.sse2.psrl.d(<4 x i32> %v, <4 x i32> <i32 15, i32 15, i32 15, i32 15>)
+  ret <4 x i32> %1
+}
+
 define <4 x i32> @sse2_psrl_d_64(<4 x i32> %v) nounwind readnone uwtable {
 ; CHECK-LABEL: @sse2_psrl_d_64
 ; CHECK: ret <4 x i32> zeroinitializer
@@ -307,6 +559,13 @@
   ret <4 x i32> %1
 }
 
+define <2 x i64> @sse2_psrl_q_0(<2 x i64> %v) nounwind readnone uwtable {
+; CHECK-LABEL: @sse2_psrl_q_0
+; CHECK: ret <2 x i64> %v
+  %1 = tail call <2 x i64> @llvm.x86.sse2.psrl.q(<2 x i64> %v, <2 x i64> zeroinitializer)
+  ret <2 x i64> %1
+}
+
 define <2 x i64> @sse2_psrl_q_15(<2 x i64> %v) nounwind readnone uwtable {
 ; CHECK-LABEL: @sse2_psrl_q_15
 ; CHECK: %1 = lshr <2 x i64> %v, <i64 15, i64 15>
@@ -322,6 +581,13 @@
   ret <2 x i64> %1
 }
 
+define <16 x i16> @avx2_psrl_w_0(<16 x i16> %v) nounwind readnone uwtable {
+; CHECK-LABEL: @avx2_psrl_w_0
+; CHECK: ret <16 x i16> %v
+  %1 = tail call <16 x i16> @llvm.x86.avx2.psrl.w(<16 x i16> %v, <8 x i16> zeroinitializer)
+  ret <16 x i16> %1
+}
+
 define <16 x i16> @avx2_psrl_w_15(<16 x i16> %v) nounwind readnone uwtable {
 ; CHECK-LABEL: @avx2_psrl_w_15
 ; CHECK: %1 = lshr <16 x i16> %v, <i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15>
@@ -330,6 +596,13 @@
   ret <16 x i16> %1
 }
 
+define <16 x i16> @avx2_psrl_w_15_splat(<16 x i16> %v) nounwind readnone uwtable {
+; CHECK-LABEL: @avx2_psrl_w_15_splat
+; CHECK: ret <16 x i16> zeroinitializer
+  %1 = tail call <16 x i16> @llvm.x86.avx2.psrl.w(<16 x i16> %v, <8 x i16> <i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15>)
+  ret <16 x i16> %1
+}
+
 define <16 x i16> @avx2_psrl_w_64(<16 x i16> %v) nounwind readnone uwtable {
 ; CHECK-LABEL: @avx2_psrl_w_64
 ; CHECK: ret <16 x i16> zeroinitializer
@@ -337,6 +610,13 @@
   ret <16 x i16> %1
 }
 
+define <8 x i32> @avx2_psrl_d_0(<8 x i32> %v) nounwind readnone uwtable {
+; CHECK-LABEL: @avx2_psrl_d_0
+; CHECK: ret <8 x i32> %v
+  %1 = tail call <8 x i32> @llvm.x86.avx2.psrl.d(<8 x i32> %v, <4 x i32> zeroinitializer)
+  ret <8 x i32> %1
+}
+
 define <8 x i32> @avx2_psrl_d_15(<8 x i32> %v) nounwind readnone uwtable {
 ; CHECK-LABEL: @avx2_psrl_d_15
 ; CHECK: %1 = lshr <8 x i32> %v, <i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15>
@@ -345,6 +625,13 @@
   ret <8 x i32> %1
 }
 
+define <8 x i32> @avx2_psrl_d_15_splat(<8 x i32> %v) nounwind readnone uwtable {
+; CHECK-LABEL: @avx2_psrl_d_15_splat
+; CHECK: ret <8 x i32> zeroinitializer
+  %1 = tail call <8 x i32> @llvm.x86.avx2.psrl.d(<8 x i32> %v, <4 x i32> <i32 15, i32 15, i32 15, i32 15>)
+  ret <8 x i32> %1
+}
+
 define <8 x i32> @avx2_psrl_d_64(<8 x i32> %v) nounwind readnone uwtable {
 ; CHECK-LABEL: @avx2_psrl_d_64
 ; CHECK: ret <8 x i32> zeroinitializer
@@ -352,6 +639,13 @@
   ret <8 x i32> %1
 }
 
+define <4 x i64> @avx2_psrl_q_0(<4 x i64> %v) nounwind readnone uwtable {
+; CHECK-LABEL: @avx2_psrl_q_0
+; CHECK: ret <4 x i64> %v
+  %1 = tail call <4 x i64> @llvm.x86.avx2.psrl.q(<4 x i64> %v, <2 x i64> zeroinitializer)
+  ret <4 x i64> %1
+}
+
 define <4 x i64> @avx2_psrl_q_15(<4 x i64> %v) nounwind readnone uwtable {
 ; CHECK-LABEL: @avx2_psrl_q_15
 ; CHECK: %1 = lshr <4 x i64> %v, <i64 15, i64 15, i64 15, i64 15>
@@ -371,6 +665,13 @@
 ; SHL - Constant Vector
 ;
 
+define <8 x i16> @sse2_psll_w_0(<8 x i16> %v) nounwind readnone uwtable {
+; CHECK-LABEL: @sse2_psll_w_0
+; CHECK: ret <8 x i16> %v
+  %1 = tail call <8 x i16> @llvm.x86.sse2.psll.w(<8 x i16> %v, <8 x i16> zeroinitializer)
+  ret <8 x i16> %1
+}
+
 define <8 x i16> @sse2_psll_w_15(<8 x i16> %v) nounwind readnone uwtable {
 ; CHECK-LABEL: @sse2_psll_w_15
 ; CHECK: %1 = shl <8 x i16> %v, <i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15>
@@ -379,6 +680,13 @@
   ret <8 x i16> %1
 }
 
+define <8 x i16> @sse2_psll_w_15_splat(<8 x i16> %v) nounwind readnone uwtable {
+; CHECK-LABEL: @sse2_psll_w_15_splat
+; CHECK: ret <8 x i16> zeroinitializer
+  %1 = tail call <8 x i16> @llvm.x86.sse2.psll.w(<8 x i16> %v, <8 x i16> <i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15>)
+  ret <8 x i16> %1
+}
+
 define <8 x i16> @sse2_psll_w_64(<8 x i16> %v) nounwind readnone uwtable {
 ; CHECK-LABEL: @sse2_psll_w_64
 ; CHECK: ret <8 x i16> zeroinitializer
@@ -386,6 +694,13 @@
   ret <8 x i16> %1
 }
 
+define <4 x i32> @sse2_psll_d_0(<4 x i32> %v) nounwind readnone uwtable {
+; CHECK-LABEL: @sse2_psll_d_0
+; CHECK: ret <4 x i32> %v
+  %1 = tail call <4 x i32> @llvm.x86.sse2.psll.d(<4 x i32> %v, <4 x i32> zeroinitializer)
+  ret <4 x i32> %1
+}
+
 define <4 x i32> @sse2_psll_d_15(<4 x i32> %v) nounwind readnone uwtable {
 ; CHECK-LABEL: @sse2_psll_d_15
 ; CHECK: %1 = shl <4 x i32> %v, <i32 15, i32 15, i32 15, i32 15>
@@ -394,6 +709,13 @@
   ret <4 x i32> %1
 }
 
+define <4 x i32> @sse2_psll_d_15_splat(<4 x i32> %v) nounwind readnone uwtable {
+; CHECK-LABEL: @sse2_psll_d_15_splat
+; CHECK: ret <4 x i32> zeroinitializer
+  %1 = tail call <4 x i32> @llvm.x86.sse2.psll.d(<4 x i32> %v, <4 x i32> <i32 15, i32 15, i32 15, i32 15>)
+  ret <4 x i32> %1
+}
+
 define <4 x i32> @sse2_psll_d_64(<4 x i32> %v) nounwind readnone uwtable {
 ; CHECK-LABEL: @sse2_psll_d_64
 ; CHECK: ret <4 x i32> zeroinitializer
@@ -401,6 +723,13 @@
   ret <4 x i32> %1
 }
 
+define <2 x i64> @sse2_psll_q_0(<2 x i64> %v) nounwind readnone uwtable {
+; CHECK-LABEL: @sse2_psll_q_0
+; CHECK: ret <2 x i64> %v
+  %1 = tail call <2 x i64> @llvm.x86.sse2.psll.q(<2 x i64> %v, <2 x i64> zeroinitializer)
+  ret <2 x i64> %1
+}
+
 define <2 x i64> @sse2_psll_q_15(<2 x i64> %v) nounwind readnone uwtable {
 ; CHECK-LABEL: @sse2_psll_q_15
 ; CHECK: %1 = shl <2 x i64> %v, <i64 15, i64 15>
@@ -416,6 +745,13 @@
   ret <2 x i64> %1
 }
 
+define <16 x i16> @avx2_psll_w_0(<16 x i16> %v) nounwind readnone uwtable {
+; CHECK-LABEL: @avx2_psll_w_0
+; CHECK: ret <16 x i16> %v
+  %1 = tail call <16 x i16> @llvm.x86.avx2.psll.w(<16 x i16> %v, <8 x i16> zeroinitializer)
+  ret <16 x i16> %1
+}
+
 define <16 x i16> @avx2_psll_w_15(<16 x i16> %v) nounwind readnone uwtable {
 ; CHECK-LABEL: @avx2_psll_w_15
 ; CHECK: %1 = shl <16 x i16> %v, <i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15>
@@ -424,6 +760,13 @@
   ret <16 x i16> %1
 }
 
+define <16 x i16> @avx2_psll_w_15_splat(<16 x i16> %v) nounwind readnone uwtable {
+; CHECK-LABEL: @avx2_psll_w_15_splat
+; CHECK: ret <16 x i16> zeroinitializer
+  %1 = tail call <16 x i16> @llvm.x86.avx2.psll.w(<16 x i16> %v, <8 x i16> <i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15>)
+  ret <16 x i16> %1
+}
+
 define <16 x i16> @avx2_psll_w_64(<16 x i16> %v) nounwind readnone uwtable {
 ; CHECK-LABEL: @avx2_psll_w_64
 ; CHECK: ret <16 x i16> zeroinitializer
@@ -431,6 +774,13 @@
   ret <16 x i16> %1
 }
 
+define <8 x i32> @avx2_psll_d_0(<8 x i32> %v) nounwind readnone uwtable {
+; CHECK-LABEL: @avx2_psll_d_0
+; CHECK: ret <8 x i32> %v
+  %1 = tail call <8 x i32> @llvm.x86.avx2.psll.d(<8 x i32> %v, <4 x i32> zeroinitializer)
+  ret <8 x i32> %1
+}
+
 define <8 x i32> @avx2_psll_d_15(<8 x i32> %v) nounwind readnone uwtable {
 ; CHECK-LABEL: @avx2_psll_d_15
 ; CHECK: %1 = shl <8 x i32> %v, <i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15>
@@ -439,6 +789,13 @@
   ret <8 x i32> %1
 }
 
+define <8 x i32> @avx2_psll_d_15_splat(<8 x i32> %v) nounwind readnone uwtable {
+; CHECK-LABEL: @avx2_psll_d_15_splat
+; CHECK: ret <8 x i32> zeroinitializer
+  %1 = tail call <8 x i32> @llvm.x86.avx2.psll.d(<8 x i32> %v, <4 x i32> <i32 15, i32 15, i32 15, i32 15>)
+  ret <8 x i32> %1
+}
+
 define <8 x i32> @avx2_psll_d_64(<8 x i32> %v) nounwind readnone uwtable {
 ; CHECK-LABEL: @avx2_psll_d_64
 ; CHECK: ret <8 x i32> zeroinitializer
@@ -446,6 +803,13 @@
   ret <8 x i32> %1
 }
 
+define <4 x i64> @avx2_psll_q_0(<4 x i64> %v) nounwind readnone uwtable {
+; CHECK-LABEL: @avx2_psll_q_0
+; CHECK: ret <4 x i64> %v
+  %1 = tail call <4 x i64> @llvm.x86.avx2.psll.q(<4 x i64> %v, <2 x i64> zeroinitializer)
+  ret <4 x i64> %1
+}
+
 define <4 x i64> @avx2_psll_q_15(<4 x i64> %v) nounwind readnone uwtable {
 ; CHECK-LABEL: @avx2_psll_q_15
 ; CHECK: %1 = shl <4 x i64> %v, <i64 15, i64 15, i64 15, i64 15>
@@ -660,6 +1024,7 @@
 declare <2 x i64> @llvm.x86.sse2.psll.q(<2 x i64>, <2 x i64>) #1
 declare <4 x i32> @llvm.x86.sse2.psll.d(<4 x i32>, <4 x i32>) #1
 declare <8 x i16> @llvm.x86.sse2.psll.w(<8 x i16>, <8 x i16>) #1
+
 declare <4 x i64> @llvm.x86.avx2.psrli.q(<4 x i64>, i32) #1
 declare <8 x i32> @llvm.x86.avx2.psrli.d(<8 x i32>, i32) #1
 declare <16 x i16> @llvm.x86.avx2.psrli.w(<16 x i16>, i32) #1
@@ -673,4 +1038,13 @@
 declare <4 x i32> @llvm.x86.sse2.psrl.d(<4 x i32>, <4 x i32>) #1
 declare <8 x i16> @llvm.x86.sse2.psrl.w(<8 x i16>, <8 x i16>) #1
 
+declare <8 x i32> @llvm.x86.avx2.psrai.d(<8 x i32>, i32) #1
+declare <16 x i16> @llvm.x86.avx2.psrai.w(<16 x i16>, i32) #1
+declare <8 x i32> @llvm.x86.avx2.psra.d(<8 x i32>, <4 x i32>) #1
+declare <16 x i16> @llvm.x86.avx2.psra.w(<16 x i16>, <8 x i16>) #1
+declare <4 x i32> @llvm.x86.sse2.psrai.d(<4 x i32>, i32) #1
+declare <8 x i16> @llvm.x86.sse2.psrai.w(<8 x i16>, i32) #1
+declare <4 x i32> @llvm.x86.sse2.psra.d(<4 x i32>, <4 x i32>) #1
+declare <8 x i16> @llvm.x86.sse2.psra.w(<8 x i16>, <8 x i16>) #1
+
 attributes #1 = { nounwind readnone }