[clang] [CIR][AArch64] Complete upstream vector saturating shift left NEON builtins (PR #221507)

Vicky Nguyen via cfe-commits cfe-commits at lists.llvm.org
Sat Sep 12 22:25:44 PDT 2026


https://github.com/iamvickynguyen updated https://github.com/llvm/llvm-project/pull/221507

>From c94c2644ab4f8585b29a5c636fb8cd8b32314ca1 Mon Sep 17 00:00:00 2001
From: Vicky Nguyen <vicky.trucviennguyen at gmail.com>
Date: Sat, 5 Sep 2026 17:45:07 -0700
Subject: [PATCH] [CIR][AArch64] Complete upstream vector saturating shift left
 NEON builtins

---
 .../include/clang/Basic/AArch64CodeGenUtils.h |  26 +-
 .../lib/CIR/CodeGen/CIRGenBuiltinAArch64.cpp  |  51 +-
 clang/test/CodeGen/AArch64/neon-intrinsics.c  | 704 ----------------
 clang/test/CodeGen/AArch64/neon/intrinsics.c  | 796 +++++++++++++++++-
 4 files changed, 853 insertions(+), 724 deletions(-)

diff --git a/clang/include/clang/Basic/AArch64CodeGenUtils.h b/clang/include/clang/Basic/AArch64CodeGenUtils.h
index 72b162ec459099..b8267f2939b97d 100644
--- a/clang/include/clang/Basic/AArch64CodeGenUtils.h
+++ b/clang/include/clang/Basic/AArch64CodeGenUtils.h
@@ -46,12 +46,18 @@ enum {
   // the intrinsic's .td.
   ArgAsWidenedRetType = (1 << 9),
 
+  // The immediate operand is `LLVMMatchType<0>` in the intrinsic
+  // definition, so it is vectorized like a data operand (e.g. the `_n_`
+  // scalar saturating shifts: vqshlb_n_s8, vqshlub_n_s8, ...).
+  // ClangIR-only; classic code-gen derives this from the intrinsic's .td.
+  VectorizeImmArg = (1 << 10),
+
   Vectorize1ArgType = Add1ArgType | VectorizeArgTypes,
   VectorRet = AddRetType | VectorizeRetType,
-  VectorRetGetArgs01 =
-      AddRetType | Add2ArgTypes | VectorizeRetType | VectorizeArgTypes,
-  FpCmpzModifiers =
-      AddRetType | VectorizeRetType | Add1ArgType | InventFloatType
+  VectorRetGetArgs01 = AddRetType | Add2ArgTypes | VectorizeRetType |
+      VectorizeArgTypes,
+  FpCmpzModifiers = AddRetType | VectorizeRetType | Add1ArgType |
+      InventFloatType
 };
 
 /// Describes an ARM or AArch64 NEON intrinsic, or an AArch64 SISD intrinsic.
@@ -592,22 +598,22 @@ const inline ARMNeonVectorIntrinsicInfo AArch64SISDIntrinsicMap[] = {
   NEONMAP1(vqrshrund_n_s64, aarch64_neon_sqrshrun, AddRetType),
   NEONMAP1(vqrshrunh_n_s16, aarch64_neon_sqrshrun, VectorRet | Use64BitVectors | ArgAsWidenedRetType),
   NEONMAP1(vqrshruns_n_s32, aarch64_neon_sqrshrun, VectorRet | Use64BitVectors | ArgAsWidenedRetType),
-  NEONMAP1(vqshlb_n_s8, aarch64_neon_sqshl, Vectorize1ArgType | Use64BitVectors),
-  NEONMAP1(vqshlb_n_u8, aarch64_neon_uqshl, Vectorize1ArgType | Use64BitVectors),
+  NEONMAP1(vqshlb_n_s8, aarch64_neon_sqshl, Vectorize1ArgType | Use64BitVectors | VectorizeImmArg),
+  NEONMAP1(vqshlb_n_u8, aarch64_neon_uqshl, Vectorize1ArgType | Use64BitVectors | VectorizeImmArg),
   NEONMAP1(vqshlb_s8, aarch64_neon_sqshl, Vectorize1ArgType | Use64BitVectors),
   NEONMAP1(vqshlb_u8, aarch64_neon_uqshl, Vectorize1ArgType | Use64BitVectors),
   NEONMAP1(vqshld_s64, aarch64_neon_sqshl, Add1ArgType),
   NEONMAP1(vqshld_u64, aarch64_neon_uqshl, Add1ArgType),
-  NEONMAP1(vqshlh_n_s16, aarch64_neon_sqshl, Vectorize1ArgType | Use64BitVectors),
-  NEONMAP1(vqshlh_n_u16, aarch64_neon_uqshl, Vectorize1ArgType | Use64BitVectors),
+  NEONMAP1(vqshlh_n_s16, aarch64_neon_sqshl, Vectorize1ArgType | Use64BitVectors | VectorizeImmArg),
+  NEONMAP1(vqshlh_n_u16, aarch64_neon_uqshl, Vectorize1ArgType | Use64BitVectors | VectorizeImmArg),
   NEONMAP1(vqshlh_s16, aarch64_neon_sqshl, Vectorize1ArgType | Use64BitVectors),
   NEONMAP1(vqshlh_u16, aarch64_neon_uqshl, Vectorize1ArgType | Use64BitVectors),
   NEONMAP1(vqshls_n_s32, aarch64_neon_sqshl, Add1ArgType),
   NEONMAP1(vqshls_n_u32, aarch64_neon_uqshl, Add1ArgType),
   NEONMAP1(vqshls_s32, aarch64_neon_sqshl, Add1ArgType),
   NEONMAP1(vqshls_u32, aarch64_neon_uqshl, Add1ArgType),
-  NEONMAP1(vqshlub_n_s8, aarch64_neon_sqshlu, Vectorize1ArgType | Use64BitVectors),
-  NEONMAP1(vqshluh_n_s16, aarch64_neon_sqshlu, Vectorize1ArgType | Use64BitVectors),
+  NEONMAP1(vqshlub_n_s8, aarch64_neon_sqshlu, Vectorize1ArgType | Use64BitVectors | VectorizeImmArg),
+  NEONMAP1(vqshluh_n_s16, aarch64_neon_sqshlu, Vectorize1ArgType | Use64BitVectors | VectorizeImmArg),
   NEONMAP1(vqshlus_n_s32, aarch64_neon_sqshlu, Add1ArgType),
   NEONMAP1(vqshrnd_n_s64, aarch64_neon_sqshrn, AddRetType),
   NEONMAP1(vqshrnd_n_u64, aarch64_neon_uqshrn, AddRetType),
diff --git a/clang/lib/CIR/CodeGen/CIRGenBuiltinAArch64.cpp b/clang/lib/CIR/CodeGen/CIRGenBuiltinAArch64.cpp
index 7f59176bc6268c..83cd4255a5bb4e 100644
--- a/clang/lib/CIR/CodeGen/CIRGenBuiltinAArch64.cpp
+++ b/clang/lib/CIR/CodeGen/CIRGenBuiltinAArch64.cpp
@@ -321,15 +321,19 @@ deriveNeonSISDIntrinsicOperandTypes(CIRGenFunction &cgf, unsigned modifier,
   // `ArgAsWidenedRetType`. When set, wrap every non-immediate data operand
   // that has the same scalar type as arg0. Checking the ICE bitmap prevents
   // an i32 immediate from being vectorized when it has the same type as a
-  // data operand (e.g. vqshrns_n_s32).
+  // data operand (e.g. vqshrns_n_s32). `VectorizeImmArg` marks the entries
+  // whose trailing immediate must be widened as well.
+  const bool widenTrailingImm = vecArgTy && ops.size() == 2 &&
+                                (modifier & VectorizeImmArg) &&
+                                (iceArguments & (1U << 1));
   llvm::SmallVector<mlir::Type> argTypes;
   argTypes.reserve(ops.size());
   for (unsigned i = 0, e = ops.size(); i != e; ++i) {
     bool isImmediate = iceArguments & (1U << i);
-    if (vecArgTy && !isImmediate && matchesArg0Ty(ops[i].getType()))
-      argTypes.push_back(vecArgTy);
-    else
-      argTypes.push_back(ops[i].getType());
+    bool asVector =
+        vecArgTy && ((!isImmediate && matchesArg0Ty(ops[i].getType())) ||
+                     (widenTrailingImm && i == 1));
+    argTypes.push_back(asVector ? mlir::Type(vecArgTy) : ops[i].getType());
   }
 
   return {funcResTy, std::move(argTypes)};
@@ -582,6 +586,23 @@ emitCommonNeonSISDBuiltinExpr(CIRGenFunction &cgf,
   case NEON::BI__builtin_neon_vqrshrnh_n_u16:
   case NEON::BI__builtin_neon_vqrshruns_n_s32:
   case NEON::BI__builtin_neon_vqrshrunh_n_s16:
+  case NEON::BI__builtin_neon_vqshlb_s8:
+  case NEON::BI__builtin_neon_vqshlb_u8:
+  case NEON::BI__builtin_neon_vqshlh_s16:
+  case NEON::BI__builtin_neon_vqshlh_u16:
+  case NEON::BI__builtin_neon_vqshls_s32:
+  case NEON::BI__builtin_neon_vqshls_u32:
+  case NEON::BI__builtin_neon_vqshld_s64:
+  case NEON::BI__builtin_neon_vqshld_u64:
+  case NEON::BI__builtin_neon_vqshlb_n_s8:
+  case NEON::BI__builtin_neon_vqshlb_n_u8:
+  case NEON::BI__builtin_neon_vqshlh_n_s16:
+  case NEON::BI__builtin_neon_vqshlh_n_u16:
+  case NEON::BI__builtin_neon_vqshls_n_s32:
+  case NEON::BI__builtin_neon_vqshls_n_u32:
+  case NEON::BI__builtin_neon_vqshlub_n_s8:
+  case NEON::BI__builtin_neon_vqshluh_n_s16:
+  case NEON::BI__builtin_neon_vqshlus_n_s32:
     break;
   }
 
@@ -1201,10 +1222,24 @@ static mlir::Value emitCommonNeonBuiltinExpr(
   case NEON::BI__builtin_neon_vqdmulh_laneq_v:
   case NEON::BI__builtin_neon_vqrdmulhq_laneq_v:
   case NEON::BI__builtin_neon_vqrdmulh_laneq_v:
+    cgf.cgm.errorNYI(expr->getSourceRange(),
+                     std::string("unimplemented AArch64 builtin call: ") +
+                         ctx.BuiltinInfo.getName(builtinID));
+    return mlir::Value{};
   case NEON::BI__builtin_neon_vqshl_n_v:
-  case NEON::BI__builtin_neon_vqshlq_n_v:
+  case NEON::BI__builtin_neon_vqshlq_n_v: {
+    llvm::StringRef intrName =
+        usgn ? "aarch64.neon.uqshl" : "aarch64.neon.sqshl";
+    return emitNeonCall(cgf.cgm, cgf.getBuilder(), {ty, ty}, ops, intrName, ty,
+                        loc, /*isConstrainedFPIntrinsic=*/false, /*shift=*/1,
+                        /*rightshift=*/false);
+  }
   case NEON::BI__builtin_neon_vqshlu_n_v:
   case NEON::BI__builtin_neon_vqshluq_n_v:
+    return emitNeonCall(cgf.cgm, cgf.getBuilder(), {ty, ty}, ops,
+                        "aarch64.neon.sqshlu", ty, loc,
+                        /*isConstrainedFPIntrinsic=*/false, /*shift=*/1,
+                        /*rightshift=*/false);
   case NEON::BI__builtin_neon_vrecpe_v:
   case NEON::BI__builtin_neon_vrecpeq_v:
   case NEON::BI__builtin_neon_vrsqrte_v:
@@ -1392,7 +1427,9 @@ static mlir::Value emitCommonNeonBuiltinExpr(
   case NEON::BI__builtin_neon_vshl_v:
   case NEON::BI__builtin_neon_vshlq_v:
   case NEON::BI__builtin_neon_vraddhn_v:
-  case NEON::BI__builtin_neon_vrsubhn_v: {
+  case NEON::BI__builtin_neon_vrsubhn_v:
+  case NEON::BI__builtin_neon_vqshl_v:
+  case NEON::BI__builtin_neon_vqshlq_v: {
     // Pick the signed/unsigned intrinsic when the builtin has both
     // (UnsignedAlts); otherwise there is a single intrinsic.
     unsigned intrinsic =
diff --git a/clang/test/CodeGen/AArch64/neon-intrinsics.c b/clang/test/CodeGen/AArch64/neon-intrinsics.c
index 7a945abca9a80d..2098abb0978893 100644
--- a/clang/test/CodeGen/AArch64/neon-intrinsics.c
+++ b/clang/test/CodeGen/AArch64/neon-intrinsics.c
@@ -2325,240 +2325,6 @@ uint64x2_t test_vcltq_f64(float64x2_t v1, float64x2_t v2) {
   return vcltq_f64(v1, v2);
 }
 
-// CHECK-LABEL: define dso_local <8 x i8> @test_vqshl_s8(
-// CHECK-SAME: <8 x i8> noundef [[A:%.*]], <8 x i8> noundef [[B:%.*]]) #[[ATTR0]] {
-// CHECK-NEXT:  [[ENTRY:.*:]]
-// CHECK-NEXT:    [[VQSHL_V_I:%.*]] = call <8 x i8> @llvm.aarch64.neon.sqshl.v8i8(<8 x i8> [[A]], <8 x i8> [[B]])
-// CHECK-NEXT:    ret <8 x i8> [[VQSHL_V_I]]
-//
-int8x8_t test_vqshl_s8(int8x8_t a, int8x8_t b) {
-  return vqshl_s8(a, b);
-}
-
-// CHECK-LABEL: define dso_local <4 x i16> @test_vqshl_s16(
-// CHECK-SAME: <4 x i16> noundef [[A:%.*]], <4 x i16> noundef [[B:%.*]]) #[[ATTR0]] {
-// CHECK-NEXT:  [[ENTRY:.*:]]
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i16> [[A]] to <8 x i8>
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast <4 x i16> [[B]] to <8 x i8>
-// CHECK-NEXT:    [[VQSHL_V_I:%.*]] = bitcast <8 x i8> [[TMP0]] to <4 x i16>
-// CHECK-NEXT:    [[VQSHL_V1_I:%.*]] = bitcast <8 x i8> [[TMP1]] to <4 x i16>
-// CHECK-NEXT:    [[VQSHL_V2_I:%.*]] = call <4 x i16> @llvm.aarch64.neon.sqshl.v4i16(<4 x i16> [[VQSHL_V_I]], <4 x i16> [[VQSHL_V1_I]])
-// CHECK-NEXT:    [[VQSHL_V3_I:%.*]] = bitcast <4 x i16> [[VQSHL_V2_I]] to <8 x i8>
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast <8 x i8> [[VQSHL_V3_I]] to <4 x i16>
-// CHECK-NEXT:    ret <4 x i16> [[TMP2]]
-//
-int16x4_t test_vqshl_s16(int16x4_t a, int16x4_t b) {
-  return vqshl_s16(a, b);
-}
-
-// CHECK-LABEL: define dso_local <2 x i32> @test_vqshl_s32(
-// CHECK-SAME: <2 x i32> noundef [[A:%.*]], <2 x i32> noundef [[B:%.*]]) #[[ATTR0]] {
-// CHECK-NEXT:  [[ENTRY:.*:]]
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x i32> [[A]] to <8 x i8>
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast <2 x i32> [[B]] to <8 x i8>
-// CHECK-NEXT:    [[VQSHL_V_I:%.*]] = bitcast <8 x i8> [[TMP0]] to <2 x i32>
-// CHECK-NEXT:    [[VQSHL_V1_I:%.*]] = bitcast <8 x i8> [[TMP1]] to <2 x i32>
-// CHECK-NEXT:    [[VQSHL_V2_I:%.*]] = call <2 x i32> @llvm.aarch64.neon.sqshl.v2i32(<2 x i32> [[VQSHL_V_I]], <2 x i32> [[VQSHL_V1_I]])
-// CHECK-NEXT:    [[VQSHL_V3_I:%.*]] = bitcast <2 x i32> [[VQSHL_V2_I]] to <8 x i8>
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast <8 x i8> [[VQSHL_V3_I]] to <2 x i32>
-// CHECK-NEXT:    ret <2 x i32> [[TMP2]]
-//
-int32x2_t test_vqshl_s32(int32x2_t a, int32x2_t b) {
-  return vqshl_s32(a, b);
-}
-
-// CHECK-LABEL: define dso_local <1 x i64> @test_vqshl_s64(
-// CHECK-SAME: <1 x i64> noundef [[A:%.*]], <1 x i64> noundef [[B:%.*]]) #[[ATTR0]] {
-// CHECK-NEXT:  [[ENTRY:.*:]]
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast <1 x i64> [[A]] to <8 x i8>
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast <1 x i64> [[B]] to <8 x i8>
-// CHECK-NEXT:    [[VQSHL_V_I:%.*]] = bitcast <8 x i8> [[TMP0]] to <1 x i64>
-// CHECK-NEXT:    [[VQSHL_V1_I:%.*]] = bitcast <8 x i8> [[TMP1]] to <1 x i64>
-// CHECK-NEXT:    [[VQSHL_V2_I:%.*]] = call <1 x i64> @llvm.aarch64.neon.sqshl.v1i64(<1 x i64> [[VQSHL_V_I]], <1 x i64> [[VQSHL_V1_I]])
-// CHECK-NEXT:    [[VQSHL_V3_I:%.*]] = bitcast <1 x i64> [[VQSHL_V2_I]] to <8 x i8>
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast <8 x i8> [[VQSHL_V3_I]] to i64
-// CHECK-NEXT:    [[REF_TMP_I_SROA_0_0_VEC_INSERT:%.*]] = insertelement <1 x i64> undef, i64 [[TMP2]], i64 0
-// CHECK-NEXT:    ret <1 x i64> [[REF_TMP_I_SROA_0_0_VEC_INSERT]]
-//
-int64x1_t test_vqshl_s64(int64x1_t a, int64x1_t b) {
-  return vqshl_s64(a, b);
-}
-
-// CHECK-LABEL: define dso_local <8 x i8> @test_vqshl_u8(
-// CHECK-SAME: <8 x i8> noundef [[A:%.*]], <8 x i8> noundef [[B:%.*]]) #[[ATTR0]] {
-// CHECK-NEXT:  [[ENTRY:.*:]]
-// CHECK-NEXT:    [[VQSHL_V_I:%.*]] = call <8 x i8> @llvm.aarch64.neon.uqshl.v8i8(<8 x i8> [[A]], <8 x i8> [[B]])
-// CHECK-NEXT:    ret <8 x i8> [[VQSHL_V_I]]
-//
-uint8x8_t test_vqshl_u8(uint8x8_t a, int8x8_t b) {
-  return vqshl_u8(a, b);
-}
-
-// CHECK-LABEL: define dso_local <4 x i16> @test_vqshl_u16(
-// CHECK-SAME: <4 x i16> noundef [[A:%.*]], <4 x i16> noundef [[B:%.*]]) #[[ATTR0]] {
-// CHECK-NEXT:  [[ENTRY:.*:]]
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i16> [[A]] to <8 x i8>
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast <4 x i16> [[B]] to <8 x i8>
-// CHECK-NEXT:    [[VQSHL_V_I:%.*]] = bitcast <8 x i8> [[TMP0]] to <4 x i16>
-// CHECK-NEXT:    [[VQSHL_V1_I:%.*]] = bitcast <8 x i8> [[TMP1]] to <4 x i16>
-// CHECK-NEXT:    [[VQSHL_V2_I:%.*]] = call <4 x i16> @llvm.aarch64.neon.uqshl.v4i16(<4 x i16> [[VQSHL_V_I]], <4 x i16> [[VQSHL_V1_I]])
-// CHECK-NEXT:    [[VQSHL_V3_I:%.*]] = bitcast <4 x i16> [[VQSHL_V2_I]] to <8 x i8>
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast <8 x i8> [[VQSHL_V3_I]] to <4 x i16>
-// CHECK-NEXT:    ret <4 x i16> [[TMP2]]
-//
-uint16x4_t test_vqshl_u16(uint16x4_t a, int16x4_t b) {
-  return vqshl_u16(a, b);
-}
-
-// CHECK-LABEL: define dso_local <2 x i32> @test_vqshl_u32(
-// CHECK-SAME: <2 x i32> noundef [[A:%.*]], <2 x i32> noundef [[B:%.*]]) #[[ATTR0]] {
-// CHECK-NEXT:  [[ENTRY:.*:]]
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x i32> [[A]] to <8 x i8>
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast <2 x i32> [[B]] to <8 x i8>
-// CHECK-NEXT:    [[VQSHL_V_I:%.*]] = bitcast <8 x i8> [[TMP0]] to <2 x i32>
-// CHECK-NEXT:    [[VQSHL_V1_I:%.*]] = bitcast <8 x i8> [[TMP1]] to <2 x i32>
-// CHECK-NEXT:    [[VQSHL_V2_I:%.*]] = call <2 x i32> @llvm.aarch64.neon.uqshl.v2i32(<2 x i32> [[VQSHL_V_I]], <2 x i32> [[VQSHL_V1_I]])
-// CHECK-NEXT:    [[VQSHL_V3_I:%.*]] = bitcast <2 x i32> [[VQSHL_V2_I]] to <8 x i8>
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast <8 x i8> [[VQSHL_V3_I]] to <2 x i32>
-// CHECK-NEXT:    ret <2 x i32> [[TMP2]]
-//
-uint32x2_t test_vqshl_u32(uint32x2_t a, int32x2_t b) {
-  return vqshl_u32(a, b);
-}
-
-// CHECK-LABEL: define dso_local <1 x i64> @test_vqshl_u64(
-// CHECK-SAME: <1 x i64> noundef [[A:%.*]], <1 x i64> noundef [[B:%.*]]) #[[ATTR0]] {
-// CHECK-NEXT:  [[ENTRY:.*:]]
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast <1 x i64> [[A]] to <8 x i8>
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast <1 x i64> [[B]] to <8 x i8>
-// CHECK-NEXT:    [[VQSHL_V_I:%.*]] = bitcast <8 x i8> [[TMP0]] to <1 x i64>
-// CHECK-NEXT:    [[VQSHL_V1_I:%.*]] = bitcast <8 x i8> [[TMP1]] to <1 x i64>
-// CHECK-NEXT:    [[VQSHL_V2_I:%.*]] = call <1 x i64> @llvm.aarch64.neon.uqshl.v1i64(<1 x i64> [[VQSHL_V_I]], <1 x i64> [[VQSHL_V1_I]])
-// CHECK-NEXT:    [[VQSHL_V3_I:%.*]] = bitcast <1 x i64> [[VQSHL_V2_I]] to <8 x i8>
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast <8 x i8> [[VQSHL_V3_I]] to i64
-// CHECK-NEXT:    [[REF_TMP_I_SROA_0_0_VEC_INSERT:%.*]] = insertelement <1 x i64> undef, i64 [[TMP2]], i64 0
-// CHECK-NEXT:    ret <1 x i64> [[REF_TMP_I_SROA_0_0_VEC_INSERT]]
-//
-uint64x1_t test_vqshl_u64(uint64x1_t a, int64x1_t b) {
-  return vqshl_u64(a, b);
-}
-
-// CHECK-LABEL: define dso_local <16 x i8> @test_vqshlq_s8(
-// CHECK-SAME: <16 x i8> noundef [[A:%.*]], <16 x i8> noundef [[B:%.*]]) #[[ATTR0]] {
-// CHECK-NEXT:  [[ENTRY:.*:]]
-// CHECK-NEXT:    [[VQSHLQ_V_I:%.*]] = call <16 x i8> @llvm.aarch64.neon.sqshl.v16i8(<16 x i8> [[A]], <16 x i8> [[B]])
-// CHECK-NEXT:    ret <16 x i8> [[VQSHLQ_V_I]]
-//
-int8x16_t test_vqshlq_s8(int8x16_t a, int8x16_t b) {
-  return vqshlq_s8(a, b);
-}
-
-// CHECK-LABEL: define dso_local <8 x i16> @test_vqshlq_s16(
-// CHECK-SAME: <8 x i16> noundef [[A:%.*]], <8 x i16> noundef [[B:%.*]]) #[[ATTR0]] {
-// CHECK-NEXT:  [[ENTRY:.*:]]
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast <8 x i16> [[A]] to <16 x i8>
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i16> [[B]] to <16 x i8>
-// CHECK-NEXT:    [[VQSHLQ_V_I:%.*]] = bitcast <16 x i8> [[TMP0]] to <8 x i16>
-// CHECK-NEXT:    [[VQSHLQ_V1_I:%.*]] = bitcast <16 x i8> [[TMP1]] to <8 x i16>
-// CHECK-NEXT:    [[VQSHLQ_V2_I:%.*]] = call <8 x i16> @llvm.aarch64.neon.sqshl.v8i16(<8 x i16> [[VQSHLQ_V_I]], <8 x i16> [[VQSHLQ_V1_I]])
-// CHECK-NEXT:    [[VQSHLQ_V3_I:%.*]] = bitcast <8 x i16> [[VQSHLQ_V2_I]] to <16 x i8>
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast <16 x i8> [[VQSHLQ_V3_I]] to <8 x i16>
-// CHECK-NEXT:    ret <8 x i16> [[TMP2]]
-//
-int16x8_t test_vqshlq_s16(int16x8_t a, int16x8_t b) {
-  return vqshlq_s16(a, b);
-}
-
-// CHECK-LABEL: define dso_local <4 x i32> @test_vqshlq_s32(
-// CHECK-SAME: <4 x i32> noundef [[A:%.*]], <4 x i32> noundef [[B:%.*]]) #[[ATTR0]] {
-// CHECK-NEXT:  [[ENTRY:.*:]]
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[A]] to <16 x i8>
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast <4 x i32> [[B]] to <16 x i8>
-// CHECK-NEXT:    [[VQSHLQ_V_I:%.*]] = bitcast <16 x i8> [[TMP0]] to <4 x i32>
-// CHECK-NEXT:    [[VQSHLQ_V1_I:%.*]] = bitcast <16 x i8> [[TMP1]] to <4 x i32>
-// CHECK-NEXT:    [[VQSHLQ_V2_I:%.*]] = call <4 x i32> @llvm.aarch64.neon.sqshl.v4i32(<4 x i32> [[VQSHLQ_V_I]], <4 x i32> [[VQSHLQ_V1_I]])
-// CHECK-NEXT:    [[VQSHLQ_V3_I:%.*]] = bitcast <4 x i32> [[VQSHLQ_V2_I]] to <16 x i8>
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast <16 x i8> [[VQSHLQ_V3_I]] to <4 x i32>
-// CHECK-NEXT:    ret <4 x i32> [[TMP2]]
-//
-int32x4_t test_vqshlq_s32(int32x4_t a, int32x4_t b) {
-  return vqshlq_s32(a, b);
-}
-
-// CHECK-LABEL: define dso_local <2 x i64> @test_vqshlq_s64(
-// CHECK-SAME: <2 x i64> noundef [[A:%.*]], <2 x i64> noundef [[B:%.*]]) #[[ATTR0]] {
-// CHECK-NEXT:  [[ENTRY:.*:]]
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x i64> [[A]] to <16 x i8>
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast <2 x i64> [[B]] to <16 x i8>
-// CHECK-NEXT:    [[VQSHLQ_V_I:%.*]] = bitcast <16 x i8> [[TMP0]] to <2 x i64>
-// CHECK-NEXT:    [[VQSHLQ_V1_I:%.*]] = bitcast <16 x i8> [[TMP1]] to <2 x i64>
-// CHECK-NEXT:    [[VQSHLQ_V2_I:%.*]] = call <2 x i64> @llvm.aarch64.neon.sqshl.v2i64(<2 x i64> [[VQSHLQ_V_I]], <2 x i64> [[VQSHLQ_V1_I]])
-// CHECK-NEXT:    [[VQSHLQ_V3_I:%.*]] = bitcast <2 x i64> [[VQSHLQ_V2_I]] to <16 x i8>
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast <16 x i8> [[VQSHLQ_V3_I]] to <2 x i64>
-// CHECK-NEXT:    ret <2 x i64> [[TMP2]]
-//
-int64x2_t test_vqshlq_s64(int64x2_t a, int64x2_t b) {
-  return vqshlq_s64(a, b);
-}
-
-// CHECK-LABEL: define dso_local <16 x i8> @test_vqshlq_u8(
-// CHECK-SAME: <16 x i8> noundef [[A:%.*]], <16 x i8> noundef [[B:%.*]]) #[[ATTR0]] {
-// CHECK-NEXT:  [[ENTRY:.*:]]
-// CHECK-NEXT:    [[VQSHLQ_V_I:%.*]] = call <16 x i8> @llvm.aarch64.neon.uqshl.v16i8(<16 x i8> [[A]], <16 x i8> [[B]])
-// CHECK-NEXT:    ret <16 x i8> [[VQSHLQ_V_I]]
-//
-uint8x16_t test_vqshlq_u8(uint8x16_t a, int8x16_t b) {
-  return vqshlq_u8(a, b);
-}
-
-// CHECK-LABEL: define dso_local <8 x i16> @test_vqshlq_u16(
-// CHECK-SAME: <8 x i16> noundef [[A:%.*]], <8 x i16> noundef [[B:%.*]]) #[[ATTR0]] {
-// CHECK-NEXT:  [[ENTRY:.*:]]
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast <8 x i16> [[A]] to <16 x i8>
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast <8 x i16> [[B]] to <16 x i8>
-// CHECK-NEXT:    [[VQSHLQ_V_I:%.*]] = bitcast <16 x i8> [[TMP0]] to <8 x i16>
-// CHECK-NEXT:    [[VQSHLQ_V1_I:%.*]] = bitcast <16 x i8> [[TMP1]] to <8 x i16>
-// CHECK-NEXT:    [[VQSHLQ_V2_I:%.*]] = call <8 x i16> @llvm.aarch64.neon.uqshl.v8i16(<8 x i16> [[VQSHLQ_V_I]], <8 x i16> [[VQSHLQ_V1_I]])
-// CHECK-NEXT:    [[VQSHLQ_V3_I:%.*]] = bitcast <8 x i16> [[VQSHLQ_V2_I]] to <16 x i8>
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast <16 x i8> [[VQSHLQ_V3_I]] to <8 x i16>
-// CHECK-NEXT:    ret <8 x i16> [[TMP2]]
-//
-uint16x8_t test_vqshlq_u16(uint16x8_t a, int16x8_t b) {
-  return vqshlq_u16(a, b);
-}
-
-// CHECK-LABEL: define dso_local <4 x i32> @test_vqshlq_u32(
-// CHECK-SAME: <4 x i32> noundef [[A:%.*]], <4 x i32> noundef [[B:%.*]]) #[[ATTR0]] {
-// CHECK-NEXT:  [[ENTRY:.*:]]
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[A]] to <16 x i8>
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast <4 x i32> [[B]] to <16 x i8>
-// CHECK-NEXT:    [[VQSHLQ_V_I:%.*]] = bitcast <16 x i8> [[TMP0]] to <4 x i32>
-// CHECK-NEXT:    [[VQSHLQ_V1_I:%.*]] = bitcast <16 x i8> [[TMP1]] to <4 x i32>
-// CHECK-NEXT:    [[VQSHLQ_V2_I:%.*]] = call <4 x i32> @llvm.aarch64.neon.uqshl.v4i32(<4 x i32> [[VQSHLQ_V_I]], <4 x i32> [[VQSHLQ_V1_I]])
-// CHECK-NEXT:    [[VQSHLQ_V3_I:%.*]] = bitcast <4 x i32> [[VQSHLQ_V2_I]] to <16 x i8>
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast <16 x i8> [[VQSHLQ_V3_I]] to <4 x i32>
-// CHECK-NEXT:    ret <4 x i32> [[TMP2]]
-//
-uint32x4_t test_vqshlq_u32(uint32x4_t a, int32x4_t b) {
-  return vqshlq_u32(a, b);
-}
-
-// CHECK-LABEL: define dso_local <2 x i64> @test_vqshlq_u64(
-// CHECK-SAME: <2 x i64> noundef [[A:%.*]], <2 x i64> noundef [[B:%.*]]) #[[ATTR0]] {
-// CHECK-NEXT:  [[ENTRY:.*:]]
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x i64> [[A]] to <16 x i8>
-// CHECK-NEXT:    [[TMP1:%.*]] = bitcast <2 x i64> [[B]] to <16 x i8>
-// CHECK-NEXT:    [[VQSHLQ_V_I:%.*]] = bitcast <16 x i8> [[TMP0]] to <2 x i64>
-// CHECK-NEXT:    [[VQSHLQ_V1_I:%.*]] = bitcast <16 x i8> [[TMP1]] to <2 x i64>
-// CHECK-NEXT:    [[VQSHLQ_V2_I:%.*]] = call <2 x i64> @llvm.aarch64.neon.uqshl.v2i64(<2 x i64> [[VQSHLQ_V_I]], <2 x i64> [[VQSHLQ_V1_I]])
-// CHECK-NEXT:    [[VQSHLQ_V3_I:%.*]] = bitcast <2 x i64> [[VQSHLQ_V2_I]] to <16 x i8>
-// CHECK-NEXT:    [[TMP2:%.*]] = bitcast <16 x i8> [[VQSHLQ_V3_I]] to <2 x i64>
-// CHECK-NEXT:    ret <2 x i64> [[TMP2]]
-//
-uint64x2_t test_vqshlq_u64(uint64x2_t a, int64x2_t b) {
-  return vqshlq_u64(a, b);
-}
-
 // CHECK-LABEL: define dso_local <8 x i8> @test_vrshl_s8(
 // CHECK-SAME: <8 x i8> noundef [[A:%.*]], <8 x i8> noundef [[B:%.*]]) #[[ATTR0]] {
 // CHECK-NEXT:  [[ENTRY:.*:]]
@@ -3476,86 +3242,6 @@ float64x2_t test_vmulxq_f64(float64x2_t a, float64x2_t b) {
   return vmulxq_f64(a, b);
 }
 
-// CHECK-LABEL: define dso_local <8 x i8> @test_vqshlu_n_s8(
-// CHECK-SAME: <8 x i8> noundef [[A:%.*]]) #[[ATTR0]] {
-// CHECK-NEXT:  [[ENTRY:.*:]]
-// CHECK-NEXT:    [[VQSHLU_N:%.*]] = call <8 x i8> @llvm.aarch64.neon.sqshlu.v8i8(<8 x i8> [[A]], <8 x i8> splat (i8 3))
-// CHECK-NEXT:    ret <8 x i8> [[VQSHLU_N]]
-//
-uint8x8_t test_vqshlu_n_s8(int8x8_t a) {
-  return vqshlu_n_s8(a, 3);
-}
-
-// CHECK-LABEL: define dso_local <4 x i16> @test_vqshlu_n_s16(
-// CHECK-SAME: <4 x i16> noundef [[A:%.*]]) #[[ATTR0]] {
-// CHECK-NEXT:  [[ENTRY:.*:]]
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i16> [[A]] to <8 x i8>
-// CHECK-NEXT:    [[VQSHLU_N:%.*]] = bitcast <8 x i8> [[TMP0]] to <4 x i16>
-// CHECK-NEXT:    [[VQSHLU_N1:%.*]] = call <4 x i16> @llvm.aarch64.neon.sqshlu.v4i16(<4 x i16> [[VQSHLU_N]], <4 x i16> splat (i16 3))
-// CHECK-NEXT:    ret <4 x i16> [[VQSHLU_N1]]
-//
-uint16x4_t test_vqshlu_n_s16(int16x4_t a) {
-  return vqshlu_n_s16(a, 3);
-}
-
-// CHECK-LABEL: define dso_local <2 x i32> @test_vqshlu_n_s32(
-// CHECK-SAME: <2 x i32> noundef [[A:%.*]]) #[[ATTR0]] {
-// CHECK-NEXT:  [[ENTRY:.*:]]
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x i32> [[A]] to <8 x i8>
-// CHECK-NEXT:    [[VQSHLU_N:%.*]] = bitcast <8 x i8> [[TMP0]] to <2 x i32>
-// CHECK-NEXT:    [[VQSHLU_N1:%.*]] = call <2 x i32> @llvm.aarch64.neon.sqshlu.v2i32(<2 x i32> [[VQSHLU_N]], <2 x i32> splat (i32 3))
-// CHECK-NEXT:    ret <2 x i32> [[VQSHLU_N1]]
-//
-uint32x2_t test_vqshlu_n_s32(int32x2_t a) {
-  return vqshlu_n_s32(a, 3);
-}
-
-// CHECK-LABEL: define dso_local <16 x i8> @test_vqshluq_n_s8(
-// CHECK-SAME: <16 x i8> noundef [[A:%.*]]) #[[ATTR0]] {
-// CHECK-NEXT:  [[ENTRY:.*:]]
-// CHECK-NEXT:    [[VQSHLU_N:%.*]] = call <16 x i8> @llvm.aarch64.neon.sqshlu.v16i8(<16 x i8> [[A]], <16 x i8> splat (i8 3))
-// CHECK-NEXT:    ret <16 x i8> [[VQSHLU_N]]
-//
-uint8x16_t test_vqshluq_n_s8(int8x16_t a) {
-  return vqshluq_n_s8(a, 3);
-}
-
-// CHECK-LABEL: define dso_local <8 x i16> @test_vqshluq_n_s16(
-// CHECK-SAME: <8 x i16> noundef [[A:%.*]]) #[[ATTR0]] {
-// CHECK-NEXT:  [[ENTRY:.*:]]
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast <8 x i16> [[A]] to <16 x i8>
-// CHECK-NEXT:    [[VQSHLU_N:%.*]] = bitcast <16 x i8> [[TMP0]] to <8 x i16>
-// CHECK-NEXT:    [[VQSHLU_N1:%.*]] = call <8 x i16> @llvm.aarch64.neon.sqshlu.v8i16(<8 x i16> [[VQSHLU_N]], <8 x i16> splat (i16 3))
-// CHECK-NEXT:    ret <8 x i16> [[VQSHLU_N1]]
-//
-uint16x8_t test_vqshluq_n_s16(int16x8_t a) {
-  return vqshluq_n_s16(a, 3);
-}
-
-// CHECK-LABEL: define dso_local <4 x i32> @test_vqshluq_n_s32(
-// CHECK-SAME: <4 x i32> noundef [[A:%.*]]) #[[ATTR0]] {
-// CHECK-NEXT:  [[ENTRY:.*:]]
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[A]] to <16 x i8>
-// CHECK-NEXT:    [[VQSHLU_N:%.*]] = bitcast <16 x i8> [[TMP0]] to <4 x i32>
-// CHECK-NEXT:    [[VQSHLU_N1:%.*]] = call <4 x i32> @llvm.aarch64.neon.sqshlu.v4i32(<4 x i32> [[VQSHLU_N]], <4 x i32> splat (i32 3))
-// CHECK-NEXT:    ret <4 x i32> [[VQSHLU_N1]]
-//
-uint32x4_t test_vqshluq_n_s32(int32x4_t a) {
-  return vqshluq_n_s32(a, 3);
-}
-
-// CHECK-LABEL: define dso_local <2 x i64> @test_vqshluq_n_s64(
-// CHECK-SAME: <2 x i64> noundef [[A:%.*]]) #[[ATTR0]] {
-// CHECK-NEXT:  [[ENTRY:.*:]]
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x i64> [[A]] to <16 x i8>
-// CHECK-NEXT:    [[VQSHLU_N:%.*]] = bitcast <16 x i8> [[TMP0]] to <2 x i64>
-// CHECK-NEXT:    [[VQSHLU_N1:%.*]] = call <2 x i64> @llvm.aarch64.neon.sqshlu.v2i64(<2 x i64> [[VQSHLU_N]], <2 x i64> splat (i64 3))
-// CHECK-NEXT:    ret <2 x i64> [[VQSHLU_N1]]
-//
-uint64x2_t test_vqshluq_n_s64(int64x2_t a) {
-  return vqshluq_n_s64(a, 3);
-}
-
 // CHECK-LABEL: define dso_local <8 x i8> @test_vshrn_n_s16(
 // CHECK-SAME: <8 x i16> noundef [[A:%.*]]) #[[ATTR0]] {
 // CHECK-NEXT:  [[ENTRY:.*:]]
@@ -4828,98 +4514,6 @@ int64x2_t test_vqdmlsl_high_s32(int64x2_t a, int32x4_t b, int32x4_t c) {
   return vqdmlsl_high_s32(a, b, c);
 }
 
-// CHECK-LABEL: define dso_local i8 @test_vqshlb_s8(
-// CHECK-SAME: i8 noundef [[A:%.*]], i8 noundef [[B:%.*]]) #[[ATTR0]] {
-// CHECK-NEXT:  [[ENTRY:.*:]]
-// CHECK-NEXT:    [[TMP0:%.*]] = insertelement <8 x i8> poison, i8 [[A]], i64 0
-// CHECK-NEXT:    [[TMP1:%.*]] = insertelement <8 x i8> poison, i8 [[B]], i64 0
-// CHECK-NEXT:    [[VQSHLB_S8_I:%.*]] = call <8 x i8> @llvm.aarch64.neon.sqshl.v8i8(<8 x i8> [[TMP0]], <8 x i8> [[TMP1]])
-// CHECK-NEXT:    [[TMP2:%.*]] = extractelement <8 x i8> [[VQSHLB_S8_I]], i64 0
-// CHECK-NEXT:    ret i8 [[TMP2]]
-//
-int8_t test_vqshlb_s8(int8_t a, int8_t b) {
-  return vqshlb_s8(a, b);
-}
-
-// CHECK-LABEL: define dso_local i16 @test_vqshlh_s16(
-// CHECK-SAME: i16 noundef [[A:%.*]], i16 noundef [[B:%.*]]) #[[ATTR0]] {
-// CHECK-NEXT:  [[ENTRY:.*:]]
-// CHECK-NEXT:    [[TMP0:%.*]] = insertelement <4 x i16> poison, i16 [[A]], i64 0
-// CHECK-NEXT:    [[TMP1:%.*]] = insertelement <4 x i16> poison, i16 [[B]], i64 0
-// CHECK-NEXT:    [[VQSHLH_S16_I:%.*]] = call <4 x i16> @llvm.aarch64.neon.sqshl.v4i16(<4 x i16> [[TMP0]], <4 x i16> [[TMP1]])
-// CHECK-NEXT:    [[TMP2:%.*]] = extractelement <4 x i16> [[VQSHLH_S16_I]], i64 0
-// CHECK-NEXT:    ret i16 [[TMP2]]
-//
-int16_t test_vqshlh_s16(int16_t a, int16_t b) {
-  return vqshlh_s16(a, b);
-}
-
-// CHECK-LABEL: define dso_local i32 @test_vqshls_s32(
-// CHECK-SAME: i32 noundef [[A:%.*]], i32 noundef [[B:%.*]]) #[[ATTR0]] {
-// CHECK-NEXT:  [[ENTRY:.*:]]
-// CHECK-NEXT:    [[VQSHLS_S32_I:%.*]] = call i32 @llvm.aarch64.neon.sqshl.i32(i32 [[A]], i32 [[B]])
-// CHECK-NEXT:    ret i32 [[VQSHLS_S32_I]]
-//
-int32_t test_vqshls_s32(int32_t a, int32_t b) {
-  return vqshls_s32(a, b);
-}
-
-// CHECK-LABEL: define dso_local i64 @test_vqshld_s64(
-// CHECK-SAME: i64 noundef [[A:%.*]], i64 noundef [[B:%.*]]) #[[ATTR0]] {
-// CHECK-NEXT:  [[ENTRY:.*:]]
-// CHECK-NEXT:    [[VQSHLD_S64_I:%.*]] = call i64 @llvm.aarch64.neon.sqshl.i64(i64 [[A]], i64 [[B]])
-// CHECK-NEXT:    ret i64 [[VQSHLD_S64_I]]
-//
-int64_t test_vqshld_s64(int64_t a, int64_t b) {
-  return vqshld_s64(a, b);
-}
-
-// CHECK-LABEL: define dso_local i8 @test_vqshlb_u8(
-// CHECK-SAME: i8 noundef [[A:%.*]], i8 noundef [[B:%.*]]) #[[ATTR0]] {
-// CHECK-NEXT:  [[ENTRY:.*:]]
-// CHECK-NEXT:    [[TMP0:%.*]] = insertelement <8 x i8> poison, i8 [[A]], i64 0
-// CHECK-NEXT:    [[TMP1:%.*]] = insertelement <8 x i8> poison, i8 [[B]], i64 0
-// CHECK-NEXT:    [[VQSHLB_U8_I:%.*]] = call <8 x i8> @llvm.aarch64.neon.uqshl.v8i8(<8 x i8> [[TMP0]], <8 x i8> [[TMP1]])
-// CHECK-NEXT:    [[TMP2:%.*]] = extractelement <8 x i8> [[VQSHLB_U8_I]], i64 0
-// CHECK-NEXT:    ret i8 [[TMP2]]
-//
-uint8_t test_vqshlb_u8(uint8_t a, int8_t b) {
-  return vqshlb_u8(a, b);
-}
-
-// CHECK-LABEL: define dso_local i16 @test_vqshlh_u16(
-// CHECK-SAME: i16 noundef [[A:%.*]], i16 noundef [[B:%.*]]) #[[ATTR0]] {
-// CHECK-NEXT:  [[ENTRY:.*:]]
-// CHECK-NEXT:    [[TMP0:%.*]] = insertelement <4 x i16> poison, i16 [[A]], i64 0
-// CHECK-NEXT:    [[TMP1:%.*]] = insertelement <4 x i16> poison, i16 [[B]], i64 0
-// CHECK-NEXT:    [[VQSHLH_U16_I:%.*]] = call <4 x i16> @llvm.aarch64.neon.uqshl.v4i16(<4 x i16> [[TMP0]], <4 x i16> [[TMP1]])
-// CHECK-NEXT:    [[TMP2:%.*]] = extractelement <4 x i16> [[VQSHLH_U16_I]], i64 0
-// CHECK-NEXT:    ret i16 [[TMP2]]
-//
-uint16_t test_vqshlh_u16(uint16_t a, int16_t b) {
-  return vqshlh_u16(a, b);
-}
-
-// CHECK-LABEL: define dso_local i32 @test_vqshls_u32(
-// CHECK-SAME: i32 noundef [[A:%.*]], i32 noundef [[B:%.*]]) #[[ATTR0]] {
-// CHECK-NEXT:  [[ENTRY:.*:]]
-// CHECK-NEXT:    [[VQSHLS_U32_I:%.*]] = call i32 @llvm.aarch64.neon.uqshl.i32(i32 [[A]], i32 [[B]])
-// CHECK-NEXT:    ret i32 [[VQSHLS_U32_I]]
-//
-uint32_t test_vqshls_u32(uint32_t a, int32_t b) {
-  return vqshls_u32(a, b);
-}
-
-// CHECK-LABEL: define dso_local i64 @test_vqshld_u64(
-// CHECK-SAME: i64 noundef [[A:%.*]], i64 noundef [[B:%.*]]) #[[ATTR0]] {
-// CHECK-NEXT:  [[ENTRY:.*:]]
-// CHECK-NEXT:    [[VQSHLD_U64_I:%.*]] = call i64 @llvm.aarch64.neon.uqshl.i64(i64 [[A]], i64 [[B]])
-// CHECK-NEXT:    ret i64 [[VQSHLD_U64_I]]
-//
-uint64_t test_vqshld_u64(uint64_t a, int64_t b) {
-  return vqshld_u64(a, b);
-}
-
 // CHECK-LABEL: define dso_local i64 @test_vrshld_s64(
 // CHECK-SAME: i64 noundef [[A:%.*]], i64 noundef [[B:%.*]]) #[[ATTR0]] {
 // CHECK-NEXT:  [[ENTRY:.*:]]
@@ -7347,304 +6941,6 @@ uint64_t test_vcaltd_f64(float64_t a, float64_t b) {
   return (uint64_t)vcaltd_f64(a, b);
 }
 
-// CHECK-LABEL: define dso_local i8 @test_vqshlb_n_s8(
-// CHECK-SAME: i8 noundef [[A:%.*]]) #[[ATTR0]] {
-// CHECK-NEXT:  [[ENTRY:.*:]]
-// CHECK-NEXT:    [[TMP0:%.*]] = insertelement <8 x i8> poison, i8 [[A]], i64 0
-// CHECK-NEXT:    [[VQSHLB_N_S8:%.*]] = call <8 x i8> @llvm.aarch64.neon.sqshl.v8i8(<8 x i8> [[TMP0]], <8 x i8> <i8 7, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison>)
-// CHECK-NEXT:    [[TMP1:%.*]] = extractelement <8 x i8> [[VQSHLB_N_S8]], i64 0
-// CHECK-NEXT:    ret i8 [[TMP1]]
-//
-int8_t test_vqshlb_n_s8(int8_t a) {
-  return (int8_t)vqshlb_n_s8(a, 7);
-}
-
-// CHECK-LABEL: define dso_local i16 @test_vqshlh_n_s16(
-// CHECK-SAME: i16 noundef [[A:%.*]]) #[[ATTR0]] {
-// CHECK-NEXT:  [[ENTRY:.*:]]
-// CHECK-NEXT:    [[TMP0:%.*]] = insertelement <4 x i16> poison, i16 [[A]], i64 0
-// CHECK-NEXT:    [[VQSHLH_N_S16:%.*]] = call <4 x i16> @llvm.aarch64.neon.sqshl.v4i16(<4 x i16> [[TMP0]], <4 x i16> <i16 15, i16 poison, i16 poison, i16 poison>)
-// CHECK-NEXT:    [[TMP1:%.*]] = extractelement <4 x i16> [[VQSHLH_N_S16]], i64 0
-// CHECK-NEXT:    ret i16 [[TMP1]]
-//
-int16_t test_vqshlh_n_s16(int16_t a) {
-  return (int16_t)vqshlh_n_s16(a, 15);
-}
-
-// CHECK-LABEL: define dso_local i32 @test_vqshls_n_s32(
-// CHECK-SAME: i32 noundef [[A:%.*]]) #[[ATTR0]] {
-// CHECK-NEXT:  [[ENTRY:.*:]]
-// CHECK-NEXT:    [[VQSHLS_N_S32:%.*]] = call i32 @llvm.aarch64.neon.sqshl.i32(i32 [[A]], i32 31)
-// CHECK-NEXT:    ret i32 [[VQSHLS_N_S32]]
-//
-int32_t test_vqshls_n_s32(int32_t a) {
-  return (int32_t)vqshls_n_s32(a, 31);
-}
-
-// CHECK-LABEL: define dso_local <8 x i8> @test_vqshl_n_s8(
-// CHECK-SAME: <8 x i8> noundef [[A:%.*]]) #[[ATTR0]] {
-// CHECK-NEXT:  [[ENTRY:.*:]]
-// CHECK-NEXT:    [[VQSHL_N:%.*]] = call <8 x i8> @llvm.aarch64.neon.sqshl.v8i8(<8 x i8> [[A]], <8 x i8> zeroinitializer)
-// CHECK-NEXT:    ret <8 x i8> [[VQSHL_N]]
-//
-int8x8_t test_vqshl_n_s8(int8x8_t a) {
-  return vqshl_n_s8(a, 0);
-}
-
-// CHECK-LABEL: define dso_local <16 x i8> @test_vqshlq_n_s8(
-// CHECK-SAME: <16 x i8> noundef [[A:%.*]]) #[[ATTR0]] {
-// CHECK-NEXT:  [[ENTRY:.*:]]
-// CHECK-NEXT:    [[VQSHL_N:%.*]] = call <16 x i8> @llvm.aarch64.neon.sqshl.v16i8(<16 x i8> [[A]], <16 x i8> zeroinitializer)
-// CHECK-NEXT:    ret <16 x i8> [[VQSHL_N]]
-//
-int8x16_t test_vqshlq_n_s8(int8x16_t a) {
-  return vqshlq_n_s8(a, 0);
-}
-
-// CHECK-LABEL: define dso_local <4 x i16> @test_vqshl_n_s16(
-// CHECK-SAME: <4 x i16> noundef [[A:%.*]]) #[[ATTR0]] {
-// CHECK-NEXT:  [[ENTRY:.*:]]
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i16> [[A]] to <8 x i8>
-// CHECK-NEXT:    [[VQSHL_N:%.*]] = bitcast <8 x i8> [[TMP0]] to <4 x i16>
-// CHECK-NEXT:    [[VQSHL_N1:%.*]] = call <4 x i16> @llvm.aarch64.neon.sqshl.v4i16(<4 x i16> [[VQSHL_N]], <4 x i16> zeroinitializer)
-// CHECK-NEXT:    ret <4 x i16> [[VQSHL_N1]]
-//
-int16x4_t test_vqshl_n_s16(int16x4_t a) {
-  return vqshl_n_s16(a, 0);
-}
-
-// CHECK-LABEL: define dso_local <8 x i16> @test_vqshlq_n_s16(
-// CHECK-SAME: <8 x i16> noundef [[A:%.*]]) #[[ATTR0]] {
-// CHECK-NEXT:  [[ENTRY:.*:]]
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast <8 x i16> [[A]] to <16 x i8>
-// CHECK-NEXT:    [[VQSHL_N:%.*]] = bitcast <16 x i8> [[TMP0]] to <8 x i16>
-// CHECK-NEXT:    [[VQSHL_N1:%.*]] = call <8 x i16> @llvm.aarch64.neon.sqshl.v8i16(<8 x i16> [[VQSHL_N]], <8 x i16> zeroinitializer)
-// CHECK-NEXT:    ret <8 x i16> [[VQSHL_N1]]
-//
-int16x8_t test_vqshlq_n_s16(int16x8_t a) {
-  return vqshlq_n_s16(a, 0);
-}
-
-// CHECK-LABEL: define dso_local <2 x i32> @test_vqshl_n_s32(
-// CHECK-SAME: <2 x i32> noundef [[A:%.*]]) #[[ATTR0]] {
-// CHECK-NEXT:  [[ENTRY:.*:]]
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x i32> [[A]] to <8 x i8>
-// CHECK-NEXT:    [[VQSHL_N:%.*]] = bitcast <8 x i8> [[TMP0]] to <2 x i32>
-// CHECK-NEXT:    [[VQSHL_N1:%.*]] = call <2 x i32> @llvm.aarch64.neon.sqshl.v2i32(<2 x i32> [[VQSHL_N]], <2 x i32> zeroinitializer)
-// CHECK-NEXT:    ret <2 x i32> [[VQSHL_N1]]
-//
-int32x2_t test_vqshl_n_s32(int32x2_t a) {
-  return vqshl_n_s32(a, 0);
-}
-
-// CHECK-LABEL: define dso_local <4 x i32> @test_vqshlq_n_s32(
-// CHECK-SAME: <4 x i32> noundef [[A:%.*]]) #[[ATTR0]] {
-// CHECK-NEXT:  [[ENTRY:.*:]]
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[A]] to <16 x i8>
-// CHECK-NEXT:    [[VQSHL_N:%.*]] = bitcast <16 x i8> [[TMP0]] to <4 x i32>
-// CHECK-NEXT:    [[VQSHL_N1:%.*]] = call <4 x i32> @llvm.aarch64.neon.sqshl.v4i32(<4 x i32> [[VQSHL_N]], <4 x i32> zeroinitializer)
-// CHECK-NEXT:    ret <4 x i32> [[VQSHL_N1]]
-//
-int32x4_t test_vqshlq_n_s32(int32x4_t a) {
-  return vqshlq_n_s32(a, 0);
-}
-
-// CHECK-LABEL: define dso_local <2 x i64> @test_vqshlq_n_s64(
-// CHECK-SAME: <2 x i64> noundef [[A:%.*]]) #[[ATTR0]] {
-// CHECK-NEXT:  [[ENTRY:.*:]]
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x i64> [[A]] to <16 x i8>
-// CHECK-NEXT:    [[VQSHL_N:%.*]] = bitcast <16 x i8> [[TMP0]] to <2 x i64>
-// CHECK-NEXT:    [[VQSHL_N1:%.*]] = call <2 x i64> @llvm.aarch64.neon.sqshl.v2i64(<2 x i64> [[VQSHL_N]], <2 x i64> zeroinitializer)
-// CHECK-NEXT:    ret <2 x i64> [[VQSHL_N1]]
-//
-int64x2_t test_vqshlq_n_s64(int64x2_t a) {
-  return vqshlq_n_s64(a, 0);
-}
-
-// CHECK-LABEL: define dso_local <8 x i8> @test_vqshl_n_u8(
-// CHECK-SAME: <8 x i8> noundef [[A:%.*]]) #[[ATTR0]] {
-// CHECK-NEXT:  [[ENTRY:.*:]]
-// CHECK-NEXT:    [[VQSHL_N:%.*]] = call <8 x i8> @llvm.aarch64.neon.uqshl.v8i8(<8 x i8> [[A]], <8 x i8> zeroinitializer)
-// CHECK-NEXT:    ret <8 x i8> [[VQSHL_N]]
-//
-uint8x8_t test_vqshl_n_u8(uint8x8_t a) {
-  return vqshl_n_u8(a, 0);
-}
-
-// CHECK-LABEL: define dso_local <16 x i8> @test_vqshlq_n_u8(
-// CHECK-SAME: <16 x i8> noundef [[A:%.*]]) #[[ATTR0]] {
-// CHECK-NEXT:  [[ENTRY:.*:]]
-// CHECK-NEXT:    [[VQSHL_N:%.*]] = call <16 x i8> @llvm.aarch64.neon.uqshl.v16i8(<16 x i8> [[A]], <16 x i8> zeroinitializer)
-// CHECK-NEXT:    ret <16 x i8> [[VQSHL_N]]
-//
-uint8x16_t test_vqshlq_n_u8(uint8x16_t a) {
-  return vqshlq_n_u8(a, 0);
-}
-
-// CHECK-LABEL: define dso_local <4 x i16> @test_vqshl_n_u16(
-// CHECK-SAME: <4 x i16> noundef [[A:%.*]]) #[[ATTR0]] {
-// CHECK-NEXT:  [[ENTRY:.*:]]
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i16> [[A]] to <8 x i8>
-// CHECK-NEXT:    [[VQSHL_N:%.*]] = bitcast <8 x i8> [[TMP0]] to <4 x i16>
-// CHECK-NEXT:    [[VQSHL_N1:%.*]] = call <4 x i16> @llvm.aarch64.neon.uqshl.v4i16(<4 x i16> [[VQSHL_N]], <4 x i16> zeroinitializer)
-// CHECK-NEXT:    ret <4 x i16> [[VQSHL_N1]]
-//
-uint16x4_t test_vqshl_n_u16(uint16x4_t a) {
-  return vqshl_n_u16(a, 0);
-}
-
-// CHECK-LABEL: define dso_local <8 x i16> @test_vqshlq_n_u16(
-// CHECK-SAME: <8 x i16> noundef [[A:%.*]]) #[[ATTR0]] {
-// CHECK-NEXT:  [[ENTRY:.*:]]
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast <8 x i16> [[A]] to <16 x i8>
-// CHECK-NEXT:    [[VQSHL_N:%.*]] = bitcast <16 x i8> [[TMP0]] to <8 x i16>
-// CHECK-NEXT:    [[VQSHL_N1:%.*]] = call <8 x i16> @llvm.aarch64.neon.uqshl.v8i16(<8 x i16> [[VQSHL_N]], <8 x i16> zeroinitializer)
-// CHECK-NEXT:    ret <8 x i16> [[VQSHL_N1]]
-//
-uint16x8_t test_vqshlq_n_u16(uint16x8_t a) {
-  return vqshlq_n_u16(a, 0);
-}
-
-// CHECK-LABEL: define dso_local <2 x i32> @test_vqshl_n_u32(
-// CHECK-SAME: <2 x i32> noundef [[A:%.*]]) #[[ATTR0]] {
-// CHECK-NEXT:  [[ENTRY:.*:]]
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x i32> [[A]] to <8 x i8>
-// CHECK-NEXT:    [[VQSHL_N:%.*]] = bitcast <8 x i8> [[TMP0]] to <2 x i32>
-// CHECK-NEXT:    [[VQSHL_N1:%.*]] = call <2 x i32> @llvm.aarch64.neon.uqshl.v2i32(<2 x i32> [[VQSHL_N]], <2 x i32> zeroinitializer)
-// CHECK-NEXT:    ret <2 x i32> [[VQSHL_N1]]
-//
-uint32x2_t test_vqshl_n_u32(uint32x2_t a) {
-  return vqshl_n_u32(a, 0);
-}
-
-// CHECK-LABEL: define dso_local <4 x i32> @test_vqshlq_n_u32(
-// CHECK-SAME: <4 x i32> noundef [[A:%.*]]) #[[ATTR0]] {
-// CHECK-NEXT:  [[ENTRY:.*:]]
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast <4 x i32> [[A]] to <16 x i8>
-// CHECK-NEXT:    [[VQSHL_N:%.*]] = bitcast <16 x i8> [[TMP0]] to <4 x i32>
-// CHECK-NEXT:    [[VQSHL_N1:%.*]] = call <4 x i32> @llvm.aarch64.neon.uqshl.v4i32(<4 x i32> [[VQSHL_N]], <4 x i32> zeroinitializer)
-// CHECK-NEXT:    ret <4 x i32> [[VQSHL_N1]]
-//
-uint32x4_t test_vqshlq_n_u32(uint32x4_t a) {
-  return vqshlq_n_u32(a, 0);
-}
-
-// CHECK-LABEL: define dso_local <2 x i64> @test_vqshlq_n_u64(
-// CHECK-SAME: <2 x i64> noundef [[A:%.*]]) #[[ATTR0]] {
-// CHECK-NEXT:  [[ENTRY:.*:]]
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast <2 x i64> [[A]] to <16 x i8>
-// CHECK-NEXT:    [[VQSHL_N:%.*]] = bitcast <16 x i8> [[TMP0]] to <2 x i64>
-// CHECK-NEXT:    [[VQSHL_N1:%.*]] = call <2 x i64> @llvm.aarch64.neon.uqshl.v2i64(<2 x i64> [[VQSHL_N]], <2 x i64> zeroinitializer)
-// CHECK-NEXT:    ret <2 x i64> [[VQSHL_N1]]
-//
-uint64x2_t test_vqshlq_n_u64(uint64x2_t a) {
-  return vqshlq_n_u64(a, 0);
-}
-
-// CHECK-LABEL: define dso_local <1 x i64> @test_vqshl_n_s64(
-// CHECK-SAME: <1 x i64> noundef [[A:%.*]]) #[[ATTR0]] {
-// CHECK-NEXT:  [[ENTRY:.*:]]
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast <1 x i64> [[A]] to <8 x i8>
-// CHECK-NEXT:    [[VQSHL_N:%.*]] = bitcast <8 x i8> [[TMP0]] to <1 x i64>
-// CHECK-NEXT:    [[VQSHL_N1:%.*]] = call <1 x i64> @llvm.aarch64.neon.sqshl.v1i64(<1 x i64> [[VQSHL_N]], <1 x i64> splat (i64 1))
-// CHECK-NEXT:    ret <1 x i64> [[VQSHL_N1]]
-//
-int64x1_t test_vqshl_n_s64(int64x1_t a) {
-  return vqshl_n_s64(a, 1);
-}
-
-// CHECK-LABEL: define dso_local i8 @test_vqshlb_n_u8(
-// CHECK-SAME: i8 noundef [[A:%.*]]) #[[ATTR0]] {
-// CHECK-NEXT:  [[ENTRY:.*:]]
-// CHECK-NEXT:    [[TMP0:%.*]] = insertelement <8 x i8> poison, i8 [[A]], i64 0
-// CHECK-NEXT:    [[VQSHLB_N_U8:%.*]] = call <8 x i8> @llvm.aarch64.neon.uqshl.v8i8(<8 x i8> [[TMP0]], <8 x i8> <i8 7, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison>)
-// CHECK-NEXT:    [[TMP1:%.*]] = extractelement <8 x i8> [[VQSHLB_N_U8]], i64 0
-// CHECK-NEXT:    ret i8 [[TMP1]]
-//
-uint8_t test_vqshlb_n_u8(uint8_t a) {
-  return (uint8_t)vqshlb_n_u8(a, 7);
-}
-
-// CHECK-LABEL: define dso_local i16 @test_vqshlh_n_u16(
-// CHECK-SAME: i16 noundef [[A:%.*]]) #[[ATTR0]] {
-// CHECK-NEXT:  [[ENTRY:.*:]]
-// CHECK-NEXT:    [[TMP0:%.*]] = insertelement <4 x i16> poison, i16 [[A]], i64 0
-// CHECK-NEXT:    [[VQSHLH_N_U16:%.*]] = call <4 x i16> @llvm.aarch64.neon.uqshl.v4i16(<4 x i16> [[TMP0]], <4 x i16> <i16 15, i16 poison, i16 poison, i16 poison>)
-// CHECK-NEXT:    [[TMP1:%.*]] = extractelement <4 x i16> [[VQSHLH_N_U16]], i64 0
-// CHECK-NEXT:    ret i16 [[TMP1]]
-//
-uint16_t test_vqshlh_n_u16(uint16_t a) {
-  return (uint16_t)vqshlh_n_u16(a, 15);
-}
-
-// CHECK-LABEL: define dso_local i32 @test_vqshls_n_u32(
-// CHECK-SAME: i32 noundef [[A:%.*]]) #[[ATTR0]] {
-// CHECK-NEXT:  [[ENTRY:.*:]]
-// CHECK-NEXT:    [[VQSHLS_N_U32:%.*]] = call i32 @llvm.aarch64.neon.uqshl.i32(i32 [[A]], i32 31)
-// CHECK-NEXT:    ret i32 [[VQSHLS_N_U32]]
-//
-uint32_t test_vqshls_n_u32(uint32_t a) {
-  return (uint32_t)vqshls_n_u32(a, 31);
-}
-
-// CHECK-LABEL: define dso_local <1 x i64> @test_vqshl_n_u64(
-// CHECK-SAME: <1 x i64> noundef [[A:%.*]]) #[[ATTR0]] {
-// CHECK-NEXT:  [[ENTRY:.*:]]
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast <1 x i64> [[A]] to <8 x i8>
-// CHECK-NEXT:    [[VQSHL_N:%.*]] = bitcast <8 x i8> [[TMP0]] to <1 x i64>
-// CHECK-NEXT:    [[VQSHL_N1:%.*]] = call <1 x i64> @llvm.aarch64.neon.uqshl.v1i64(<1 x i64> [[VQSHL_N]], <1 x i64> splat (i64 1))
-// CHECK-NEXT:    ret <1 x i64> [[VQSHL_N1]]
-//
-uint64x1_t test_vqshl_n_u64(uint64x1_t a) {
-  return vqshl_n_u64(a, 1);
-}
-
-// CHECK-LABEL: define dso_local i8 @test_vqshlub_n_s8(
-// CHECK-SAME: i8 noundef [[A:%.*]]) #[[ATTR0]] {
-// CHECK-NEXT:  [[ENTRY:.*:]]
-// CHECK-NEXT:    [[TMP0:%.*]] = insertelement <8 x i8> poison, i8 [[A]], i64 0
-// CHECK-NEXT:    [[VQSHLUB_N_S8:%.*]] = call <8 x i8> @llvm.aarch64.neon.sqshlu.v8i8(<8 x i8> [[TMP0]], <8 x i8> <i8 7, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison>)
-// CHECK-NEXT:    [[TMP1:%.*]] = extractelement <8 x i8> [[VQSHLUB_N_S8]], i64 0
-// CHECK-NEXT:    ret i8 [[TMP1]]
-//
-int8_t test_vqshlub_n_s8(int8_t a) {
-  return (int8_t)vqshlub_n_s8(a, 7);
-}
-
-// CHECK-LABEL: define dso_local i16 @test_vqshluh_n_s16(
-// CHECK-SAME: i16 noundef [[A:%.*]]) #[[ATTR0]] {
-// CHECK-NEXT:  [[ENTRY:.*:]]
-// CHECK-NEXT:    [[TMP0:%.*]] = insertelement <4 x i16> poison, i16 [[A]], i64 0
-// CHECK-NEXT:    [[VQSHLUH_N_S16:%.*]] = call <4 x i16> @llvm.aarch64.neon.sqshlu.v4i16(<4 x i16> [[TMP0]], <4 x i16> <i16 15, i16 poison, i16 poison, i16 poison>)
-// CHECK-NEXT:    [[TMP1:%.*]] = extractelement <4 x i16> [[VQSHLUH_N_S16]], i64 0
-// CHECK-NEXT:    ret i16 [[TMP1]]
-//
-int16_t test_vqshluh_n_s16(int16_t a) {
-  return (int16_t)vqshluh_n_s16(a, 15);
-}
-
-// CHECK-LABEL: define dso_local i32 @test_vqshlus_n_s32(
-// CHECK-SAME: i32 noundef [[A:%.*]]) #[[ATTR0]] {
-// CHECK-NEXT:  [[ENTRY:.*:]]
-// CHECK-NEXT:    [[VQSHLUS_N_S32:%.*]] = call i32 @llvm.aarch64.neon.sqshlu.i32(i32 [[A]], i32 31)
-// CHECK-NEXT:    ret i32 [[VQSHLUS_N_S32]]
-//
-int32_t test_vqshlus_n_s32(int32_t a) {
-  return (int32_t)vqshlus_n_s32(a, 31);
-}
-
-// CHECK-LABEL: define dso_local <1 x i64> @test_vqshlu_n_s64(
-// CHECK-SAME: <1 x i64> noundef [[A:%.*]]) #[[ATTR0]] {
-// CHECK-NEXT:  [[ENTRY:.*:]]
-// CHECK-NEXT:    [[TMP0:%.*]] = bitcast <1 x i64> [[A]] to <8 x i8>
-// CHECK-NEXT:    [[VQSHLU_N:%.*]] = bitcast <8 x i8> [[TMP0]] to <1 x i64>
-// CHECK-NEXT:    [[VQSHLU_N1:%.*]] = call <1 x i64> @llvm.aarch64.neon.sqshlu.v1i64(<1 x i64> [[VQSHLU_N]], <1 x i64> splat (i64 1))
-// CHECK-NEXT:    ret <1 x i64> [[VQSHLU_N1]]
-//
-uint64x1_t test_vqshlu_n_s64(int64x1_t a) {
-  return vqshlu_n_s64(a, 1);
-}
-
 
 
 // CHECK-LABEL: define dso_local <8 x i8> @test_vreinterpret_s8_s16(
diff --git a/clang/test/CodeGen/AArch64/neon/intrinsics.c b/clang/test/CodeGen/AArch64/neon/intrinsics.c
index 6d46e7ad8a8f3c..7175c616cf484f 100644
--- a/clang/test/CodeGen/AArch64/neon/intrinsics.c
+++ b/clang/test/CodeGen/AArch64/neon/intrinsics.c
@@ -561,9 +561,8 @@ float64_t test_vmaxvq_f64(float64x2_t a) {
 }
 
 //===------------------------------------------------------===//
-// 2.1.3.2 Vector Saturating Shift Left
-//
-// TODO: Implement the remaining intrinsics from this group.
+// 2.1.3.1.2 Vector Saturating Shift Left
+// https://arm-software.github.io/acle/neon_intrinsics/advsimd.html#vector-saturating-shift-left
 //===------------------------------------------------------===//
 
 // ALL-LABEL: @test_vqshlud_n_s64(
@@ -600,6 +599,797 @@ int64_t test_vqshld_n_s64(int64_t a) {
   return (int64_t)vqshld_n_s64(a, 63);
 }
 
+// LLVM-LABEL: @test_vqshl_s8(
+// CIR-LABEL: @vqshl_s8(
+int8x8_t test_vqshl_s8(int8x8_t a, int8x8_t b) {
+// CIR: {{%.*}} = cir.call_llvm_intrinsic "aarch64.neon.sqshl" {{%.*}}, {{%.*}} : (!cir.vector<8 x !s8i>, !cir.vector<8 x !s8i>) -> !cir.vector<8 x !s8i>
+
+// LLVM-SAME: <8 x i8> {{.*}}[[A:%.*]], <8 x i8> {{.*}}[[B:%.*]])
+// LLVM: [[VQSHL_V_I:%.*]] = call <8 x i8> @llvm.aarch64.neon.sqshl.v8i8(<8 x i8> [[A]], <8 x i8> [[B]])
+// LLVM: ret <8 x i8> [[VQSHL_V_I]]
+  return vqshl_s8(a, b);
+}
+
+// LLVM-LABEL: @test_vqshl_s16(
+// CIR-LABEL: @vqshl_s16(
+int16x4_t test_vqshl_s16(int16x4_t a, int16x4_t b) {
+// CIR: {{%.*}} = cir.call_llvm_intrinsic "aarch64.neon.sqshl" {{%.*}}, {{%.*}} : (!cir.vector<4 x !s16i>, !cir.vector<4 x !s16i>) -> !cir.vector<4 x !s16i>
+
+// LLVM-SAME: <4 x i16> {{.*}}[[A:%.*]], <4 x i16> {{.*}}[[B:%.*]])
+// LLVM: [[TMP0:%.*]] = bitcast <4 x i16> [[A]] to <8 x i8>
+// LLVM: [[TMP1:%.*]] = bitcast <4 x i16> [[B]] to <8 x i8>
+// LLVM: [[VQSHL_V_I:%.*]] = bitcast <8 x i8> [[TMP0]] to <4 x i16>
+// LLVM: [[VQSHL_V1_I:%.*]] = bitcast <8 x i8> [[TMP1]] to <4 x i16>
+// LLVM: [[VQSHL_V2_I:%.*]] = call <4 x i16> @llvm.aarch64.neon.sqshl.v4i16(<4 x i16> [[VQSHL_V_I]], <4 x i16> [[VQSHL_V1_I]])
+// LLVM: [[VQSHL_V3_I:%.*]] = bitcast <4 x i16> [[VQSHL_V2_I]] to <8 x i8>
+// LLVM: [[TMP2:%.*]] = bitcast <8 x i8> [[VQSHL_V3_I]] to <4 x i16>
+// LLVM: ret <4 x i16> [[TMP2]]
+  return vqshl_s16(a, b);
+}
+
+// LLVM-LABEL: @test_vqshl_s32(
+// CIR-LABEL: @vqshl_s32(
+int32x2_t test_vqshl_s32(int32x2_t a, int32x2_t b) {
+// CIR: {{%.*}} = cir.call_llvm_intrinsic "aarch64.neon.sqshl" {{%.*}}, {{%.*}} : (!cir.vector<2 x !s32i>, !cir.vector<2 x !s32i>) -> !cir.vector<2 x !s32i>
+
+// LLVM-SAME: <2 x i32> {{.*}}[[A:%.*]], <2 x i32> {{.*}}[[B:%.*]])
+// LLVM: [[TMP0:%.*]] = bitcast <2 x i32> [[A]] to <8 x i8>
+// LLVM: [[TMP1:%.*]] = bitcast <2 x i32> [[B]] to <8 x i8>
+// LLVM: [[VQSHL_V_I:%.*]] = bitcast <8 x i8> [[TMP0]] to <2 x i32>
+// LLVM: [[VQSHL_V1_I:%.*]] = bitcast <8 x i8> [[TMP1]] to <2 x i32>
+// LLVM: [[VQSHL_V2_I:%.*]] = call <2 x i32> @llvm.aarch64.neon.sqshl.v2i32(<2 x i32> [[VQSHL_V_I]], <2 x i32> [[VQSHL_V1_I]])
+// LLVM: [[VQSHL_V3_I:%.*]] = bitcast <2 x i32> [[VQSHL_V2_I]] to <8 x i8>
+// LLVM: [[TMP2:%.*]] = bitcast <8 x i8> [[VQSHL_V3_I]] to <2 x i32>
+// LLVM: ret <2 x i32> [[TMP2]]
+  return vqshl_s32(a, b);
+}
+
+// LLVM-LABEL: @test_vqshl_s64(
+// CIR-LABEL: @vqshl_s64(
+int64x1_t test_vqshl_s64(int64x1_t a, int64x1_t b) {
+// CIR: {{%.*}} = cir.call_llvm_intrinsic "aarch64.neon.sqshl" {{%.*}}, {{%.*}} : (!cir.vector<1 x !s64i>, !cir.vector<1 x !s64i>) -> !cir.vector<1 x !s64i>
+
+// LLVM-SAME: <1 x i64> {{.*}}[[A:%.*]], <1 x i64> {{.*}}[[B:%.*]])
+// LLVM: [[TMP0:%.*]] = bitcast <1 x i64> [[A]] to <8 x i8>
+// LLVM: [[TMP1:%.*]] = bitcast <1 x i64> [[B]] to <8 x i8>
+// LLVM: [[VQSHL_V_I:%.*]] = bitcast <8 x i8> [[TMP0]] to <1 x i64>
+// LLVM: [[VQSHL_V1_I:%.*]] = bitcast <8 x i8> [[TMP1]] to <1 x i64>
+// LLVM: [[VQSHL_V2_I:%.*]] = call <1 x i64> @llvm.aarch64.neon.sqshl.v1i64(<1 x i64> [[VQSHL_V_I]], <1 x i64> [[VQSHL_V1_I]])
+// LLVM: [[VQSHL_V3_I:%.*]] = bitcast <1 x i64> [[VQSHL_V2_I]] to <8 x i8>
+// LLVM: [[TMP2:%.*]] = bitcast <8 x i8> [[VQSHL_V3_I]] to i64
+// LLVM: [[REF_TMP_I_SROA_0_0_VEC_INSERT:%.*]] = insertelement <1 x i64> undef, i64 [[TMP2]], i64 0
+// LLVM: ret <1 x i64> [[REF_TMP_I_SROA_0_0_VEC_INSERT]]
+  return vqshl_s64(a, b);
+}
+
+// LLVM-LABEL: @test_vqshl_u8(
+// CIR-LABEL: @vqshl_u8(
+uint8x8_t test_vqshl_u8(uint8x8_t a, int8x8_t b) {
+// CIR: {{%.*}} = cir.call_llvm_intrinsic "aarch64.neon.uqshl" {{%.*}}, {{%.*}} : (!cir.vector<8 x !u8i>, !cir.vector<8 x !u8i>) -> !cir.vector<8 x !u8i>
+
+// LLVM-SAME: <8 x i8> {{.*}}[[A:%.*]], <8 x i8> {{.*}}[[B:%.*]])
+// LLVM: [[VQSHL_V_I:%.*]] = call <8 x i8> @llvm.aarch64.neon.uqshl.v8i8(<8 x i8> [[A]], <8 x i8> [[B]])
+// LLVM: ret <8 x i8> [[VQSHL_V_I]]
+  return vqshl_u8(a, b);
+}
+
+// LLVM-LABEL: @test_vqshl_u16(
+// CIR-LABEL: @vqshl_u16(
+uint16x4_t test_vqshl_u16(uint16x4_t a, int16x4_t b) {
+// CIR: {{%.*}} = cir.call_llvm_intrinsic "aarch64.neon.uqshl" {{%.*}}, {{%.*}} : (!cir.vector<4 x !u16i>, !cir.vector<4 x !u16i>) -> !cir.vector<4 x !u16i>
+
+// LLVM-SAME: <4 x i16> {{.*}}[[A:%.*]], <4 x i16> {{.*}}[[B:%.*]])
+// LLVM: [[TMP0:%.*]] = bitcast <4 x i16> [[A]] to <8 x i8>
+// LLVM: [[TMP1:%.*]] = bitcast <4 x i16> [[B]] to <8 x i8>
+// LLVM: [[VQSHL_V_I:%.*]] = bitcast <8 x i8> [[TMP0]] to <4 x i16>
+// LLVM: [[VQSHL_V1_I:%.*]] = bitcast <8 x i8> [[TMP1]] to <4 x i16>
+// LLVM: [[VQSHL_V2_I:%.*]] = call <4 x i16> @llvm.aarch64.neon.uqshl.v4i16(<4 x i16> [[VQSHL_V_I]], <4 x i16> [[VQSHL_V1_I]])
+// LLVM: [[VQSHL_V3_I:%.*]] = bitcast <4 x i16> [[VQSHL_V2_I]] to <8 x i8>
+// LLVM: [[TMP2:%.*]] = bitcast <8 x i8> [[VQSHL_V3_I]] to <4 x i16>
+// LLVM: ret <4 x i16> [[TMP2]]
+  return vqshl_u16(a, b);
+}
+
+// LLVM-LABEL: @test_vqshl_u32(
+// CIR-LABEL: @vqshl_u32(
+uint32x2_t test_vqshl_u32(uint32x2_t a, int32x2_t b) {
+// CIR: {{%.*}} = cir.call_llvm_intrinsic "aarch64.neon.uqshl" {{%.*}}, {{%.*}} : (!cir.vector<2 x !u32i>, !cir.vector<2 x !u32i>) -> !cir.vector<2 x !u32i>
+
+// LLVM-SAME: <2 x i32> {{.*}}[[A:%.*]], <2 x i32> {{.*}}[[B:%.*]])
+// LLVM: [[TMP0:%.*]] = bitcast <2 x i32> [[A]] to <8 x i8>
+// LLVM: [[TMP1:%.*]] = bitcast <2 x i32> [[B]] to <8 x i8>
+// LLVM: [[VQSHL_V_I:%.*]] = bitcast <8 x i8> [[TMP0]] to <2 x i32>
+// LLVM: [[VQSHL_V1_I:%.*]] = bitcast <8 x i8> [[TMP1]] to <2 x i32>
+// LLVM: [[VQSHL_V2_I:%.*]] = call <2 x i32> @llvm.aarch64.neon.uqshl.v2i32(<2 x i32> [[VQSHL_V_I]], <2 x i32> [[VQSHL_V1_I]])
+// LLVM: [[VQSHL_V3_I:%.*]] = bitcast <2 x i32> [[VQSHL_V2_I]] to <8 x i8>
+// LLVM: [[TMP2:%.*]] = bitcast <8 x i8> [[VQSHL_V3_I]] to <2 x i32>
+// LLVM: ret <2 x i32> [[TMP2]]
+  return vqshl_u32(a, b);
+}
+
+// LLVM-LABEL: @test_vqshl_u64(
+// CIR-LABEL: @vqshl_u64(
+uint64x1_t test_vqshl_u64(uint64x1_t a, int64x1_t b) {
+// CIR: {{%.*}} = cir.call_llvm_intrinsic "aarch64.neon.uqshl" {{%.*}}, {{%.*}} : (!cir.vector<1 x !u64i>, !cir.vector<1 x !u64i>) -> !cir.vector<1 x !u64i>
+
+// LLVM-SAME: <1 x i64> {{.*}}[[A:%.*]], <1 x i64> {{.*}}[[B:%.*]])
+// LLVM: [[TMP0:%.*]] = bitcast <1 x i64> [[A]] to <8 x i8>
+// LLVM: [[TMP1:%.*]] = bitcast <1 x i64> [[B]] to <8 x i8>
+// LLVM: [[VQSHL_V_I:%.*]] = bitcast <8 x i8> [[TMP0]] to <1 x i64>
+// LLVM: [[VQSHL_V1_I:%.*]] = bitcast <8 x i8> [[TMP1]] to <1 x i64>
+// LLVM: [[VQSHL_V2_I:%.*]] = call <1 x i64> @llvm.aarch64.neon.uqshl.v1i64(<1 x i64> [[VQSHL_V_I]], <1 x i64> [[VQSHL_V1_I]])
+// LLVM: [[VQSHL_V3_I:%.*]] = bitcast <1 x i64> [[VQSHL_V2_I]] to <8 x i8>
+// LLVM: [[TMP2:%.*]] = bitcast <8 x i8> [[VQSHL_V3_I]] to i64
+// LLVM: [[REF_TMP_I_SROA_0_0_VEC_INSERT:%.*]] = insertelement <1 x i64> undef, i64 [[TMP2]], i64 0
+// LLVM: ret <1 x i64> [[REF_TMP_I_SROA_0_0_VEC_INSERT]]
+  return vqshl_u64(a, b);
+}
+
+// LLVM-LABEL: @test_vqshlq_s8(
+// CIR-LABEL: @vqshlq_s8(
+int8x16_t test_vqshlq_s8(int8x16_t a, int8x16_t b) {
+// CIR: {{%.*}} = cir.call_llvm_intrinsic "aarch64.neon.sqshl" {{%.*}}, {{%.*}} : (!cir.vector<16 x !s8i>, !cir.vector<16 x !s8i>) -> !cir.vector<16 x !s8i>
+
+// LLVM-SAME: <16 x i8> {{.*}}[[A:%.*]], <16 x i8> {{.*}}[[B:%.*]])
+// LLVM: [[VQSHLQ_V_I:%.*]] = call <16 x i8> @llvm.aarch64.neon.sqshl.v16i8(<16 x i8> [[A]], <16 x i8> [[B]])
+// LLVM: ret <16 x i8> [[VQSHLQ_V_I]]
+  return vqshlq_s8(a, b);
+}
+
+// LLVM-LABEL: @test_vqshlq_s16(
+// CIR-LABEL: @vqshlq_s16(
+int16x8_t test_vqshlq_s16(int16x8_t a, int16x8_t b) {
+// CIR: {{%.*}} = cir.call_llvm_intrinsic "aarch64.neon.sqshl" {{%.*}}, {{%.*}} : (!cir.vector<8 x !s16i>, !cir.vector<8 x !s16i>) -> !cir.vector<8 x !s16i>
+
+// LLVM-SAME: <8 x i16> {{.*}}[[A:%.*]], <8 x i16> {{.*}}[[B:%.*]])
+// LLVM: [[TMP0:%.*]] = bitcast <8 x i16> [[A]] to <16 x i8>
+// LLVM: [[TMP1:%.*]] = bitcast <8 x i16> [[B]] to <16 x i8>
+// LLVM: [[VQSHLQ_V_I:%.*]] = bitcast <16 x i8> [[TMP0]] to <8 x i16>
+// LLVM: [[VQSHLQ_V1_I:%.*]] = bitcast <16 x i8> [[TMP1]] to <8 x i16>
+// LLVM: [[VQSHLQ_V2_I:%.*]] = call <8 x i16> @llvm.aarch64.neon.sqshl.v8i16(<8 x i16> [[VQSHLQ_V_I]], <8 x i16> [[VQSHLQ_V1_I]])
+// LLVM: [[VQSHLQ_V3_I:%.*]] = bitcast <8 x i16> [[VQSHLQ_V2_I]] to <16 x i8>
+// LLVM: [[TMP2:%.*]] = bitcast <16 x i8> [[VQSHLQ_V3_I]] to <8 x i16>
+// LLVM: ret <8 x i16> [[TMP2]]
+  return vqshlq_s16(a, b);
+}
+
+// LLVM-LABEL: @test_vqshlq_s32(
+// CIR-LABEL: @vqshlq_s32(
+int32x4_t test_vqshlq_s32(int32x4_t a, int32x4_t b) {
+// CIR: {{%.*}} = cir.call_llvm_intrinsic "aarch64.neon.sqshl" {{%.*}}, {{%.*}} : (!cir.vector<4 x !s32i>, !cir.vector<4 x !s32i>) -> !cir.vector<4 x !s32i>
+
+// LLVM-SAME: <4 x i32> {{.*}}[[A:%.*]], <4 x i32> {{.*}}[[B:%.*]])
+// LLVM: [[TMP0:%.*]] = bitcast <4 x i32> [[A]] to <16 x i8>
+// LLVM: [[TMP1:%.*]] = bitcast <4 x i32> [[B]] to <16 x i8>
+// LLVM: [[VQSHLQ_V_I:%.*]] = bitcast <16 x i8> [[TMP0]] to <4 x i32>
+// LLVM: [[VQSHLQ_V1_I:%.*]] = bitcast <16 x i8> [[TMP1]] to <4 x i32>
+// LLVM: [[VQSHLQ_V2_I:%.*]] = call <4 x i32> @llvm.aarch64.neon.sqshl.v4i32(<4 x i32> [[VQSHLQ_V_I]], <4 x i32> [[VQSHLQ_V1_I]])
+// LLVM: [[VQSHLQ_V3_I:%.*]] = bitcast <4 x i32> [[VQSHLQ_V2_I]] to <16 x i8>
+// LLVM: [[TMP2:%.*]] = bitcast <16 x i8> [[VQSHLQ_V3_I]] to <4 x i32>
+// LLVM: ret <4 x i32> [[TMP2]]
+  return vqshlq_s32(a, b);
+}
+
+// LLVM-LABEL: @test_vqshlq_s64(
+// CIR-LABEL: @vqshlq_s64(
+int64x2_t test_vqshlq_s64(int64x2_t a, int64x2_t b) {
+// CIR: {{%.*}} = cir.call_llvm_intrinsic "aarch64.neon.sqshl" {{%.*}}, {{%.*}} : (!cir.vector<2 x !s64i>, !cir.vector<2 x !s64i>) -> !cir.vector<2 x !s64i>
+
+// LLVM-SAME: <2 x i64> {{.*}}[[A:%.*]], <2 x i64> {{.*}}[[B:%.*]])
+// LLVM: [[TMP0:%.*]] = bitcast <2 x i64> [[A]] to <16 x i8>
+// LLVM: [[TMP1:%.*]] = bitcast <2 x i64> [[B]] to <16 x i8>
+// LLVM: [[VQSHLQ_V_I:%.*]] = bitcast <16 x i8> [[TMP0]] to <2 x i64>
+// LLVM: [[VQSHLQ_V1_I:%.*]] = bitcast <16 x i8> [[TMP1]] to <2 x i64>
+// LLVM: [[VQSHLQ_V2_I:%.*]] = call <2 x i64> @llvm.aarch64.neon.sqshl.v2i64(<2 x i64> [[VQSHLQ_V_I]], <2 x i64> [[VQSHLQ_V1_I]])
+// LLVM: [[VQSHLQ_V3_I:%.*]] = bitcast <2 x i64> [[VQSHLQ_V2_I]] to <16 x i8>
+// LLVM: [[TMP2:%.*]] = bitcast <16 x i8> [[VQSHLQ_V3_I]] to <2 x i64>
+// LLVM: ret <2 x i64> [[TMP2]]
+  return vqshlq_s64(a, b);
+}
+
+// LLVM-LABEL: @test_vqshlq_u8(
+// CIR-LABEL: @vqshlq_u8(
+uint8x16_t test_vqshlq_u8(uint8x16_t a, int8x16_t b) {
+// CIR: {{%.*}} = cir.call_llvm_intrinsic "aarch64.neon.uqshl" {{%.*}}, {{%.*}} : (!cir.vector<16 x !u8i>, !cir.vector<16 x !u8i>) -> !cir.vector<16 x !u8i>
+
+// LLVM-SAME: <16 x i8> {{.*}}[[A:%.*]], <16 x i8> {{.*}}[[B:%.*]])
+// LLVM: [[VQSHLQ_V_I:%.*]] = call <16 x i8> @llvm.aarch64.neon.uqshl.v16i8(<16 x i8> [[A]], <16 x i8> [[B]])
+// LLVM: ret <16 x i8> [[VQSHLQ_V_I]]
+  return vqshlq_u8(a, b);
+}
+
+// LLVM-LABEL: @test_vqshlq_u16(
+// CIR-LABEL: @vqshlq_u16(
+uint16x8_t test_vqshlq_u16(uint16x8_t a, int16x8_t b) {
+// CIR: {{%.*}} = cir.call_llvm_intrinsic "aarch64.neon.uqshl" {{%.*}}, {{%.*}} : (!cir.vector<8 x !u16i>, !cir.vector<8 x !u16i>) -> !cir.vector<8 x !u16i>
+
+// LLVM-SAME: <8 x i16> {{.*}}[[A:%.*]], <8 x i16> {{.*}}[[B:%.*]])
+// LLVM: [[TMP0:%.*]] = bitcast <8 x i16> [[A]] to <16 x i8>
+// LLVM: [[TMP1:%.*]] = bitcast <8 x i16> [[B]] to <16 x i8>
+// LLVM: [[VQSHLQ_V_I:%.*]] = bitcast <16 x i8> [[TMP0]] to <8 x i16>
+// LLVM: [[VQSHLQ_V1_I:%.*]] = bitcast <16 x i8> [[TMP1]] to <8 x i16>
+// LLVM: [[VQSHLQ_V2_I:%.*]] = call <8 x i16> @llvm.aarch64.neon.uqshl.v8i16(<8 x i16> [[VQSHLQ_V_I]], <8 x i16> [[VQSHLQ_V1_I]])
+// LLVM: [[VQSHLQ_V3_I:%.*]] = bitcast <8 x i16> [[VQSHLQ_V2_I]] to <16 x i8>
+// LLVM: [[TMP2:%.*]] = bitcast <16 x i8> [[VQSHLQ_V3_I]] to <8 x i16>
+// LLVM: ret <8 x i16> [[TMP2]]
+  return vqshlq_u16(a, b);
+}
+
+// LLVM-LABEL: @test_vqshlq_u32(
+// CIR-LABEL: @vqshlq_u32(
+uint32x4_t test_vqshlq_u32(uint32x4_t a, int32x4_t b) {
+// CIR: {{%.*}} = cir.call_llvm_intrinsic "aarch64.neon.uqshl" {{%.*}}, {{%.*}} : (!cir.vector<4 x !u32i>, !cir.vector<4 x !u32i>) -> !cir.vector<4 x !u32i>
+
+// LLVM-SAME: <4 x i32> {{.*}}[[A:%.*]], <4 x i32> {{.*}}[[B:%.*]])
+// LLVM: [[TMP0:%.*]] = bitcast <4 x i32> [[A]] to <16 x i8>
+// LLVM: [[TMP1:%.*]] = bitcast <4 x i32> [[B]] to <16 x i8>
+// LLVM: [[VQSHLQ_V_I:%.*]] = bitcast <16 x i8> [[TMP0]] to <4 x i32>
+// LLVM: [[VQSHLQ_V1_I:%.*]] = bitcast <16 x i8> [[TMP1]] to <4 x i32>
+// LLVM: [[VQSHLQ_V2_I:%.*]] = call <4 x i32> @llvm.aarch64.neon.uqshl.v4i32(<4 x i32> [[VQSHLQ_V_I]], <4 x i32> [[VQSHLQ_V1_I]])
+// LLVM: [[VQSHLQ_V3_I:%.*]] = bitcast <4 x i32> [[VQSHLQ_V2_I]] to <16 x i8>
+// LLVM: [[TMP2:%.*]] = bitcast <16 x i8> [[VQSHLQ_V3_I]] to <4 x i32>
+// LLVM: ret <4 x i32> [[TMP2]]
+  return vqshlq_u32(a, b);
+}
+
+// LLVM-LABEL: @test_vqshlq_u64(
+// CIR-LABEL: @vqshlq_u64(
+uint64x2_t test_vqshlq_u64(uint64x2_t a, int64x2_t b) {
+// CIR: {{%.*}} = cir.call_llvm_intrinsic "aarch64.neon.uqshl" {{%.*}}, {{%.*}} : (!cir.vector<2 x !u64i>, !cir.vector<2 x !u64i>) -> !cir.vector<2 x !u64i>
+
+// LLVM-SAME: <2 x i64> {{.*}}[[A:%.*]], <2 x i64> {{.*}}[[B:%.*]])
+// LLVM: [[TMP0:%.*]] = bitcast <2 x i64> [[A]] to <16 x i8>
+// LLVM: [[TMP1:%.*]] = bitcast <2 x i64> [[B]] to <16 x i8>
+// LLVM: [[VQSHLQ_V_I:%.*]] = bitcast <16 x i8> [[TMP0]] to <2 x i64>
+// LLVM: [[VQSHLQ_V1_I:%.*]] = bitcast <16 x i8> [[TMP1]] to <2 x i64>
+// LLVM: [[VQSHLQ_V2_I:%.*]] = call <2 x i64> @llvm.aarch64.neon.uqshl.v2i64(<2 x i64> [[VQSHLQ_V_I]], <2 x i64> [[VQSHLQ_V1_I]])
+// LLVM: [[VQSHLQ_V3_I:%.*]] = bitcast <2 x i64> [[VQSHLQ_V2_I]] to <16 x i8>
+// LLVM: [[TMP2:%.*]] = bitcast <16 x i8> [[VQSHLQ_V3_I]] to <2 x i64>
+// LLVM: ret <2 x i64> [[TMP2]]
+  return vqshlq_u64(a, b);
+}
+
+// LLVM-LABEL: @test_vqshlb_s8(
+// CIR-LABEL: @vqshlb_s8(
+int8_t test_vqshlb_s8(int8_t a, int8_t b) {
+// CIR: {{%.*}} = cir.call_llvm_intrinsic "aarch64.neon.sqshl" {{%.*}}, {{%.*}} : (!cir.vector<8 x !s8i>, !cir.vector<8 x !s8i>) -> !cir.vector<8 x !s8i>
+
+// LLVM-SAME: i8 {{.*}}[[A:%.*]], i8 {{.*}}[[B:%.*]])
+// LLVM: [[TMP0:%.*]] = insertelement <8 x i8> poison, i8 [[A]], i64 0
+// LLVM: [[TMP1:%.*]] = insertelement <8 x i8> poison, i8 [[B]], i64 0
+// LLVM: [[VQSHLB_S8_I:%.*]] = call <8 x i8> @llvm.aarch64.neon.sqshl.v8i8(<8 x i8> [[TMP0]], <8 x i8> [[TMP1]])
+// LLVM: [[TMP2:%.*]] = extractelement <8 x i8> [[VQSHLB_S8_I]], i64 0
+// LLVM: ret i8 [[TMP2]]
+  return vqshlb_s8(a, b);
+}
+
+// LLVM-LABEL: @test_vqshlh_s16(
+// CIR-LABEL: @vqshlh_s16(
+int16_t test_vqshlh_s16(int16_t a, int16_t b) {
+// CIR: {{%.*}} = cir.call_llvm_intrinsic "aarch64.neon.sqshl" {{%.*}}, {{%.*}} : (!cir.vector<4 x !s16i>, !cir.vector<4 x !s16i>) -> !cir.vector<4 x !s16i>
+
+// LLVM-SAME: i16 {{.*}}[[A:%.*]], i16 {{.*}}[[B:%.*]])
+// LLVM: [[TMP0:%.*]] = insertelement <4 x i16> poison, i16 [[A]], i64 0
+// LLVM: [[TMP1:%.*]] = insertelement <4 x i16> poison, i16 [[B]], i64 0
+// LLVM: [[VQSHLH_S16_I:%.*]] = call <4 x i16> @llvm.aarch64.neon.sqshl.v4i16(<4 x i16> [[TMP0]], <4 x i16> [[TMP1]])
+// LLVM: [[TMP2:%.*]] = extractelement <4 x i16> [[VQSHLH_S16_I]], i64 0
+// LLVM: ret i16 [[TMP2]]
+  return vqshlh_s16(a, b);
+}
+
+// LLVM-LABEL: @test_vqshls_s32(
+// CIR-LABEL: @vqshls_s32(
+int32_t test_vqshls_s32(int32_t a, int32_t b) {
+// CIR: {{%.*}} = cir.call_llvm_intrinsic "aarch64.neon.sqshl" {{%.*}}, {{%.*}} : (!s32i, !s32i) -> !s32i
+
+// LLVM-SAME: i32 {{.*}}[[A:%.*]], i32 {{.*}}[[B:%.*]])
+// LLVM: [[VQSHLS_S32_I:%.*]] = call i32 @llvm.aarch64.neon.sqshl.i32(i32 [[A]], i32 [[B]])
+// LLVM: ret i32 [[VQSHLS_S32_I]]
+  return vqshls_s32(a, b);
+}
+
+// LLVM-LABEL: @test_vqshld_s64(
+// CIR-LABEL: @vqshld_s64(
+int64_t test_vqshld_s64(int64_t a, int64_t b) {
+// CIR: {{%.*}} = cir.call_llvm_intrinsic "aarch64.neon.sqshl" {{%.*}}, {{%.*}} : (!s64i, !s64i) -> !s64i
+
+// LLVM-SAME: i64 {{.*}}[[A:%.*]], i64 {{.*}}[[B:%.*]])
+// LLVM: [[VQSHLD_S64_I:%.*]] = call i64 @llvm.aarch64.neon.sqshl.i64(i64 [[A]], i64 [[B]])
+// LLVM: ret i64 [[VQSHLD_S64_I]]
+  return vqshld_s64(a, b);
+}
+
+// LLVM-LABEL: @test_vqshlb_u8(
+// CIR-LABEL: @vqshlb_u8(
+uint8_t test_vqshlb_u8(uint8_t a, int8_t b) {
+// CIR: {{%.*}} = cir.call_llvm_intrinsic "aarch64.neon.uqshl" {{%.*}}, {{%.*}} : (!cir.vector<8 x !u8i>, !cir.vector<8 x !u8i>) -> !cir.vector<8 x !u8i>
+
+// LLVM-SAME: i8 {{.*}}[[A:%.*]], i8 {{.*}}[[B:%.*]])
+// LLVM: [[TMP0:%.*]] = insertelement <8 x i8> poison, i8 [[A]], i64 0
+// LLVM: [[TMP1:%.*]] = insertelement <8 x i8> poison, i8 [[B]], i64 0
+// LLVM: [[VQSHLB_U8_I:%.*]] = call <8 x i8> @llvm.aarch64.neon.uqshl.v8i8(<8 x i8> [[TMP0]], <8 x i8> [[TMP1]])
+// LLVM: [[TMP2:%.*]] = extractelement <8 x i8> [[VQSHLB_U8_I]], i64 0
+// LLVM: ret i8 [[TMP2]]
+  return vqshlb_u8(a, b);
+}
+
+// LLVM-LABEL: @test_vqshlh_u16(
+// CIR-LABEL: @vqshlh_u16(
+uint16_t test_vqshlh_u16(uint16_t a, int16_t b) {
+// CIR: {{%.*}} = cir.call_llvm_intrinsic "aarch64.neon.uqshl" {{%.*}}, {{%.*}} : (!cir.vector<4 x !u16i>, !cir.vector<4 x !u16i>) -> !cir.vector<4 x !u16i>
+
+// LLVM-SAME: i16 {{.*}}[[A:%.*]], i16 {{.*}}[[B:%.*]])
+// LLVM: [[TMP0:%.*]] = insertelement <4 x i16> poison, i16 [[A]], i64 0
+// LLVM: [[TMP1:%.*]] = insertelement <4 x i16> poison, i16 [[B]], i64 0
+// LLVM: [[VQSHLH_U16_I:%.*]] = call <4 x i16> @llvm.aarch64.neon.uqshl.v4i16(<4 x i16> [[TMP0]], <4 x i16> [[TMP1]])
+// LLVM: [[TMP2:%.*]] = extractelement <4 x i16> [[VQSHLH_U16_I]], i64 0
+// LLVM: ret i16 [[TMP2]]
+  return vqshlh_u16(a, b);
+}
+
+// LLVM-LABEL: @test_vqshls_u32(
+// CIR-LABEL: @vqshls_u32(
+uint32_t test_vqshls_u32(uint32_t a, int32_t b) {
+// CIR: {{%.*}} = cir.call_llvm_intrinsic "aarch64.neon.uqshl" {{%.*}}, {{%.*}} : (!u32i, !s32i) -> !u32i
+
+// LLVM-SAME: i32 {{.*}}[[A:%.*]], i32 {{.*}}[[B:%.*]])
+// LLVM: [[VQSHLS_U32_I:%.*]] = call i32 @llvm.aarch64.neon.uqshl.i32(i32 [[A]], i32 [[B]])
+// LLVM: ret i32 [[VQSHLS_U32_I]]
+  return vqshls_u32(a, b);
+}
+
+// LLVM-LABEL: @test_vqshld_u64(
+// CIR-LABEL: @vqshld_u64(
+uint64_t test_vqshld_u64(uint64_t a, int64_t b) {
+// CIR: {{%.*}} = cir.call_llvm_intrinsic "aarch64.neon.uqshl" {{%.*}}, {{%.*}} : (!u64i, !s64i) -> !u64i
+
+// LLVM-SAME: i64 {{.*}}[[A:%.*]], i64 {{.*}}[[B:%.*]])
+// LLVM: [[VQSHLD_U64_I:%.*]] = call i64 @llvm.aarch64.neon.uqshl.i64(i64 [[A]], i64 [[B]])
+// LLVM: ret i64 [[VQSHLD_U64_I]]
+  return vqshld_u64(a, b);
+}
+
+// ALL-LABEL: @test_vqshlu_n_s8(
+uint8x8_t test_vqshlu_n_s8(int8x8_t a) {
+// CIR: [[IMM:%.*]] = cir.const #cir.int<3> : !u8i
+// CIR: [[AMT:%.*]] = cir.vec.splat [[IMM]] : !u8i, !cir.vector<8 x !u8i>
+// CIR: {{%.*}} = cir.call_llvm_intrinsic "aarch64.neon.sqshlu" {{%.*}}, [[AMT]] : (!cir.vector<8 x !u8i>, !cir.vector<8 x !u8i>) -> !cir.vector<8 x !u8i>
+
+// LLVM-SAME: <8 x i8> {{.*}}[[A:%.*]])
+// LLVM: [[VQSHLU_N:%.*]] = call <8 x i8> @llvm.aarch64.neon.sqshlu.v8i8(<8 x i8> [[A]], <8 x i8> splat (i8 3))
+// LLVM: ret <8 x i8> [[VQSHLU_N]]
+  return vqshlu_n_s8(a, 3);
+}
+
+// ALL-LABEL: @test_vqshlu_n_s16(
+uint16x4_t test_vqshlu_n_s16(int16x4_t a) {
+// CIR: [[IMM:%.*]] = cir.const #cir.int<3> : !u16i
+// CIR: [[AMT:%.*]] = cir.vec.splat [[IMM]] : !u16i, !cir.vector<4 x !u16i>
+// CIR: {{%.*}} = cir.call_llvm_intrinsic "aarch64.neon.sqshlu" {{%.*}}, [[AMT]] : (!cir.vector<4 x !u16i>, !cir.vector<4 x !u16i>) -> !cir.vector<4 x !u16i>
+
+// LLVM-SAME: <4 x i16> {{.*}}[[A:%.*]])
+// LLVM: [[TMP0:%.*]] = bitcast <4 x i16> [[A]] to <8 x i8>
+// LLVM: [[VQSHLU_N:%.*]] = bitcast <8 x i8> [[TMP0]] to <4 x i16>
+// LLVM: [[VQSHLU_N1:%.*]] = call <4 x i16> @llvm.aarch64.neon.sqshlu.v4i16(<4 x i16> [[VQSHLU_N]], <4 x i16> splat (i16 3))
+// LLVM: ret <4 x i16> [[VQSHLU_N1]]
+  return vqshlu_n_s16(a, 3);
+}
+
+// ALL-LABEL: @test_vqshlu_n_s32(
+uint32x2_t test_vqshlu_n_s32(int32x2_t a) {
+// CIR: [[IMM:%.*]] = cir.const #cir.int<3> : !u32i
+// CIR: [[AMT:%.*]] = cir.vec.splat [[IMM]] : !u32i, !cir.vector<2 x !u32i>
+// CIR: {{%.*}} = cir.call_llvm_intrinsic "aarch64.neon.sqshlu" {{%.*}}, [[AMT]] : (!cir.vector<2 x !u32i>, !cir.vector<2 x !u32i>) -> !cir.vector<2 x !u32i>
+
+// LLVM-SAME: <2 x i32> {{.*}}[[A:%.*]])
+// LLVM: [[TMP0:%.*]] = bitcast <2 x i32> [[A]] to <8 x i8>
+// LLVM: [[VQSHLU_N:%.*]] = bitcast <8 x i8> [[TMP0]] to <2 x i32>
+// LLVM: [[VQSHLU_N1:%.*]] = call <2 x i32> @llvm.aarch64.neon.sqshlu.v2i32(<2 x i32> [[VQSHLU_N]], <2 x i32> splat (i32 3))
+// LLVM: ret <2 x i32> [[VQSHLU_N1]]
+  return vqshlu_n_s32(a, 3);
+}
+
+// ALL-LABEL: @test_vqshluq_n_s8(
+uint8x16_t test_vqshluq_n_s8(int8x16_t a) {
+// CIR: [[IMM:%.*]] = cir.const #cir.int<3> : !u8i
+// CIR: [[AMT:%.*]] = cir.vec.splat [[IMM]] : !u8i, !cir.vector<16 x !u8i>
+// CIR: {{%.*}} = cir.call_llvm_intrinsic "aarch64.neon.sqshlu" {{%.*}}, [[AMT]] : (!cir.vector<16 x !u8i>, !cir.vector<16 x !u8i>) -> !cir.vector<16 x !u8i>
+
+// LLVM-SAME: <16 x i8> {{.*}}[[A:%.*]])
+// LLVM: [[VQSHLU_N:%.*]] = call <16 x i8> @llvm.aarch64.neon.sqshlu.v16i8(<16 x i8> [[A]], <16 x i8> splat (i8 3))
+// LLVM: ret <16 x i8> [[VQSHLU_N]]
+  return vqshluq_n_s8(a, 3);
+}
+
+// ALL-LABEL: @test_vqshluq_n_s16(
+uint16x8_t test_vqshluq_n_s16(int16x8_t a) {
+// CIR: [[IMM:%.*]] = cir.const #cir.int<3> : !u16i
+// CIR: [[AMT:%.*]] = cir.vec.splat [[IMM]] : !u16i, !cir.vector<8 x !u16i>
+// CIR: {{%.*}} = cir.call_llvm_intrinsic "aarch64.neon.sqshlu" {{%.*}}, [[AMT]] : (!cir.vector<8 x !u16i>, !cir.vector<8 x !u16i>) -> !cir.vector<8 x !u16i>
+
+// LLVM-SAME: <8 x i16> {{.*}}[[A:%.*]])
+// LLVM: [[TMP0:%.*]] = bitcast <8 x i16> [[A]] to <16 x i8>
+// LLVM: [[VQSHLU_N:%.*]] = bitcast <16 x i8> [[TMP0]] to <8 x i16>
+// LLVM: [[VQSHLU_N1:%.*]] = call <8 x i16> @llvm.aarch64.neon.sqshlu.v8i16(<8 x i16> [[VQSHLU_N]], <8 x i16> splat (i16 3))
+// LLVM: ret <8 x i16> [[VQSHLU_N1]]
+  return vqshluq_n_s16(a, 3);
+}
+
+// ALL-LABEL: @test_vqshluq_n_s32(
+uint32x4_t test_vqshluq_n_s32(int32x4_t a) {
+// CIR: [[IMM:%.*]] = cir.const #cir.int<3> : !u32i
+// CIR: [[AMT:%.*]] = cir.vec.splat [[IMM]] : !u32i, !cir.vector<4 x !u32i>
+// CIR: {{%.*}} = cir.call_llvm_intrinsic "aarch64.neon.sqshlu" {{%.*}}, [[AMT]] : (!cir.vector<4 x !u32i>, !cir.vector<4 x !u32i>) -> !cir.vector<4 x !u32i>
+
+// LLVM-SAME: <4 x i32> {{.*}}[[A:%.*]])
+// LLVM: [[TMP0:%.*]] = bitcast <4 x i32> [[A]] to <16 x i8>
+// LLVM: [[VQSHLU_N:%.*]] = bitcast <16 x i8> [[TMP0]] to <4 x i32>
+// LLVM: [[VQSHLU_N1:%.*]] = call <4 x i32> @llvm.aarch64.neon.sqshlu.v4i32(<4 x i32> [[VQSHLU_N]], <4 x i32> splat (i32 3))
+// LLVM: ret <4 x i32> [[VQSHLU_N1]]
+  return vqshluq_n_s32(a, 3);
+}
+
+// ALL-LABEL: @test_vqshluq_n_s64(
+uint64x2_t test_vqshluq_n_s64(int64x2_t a) {
+// CIR: [[IMM:%.*]] = cir.const #cir.int<3> : !u64i
+// CIR: [[AMT:%.*]] = cir.vec.splat [[IMM]] : !u64i, !cir.vector<2 x !u64i>
+// CIR: {{%.*}} = cir.call_llvm_intrinsic "aarch64.neon.sqshlu" {{%.*}}, [[AMT]] : (!cir.vector<2 x !u64i>, !cir.vector<2 x !u64i>) -> !cir.vector<2 x !u64i>
+
+// LLVM-SAME: <2 x i64> {{.*}}[[A:%.*]])
+// LLVM: [[TMP0:%.*]] = bitcast <2 x i64> [[A]] to <16 x i8>
+// LLVM: [[VQSHLU_N:%.*]] = bitcast <16 x i8> [[TMP0]] to <2 x i64>
+// LLVM: [[VQSHLU_N1:%.*]] = call <2 x i64> @llvm.aarch64.neon.sqshlu.v2i64(<2 x i64> [[VQSHLU_N]], <2 x i64> splat (i64 3))
+// LLVM: ret <2 x i64> [[VQSHLU_N1]]
+  return vqshluq_n_s64(a, 3);
+}
+
+// ALL-LABEL: @test_vqshlb_n_s8(
+int8_t test_vqshlb_n_s8(int8_t a) {
+// CIR: [[IMM:%.*]] = cir.const #cir.int<7> : !s8i
+// CIR: [[AMT:%.*]] = cir.vec.insert [[IMM]], {{.*}} : !cir.vector<8 x !s8i>
+// CIR: {{%.*}} = cir.call_llvm_intrinsic "aarch64.neon.sqshl" {{%.*}}, [[AMT]] : (!cir.vector<8 x !s8i>, !cir.vector<8 x !s8i>) -> !cir.vector<8 x !s8i>
+
+// LLVM-SAME: i8 {{.*}}[[A:%.*]])
+// LLVM: [[TMP0:%.*]] = insertelement <8 x i8> poison, i8 [[A]], i64 0
+// LLVM: [[VQSHLB_N_S8:%.*]] = call <8 x i8> @llvm.aarch64.neon.sqshl.v8i8(<8 x i8> [[TMP0]], <8 x i8> <i8 7, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison>)
+// LLVM: [[TMP1:%.*]] = extractelement <8 x i8> [[VQSHLB_N_S8]], i64 0
+// LLVM: ret i8 [[TMP1]]
+  return (int8_t)vqshlb_n_s8(a, 7);
+}
+
+// ALL-LABEL: @test_vqshlh_n_s16(
+int16_t test_vqshlh_n_s16(int16_t a) {
+// CIR: [[IMM:%.*]] = cir.const #cir.int<15> : !s16i
+// CIR: [[AMT:%.*]] = cir.vec.insert [[IMM]], {{.*}} : !cir.vector<4 x !s16i>
+// CIR: {{%.*}} = cir.call_llvm_intrinsic "aarch64.neon.sqshl" {{%.*}}, [[AMT]] : (!cir.vector<4 x !s16i>, !cir.vector<4 x !s16i>) -> !cir.vector<4 x !s16i>
+
+// LLVM-SAME: i16 {{.*}}[[A:%.*]])
+// LLVM: [[TMP0:%.*]] = insertelement <4 x i16> poison, i16 [[A]], i64 0
+// LLVM: [[VQSHLH_N_S16:%.*]] = call <4 x i16> @llvm.aarch64.neon.sqshl.v4i16(<4 x i16> [[TMP0]], <4 x i16> <i16 15, i16 poison, i16 poison, i16 poison>)
+// LLVM: [[TMP1:%.*]] = extractelement <4 x i16> [[VQSHLH_N_S16]], i64 0
+// LLVM: ret i16 [[TMP1]]
+  return (int16_t)vqshlh_n_s16(a, 15);
+}
+
+// ALL-LABEL: @test_vqshls_n_s32(
+int32_t test_vqshls_n_s32(int32_t a) {
+// CIR: [[AMT:%.*]] = cir.const #cir.int<31> : !s32i
+// CIR: {{%.*}} = cir.call_llvm_intrinsic "aarch64.neon.sqshl" {{%.*}}, [[AMT]] : (!s32i, !s32i) -> !s32i
+
+// LLVM-SAME: i32 {{.*}}[[A:%.*]])
+// LLVM: [[VQSHLS_N_S32:%.*]] = call i32 @llvm.aarch64.neon.sqshl.i32(i32 [[A]], i32 31)
+// LLVM: ret i32 [[VQSHLS_N_S32]]
+  return (int32_t)vqshls_n_s32(a, 31);
+}
+
+// ALL-LABEL: @test_vqshl_n_s8(
+int8x8_t test_vqshl_n_s8(int8x8_t a) {
+// CIR: [[IMM:%.*]] = cir.const #cir.int<0> : !s8i
+// CIR: [[AMT:%.*]] = cir.vec.splat [[IMM]] : !s8i, !cir.vector<8 x !s8i>
+// CIR: {{%.*}} = cir.call_llvm_intrinsic "aarch64.neon.sqshl" {{%.*}}, [[AMT]] : (!cir.vector<8 x !s8i>, !cir.vector<8 x !s8i>) -> !cir.vector<8 x !s8i>
+
+// LLVM-SAME: <8 x i8> {{.*}}[[A:%.*]])
+// LLVM: [[VQSHL_N:%.*]] = call <8 x i8> @llvm.aarch64.neon.sqshl.v8i8(<8 x i8> [[A]], <8 x i8> zeroinitializer)
+// LLVM: ret <8 x i8> [[VQSHL_N]]
+  return vqshl_n_s8(a, 0);
+}
+
+// ALL-LABEL: @test_vqshlq_n_s8(
+int8x16_t test_vqshlq_n_s8(int8x16_t a) {
+// CIR: [[IMM:%.*]] = cir.const #cir.int<0> : !s8i
+// CIR: [[AMT:%.*]] = cir.vec.splat [[IMM]] : !s8i, !cir.vector<16 x !s8i>
+// CIR: {{%.*}} = cir.call_llvm_intrinsic "aarch64.neon.sqshl" {{%.*}}, [[AMT]] : (!cir.vector<16 x !s8i>, !cir.vector<16 x !s8i>) -> !cir.vector<16 x !s8i>
+
+// LLVM-SAME: <16 x i8> {{.*}}[[A:%.*]])
+// LLVM: [[VQSHL_N:%.*]] = call <16 x i8> @llvm.aarch64.neon.sqshl.v16i8(<16 x i8> [[A]], <16 x i8> zeroinitializer)
+// LLVM: ret <16 x i8> [[VQSHL_N]]
+  return vqshlq_n_s8(a, 0);
+}
+
+// ALL-LABEL: @test_vqshl_n_s16(
+int16x4_t test_vqshl_n_s16(int16x4_t a) {
+// CIR: [[IMM:%.*]] = cir.const #cir.int<0> : !s16i
+// CIR: [[AMT:%.*]] = cir.vec.splat [[IMM]] : !s16i, !cir.vector<4 x !s16i>
+// CIR: {{%.*}} = cir.call_llvm_intrinsic "aarch64.neon.sqshl" {{%.*}}, [[AMT]] : (!cir.vector<4 x !s16i>, !cir.vector<4 x !s16i>) -> !cir.vector<4 x !s16i>
+
+// LLVM-SAME: <4 x i16> {{.*}}[[A:%.*]])
+// LLVM: [[TMP0:%.*]] = bitcast <4 x i16> [[A]] to <8 x i8>
+// LLVM: [[VQSHL_N:%.*]] = bitcast <8 x i8> [[TMP0]] to <4 x i16>
+// LLVM: [[VQSHL_N1:%.*]] = call <4 x i16> @llvm.aarch64.neon.sqshl.v4i16(<4 x i16> [[VQSHL_N]], <4 x i16> zeroinitializer)
+// LLVM: ret <4 x i16> [[VQSHL_N1]]
+  return vqshl_n_s16(a, 0);
+}
+
+// ALL-LABEL: @test_vqshlq_n_s16(
+int16x8_t test_vqshlq_n_s16(int16x8_t a) {
+// CIR: [[IMM:%.*]] = cir.const #cir.int<0> : !s16i
+// CIR: [[AMT:%.*]] = cir.vec.splat [[IMM]] : !s16i, !cir.vector<8 x !s16i>
+// CIR: {{%.*}} = cir.call_llvm_intrinsic "aarch64.neon.sqshl" {{%.*}}, [[AMT]] : (!cir.vector<8 x !s16i>, !cir.vector<8 x !s16i>) -> !cir.vector<8 x !s16i>
+
+// LLVM-SAME: <8 x i16> {{.*}}[[A:%.*]])
+// LLVM: [[TMP0:%.*]] = bitcast <8 x i16> [[A]] to <16 x i8>
+// LLVM: [[VQSHL_N:%.*]] = bitcast <16 x i8> [[TMP0]] to <8 x i16>
+// LLVM: [[VQSHL_N1:%.*]] = call <8 x i16> @llvm.aarch64.neon.sqshl.v8i16(<8 x i16> [[VQSHL_N]], <8 x i16> zeroinitializer)
+// LLVM: ret <8 x i16> [[VQSHL_N1]]
+  return vqshlq_n_s16(a, 0);
+}
+
+// ALL-LABEL: @test_vqshl_n_s32(
+int32x2_t test_vqshl_n_s32(int32x2_t a) {
+// CIR: [[IMM:%.*]] = cir.const #cir.int<0> : !s32i
+// CIR: [[AMT:%.*]] = cir.vec.splat [[IMM]] : !s32i, !cir.vector<2 x !s32i>
+// CIR: {{%.*}} = cir.call_llvm_intrinsic "aarch64.neon.sqshl" {{%.*}}, [[AMT]] : (!cir.vector<2 x !s32i>, !cir.vector<2 x !s32i>) -> !cir.vector<2 x !s32i>
+
+// LLVM-SAME: <2 x i32> {{.*}}[[A:%.*]])
+// LLVM: [[TMP0:%.*]] = bitcast <2 x i32> [[A]] to <8 x i8>
+// LLVM: [[VQSHL_N:%.*]] = bitcast <8 x i8> [[TMP0]] to <2 x i32>
+// LLVM: [[VQSHL_N1:%.*]] = call <2 x i32> @llvm.aarch64.neon.sqshl.v2i32(<2 x i32> [[VQSHL_N]], <2 x i32> zeroinitializer)
+// LLVM: ret <2 x i32> [[VQSHL_N1]]
+  return vqshl_n_s32(a, 0);
+}
+
+// ALL-LABEL: @test_vqshlq_n_s32(
+int32x4_t test_vqshlq_n_s32(int32x4_t a) {
+// CIR: [[IMM:%.*]] = cir.const #cir.int<0> : !s32i
+// CIR: [[AMT:%.*]] = cir.vec.splat [[IMM]] : !s32i, !cir.vector<4 x !s32i>
+// CIR: {{%.*}} = cir.call_llvm_intrinsic "aarch64.neon.sqshl" {{%.*}}, [[AMT]] : (!cir.vector<4 x !s32i>, !cir.vector<4 x !s32i>) -> !cir.vector<4 x !s32i>
+
+// LLVM-SAME: <4 x i32> {{.*}}[[A:%.*]])
+// LLVM: [[TMP0:%.*]] = bitcast <4 x i32> [[A]] to <16 x i8>
+// LLVM: [[VQSHL_N:%.*]] = bitcast <16 x i8> [[TMP0]] to <4 x i32>
+// LLVM: [[VQSHL_N1:%.*]] = call <4 x i32> @llvm.aarch64.neon.sqshl.v4i32(<4 x i32> [[VQSHL_N]], <4 x i32> zeroinitializer)
+// LLVM: ret <4 x i32> [[VQSHL_N1]]
+  return vqshlq_n_s32(a, 0);
+}
+
+// ALL-LABEL: @test_vqshlq_n_s64(
+int64x2_t test_vqshlq_n_s64(int64x2_t a) {
+// CIR: [[IMM:%.*]] = cir.const #cir.int<0> : !s64i
+// CIR: [[AMT:%.*]] = cir.vec.splat [[IMM]] : !s64i, !cir.vector<2 x !s64i>
+// CIR: {{%.*}} = cir.call_llvm_intrinsic "aarch64.neon.sqshl" {{%.*}}, [[AMT]] : (!cir.vector<2 x !s64i>, !cir.vector<2 x !s64i>) -> !cir.vector<2 x !s64i>
+
+// LLVM-SAME: <2 x i64> {{.*}}[[A:%.*]])
+// LLVM: [[TMP0:%.*]] = bitcast <2 x i64> [[A]] to <16 x i8>
+// LLVM: [[VQSHL_N:%.*]] = bitcast <16 x i8> [[TMP0]] to <2 x i64>
+// LLVM: [[VQSHL_N1:%.*]] = call <2 x i64> @llvm.aarch64.neon.sqshl.v2i64(<2 x i64> [[VQSHL_N]], <2 x i64> zeroinitializer)
+// LLVM: ret <2 x i64> [[VQSHL_N1]]
+  return vqshlq_n_s64(a, 0);
+}
+
+// ALL-LABEL: @test_vqshl_n_u8(
+uint8x8_t test_vqshl_n_u8(uint8x8_t a) {
+// CIR: [[IMM:%.*]] = cir.const #cir.int<0> : !u8i
+// CIR: [[AMT:%.*]] = cir.vec.splat [[IMM]] : !u8i, !cir.vector<8 x !u8i>
+// CIR: {{%.*}} = cir.call_llvm_intrinsic "aarch64.neon.uqshl" {{%.*}}, [[AMT]] : (!cir.vector<8 x !u8i>, !cir.vector<8 x !u8i>) -> !cir.vector<8 x !u8i>
+
+// LLVM-SAME: <8 x i8> {{.*}}[[A:%.*]])
+// LLVM: [[VQSHL_N:%.*]] = call <8 x i8> @llvm.aarch64.neon.uqshl.v8i8(<8 x i8> [[A]], <8 x i8> zeroinitializer)
+// LLVM: ret <8 x i8> [[VQSHL_N]]
+  return vqshl_n_u8(a, 0);
+}
+
+// ALL-LABEL: @test_vqshlq_n_u8(
+uint8x16_t test_vqshlq_n_u8(uint8x16_t a) {
+// CIR: [[IMM:%.*]] = cir.const #cir.int<0> : !u8i
+// CIR: [[AMT:%.*]] = cir.vec.splat [[IMM]] : !u8i, !cir.vector<16 x !u8i>
+// CIR: {{%.*}} = cir.call_llvm_intrinsic "aarch64.neon.uqshl" {{%.*}}, [[AMT]] : (!cir.vector<16 x !u8i>, !cir.vector<16 x !u8i>) -> !cir.vector<16 x !u8i>
+
+// LLVM-SAME: <16 x i8> {{.*}}[[A:%.*]])
+// LLVM: [[VQSHL_N:%.*]] = call <16 x i8> @llvm.aarch64.neon.uqshl.v16i8(<16 x i8> [[A]], <16 x i8> zeroinitializer)
+// LLVM: ret <16 x i8> [[VQSHL_N]]
+  return vqshlq_n_u8(a, 0);
+}
+
+// ALL-LABEL: @test_vqshl_n_u16(
+uint16x4_t test_vqshl_n_u16(uint16x4_t a) {
+// CIR: [[IMM:%.*]] = cir.const #cir.int<0> : !u16i
+// CIR: [[AMT:%.*]] = cir.vec.splat [[IMM]] : !u16i, !cir.vector<4 x !u16i>
+// CIR: {{%.*}} = cir.call_llvm_intrinsic "aarch64.neon.uqshl" {{%.*}}, [[AMT]] : (!cir.vector<4 x !u16i>, !cir.vector<4 x !u16i>) -> !cir.vector<4 x !u16i>
+
+// LLVM-SAME: <4 x i16> {{.*}}[[A:%.*]])
+// LLVM: [[TMP0:%.*]] = bitcast <4 x i16> [[A]] to <8 x i8>
+// LLVM: [[VQSHL_N:%.*]] = bitcast <8 x i8> [[TMP0]] to <4 x i16>
+// LLVM: [[VQSHL_N1:%.*]] = call <4 x i16> @llvm.aarch64.neon.uqshl.v4i16(<4 x i16> [[VQSHL_N]], <4 x i16> zeroinitializer)
+// LLVM: ret <4 x i16> [[VQSHL_N1]]
+  return vqshl_n_u16(a, 0);
+}
+
+// ALL-LABEL: @test_vqshlq_n_u16(
+uint16x8_t test_vqshlq_n_u16(uint16x8_t a) {
+// CIR: [[IMM:%.*]] = cir.const #cir.int<0> : !u16i
+// CIR: [[AMT:%.*]] = cir.vec.splat [[IMM]] : !u16i, !cir.vector<8 x !u16i>
+// CIR: {{%.*}} = cir.call_llvm_intrinsic "aarch64.neon.uqshl" {{%.*}}, [[AMT]] : (!cir.vector<8 x !u16i>, !cir.vector<8 x !u16i>) -> !cir.vector<8 x !u16i>
+
+// LLVM-SAME: <8 x i16> {{.*}}[[A:%.*]])
+// LLVM: [[TMP0:%.*]] = bitcast <8 x i16> [[A]] to <16 x i8>
+// LLVM: [[VQSHL_N:%.*]] = bitcast <16 x i8> [[TMP0]] to <8 x i16>
+// LLVM: [[VQSHL_N1:%.*]] = call <8 x i16> @llvm.aarch64.neon.uqshl.v8i16(<8 x i16> [[VQSHL_N]], <8 x i16> zeroinitializer)
+// LLVM: ret <8 x i16> [[VQSHL_N1]]
+  return vqshlq_n_u16(a, 0);
+}
+
+// ALL-LABEL: @test_vqshl_n_u32(
+uint32x2_t test_vqshl_n_u32(uint32x2_t a) {
+// CIR: [[IMM:%.*]] = cir.const #cir.int<0> : !u32i
+// CIR: [[AMT:%.*]] = cir.vec.splat [[IMM]] : !u32i, !cir.vector<2 x !u32i>
+// CIR: {{%.*}} = cir.call_llvm_intrinsic "aarch64.neon.uqshl" {{%.*}}, [[AMT]] : (!cir.vector<2 x !u32i>, !cir.vector<2 x !u32i>) -> !cir.vector<2 x !u32i>
+
+// LLVM-SAME: <2 x i32> {{.*}}[[A:%.*]])
+// LLVM: [[TMP0:%.*]] = bitcast <2 x i32> [[A]] to <8 x i8>
+// LLVM: [[VQSHL_N:%.*]] = bitcast <8 x i8> [[TMP0]] to <2 x i32>
+// LLVM: [[VQSHL_N1:%.*]] = call <2 x i32> @llvm.aarch64.neon.uqshl.v2i32(<2 x i32> [[VQSHL_N]], <2 x i32> zeroinitializer)
+// LLVM: ret <2 x i32> [[VQSHL_N1]]
+  return vqshl_n_u32(a, 0);
+}
+
+// ALL-LABEL: @test_vqshlq_n_u32(
+uint32x4_t test_vqshlq_n_u32(uint32x4_t a) {
+// CIR: [[IMM:%.*]] = cir.const #cir.int<0> : !u32i
+// CIR: [[AMT:%.*]] = cir.vec.splat [[IMM]] : !u32i, !cir.vector<4 x !u32i>
+// CIR: {{%.*}} = cir.call_llvm_intrinsic "aarch64.neon.uqshl" {{%.*}}, [[AMT]] : (!cir.vector<4 x !u32i>, !cir.vector<4 x !u32i>) -> !cir.vector<4 x !u32i>
+
+// LLVM-SAME: <4 x i32> {{.*}}[[A:%.*]])
+// LLVM: [[TMP0:%.*]] = bitcast <4 x i32> [[A]] to <16 x i8>
+// LLVM: [[VQSHL_N:%.*]] = bitcast <16 x i8> [[TMP0]] to <4 x i32>
+// LLVM: [[VQSHL_N1:%.*]] = call <4 x i32> @llvm.aarch64.neon.uqshl.v4i32(<4 x i32> [[VQSHL_N]], <4 x i32> zeroinitializer)
+// LLVM: ret <4 x i32> [[VQSHL_N1]]
+  return vqshlq_n_u32(a, 0);
+}
+
+// ALL-LABEL: @test_vqshlq_n_u64(
+uint64x2_t test_vqshlq_n_u64(uint64x2_t a) {
+// CIR: [[IMM:%.*]] = cir.const #cir.int<0> : !u64i
+// CIR: [[AMT:%.*]] = cir.vec.splat [[IMM]] : !u64i, !cir.vector<2 x !u64i>
+// CIR: {{%.*}} = cir.call_llvm_intrinsic "aarch64.neon.uqshl" {{%.*}}, [[AMT]] : (!cir.vector<2 x !u64i>, !cir.vector<2 x !u64i>) -> !cir.vector<2 x !u64i>
+
+// LLVM-SAME: <2 x i64> {{.*}}[[A:%.*]])
+// LLVM: [[TMP0:%.*]] = bitcast <2 x i64> [[A]] to <16 x i8>
+// LLVM: [[VQSHL_N:%.*]] = bitcast <16 x i8> [[TMP0]] to <2 x i64>
+// LLVM: [[VQSHL_N1:%.*]] = call <2 x i64> @llvm.aarch64.neon.uqshl.v2i64(<2 x i64> [[VQSHL_N]], <2 x i64> zeroinitializer)
+// LLVM: ret <2 x i64> [[VQSHL_N1]]
+  return vqshlq_n_u64(a, 0);
+}
+
+// ALL-LABEL: @test_vqshl_n_s64(
+int64x1_t test_vqshl_n_s64(int64x1_t a) {
+// CIR: [[IMM:%.*]] = cir.const #cir.int<1> : !s64i
+// CIR: [[AMT:%.*]] = cir.vec.splat [[IMM]] : !s64i, !cir.vector<1 x !s64i>
+// CIR: {{%.*}} = cir.call_llvm_intrinsic "aarch64.neon.sqshl" {{%.*}}, [[AMT]] : (!cir.vector<1 x !s64i>, !cir.vector<1 x !s64i>) -> !cir.vector<1 x !s64i>
+
+// LLVM-SAME: <1 x i64> {{.*}}[[A:%.*]])
+// LLVM: [[TMP0:%.*]] = bitcast <1 x i64> [[A]] to <8 x i8>
+// LLVM: [[VQSHL_N:%.*]] = bitcast <8 x i8> [[TMP0]] to <1 x i64>
+// LLVM: [[VQSHL_N1:%.*]] = call <1 x i64> @llvm.aarch64.neon.sqshl.v1i64(<1 x i64> [[VQSHL_N]], <1 x i64> splat (i64 1))
+// LLVM: ret <1 x i64> [[VQSHL_N1]]
+  return vqshl_n_s64(a, 1);
+}
+
+// ALL-LABEL: @test_vqshlb_n_u8(
+uint8_t test_vqshlb_n_u8(uint8_t a) {
+// CIR: [[IMM:%.*]] = cir.const #cir.int<7> : !u8i
+// CIR: [[AMT:%.*]] = cir.vec.insert [[IMM]], {{.*}} : !cir.vector<8 x !u8i>
+// CIR: {{%.*}} = cir.call_llvm_intrinsic "aarch64.neon.uqshl" {{%.*}}, [[AMT]] : (!cir.vector<8 x !u8i>, !cir.vector<8 x !u8i>) -> !cir.vector<8 x !u8i>
+
+// LLVM-SAME: i8 {{.*}}[[A:%.*]])
+// LLVM: [[TMP0:%.*]] = insertelement <8 x i8> poison, i8 [[A]], i64 0
+// LLVM: [[VQSHLB_N_U8:%.*]] = call <8 x i8> @llvm.aarch64.neon.uqshl.v8i8(<8 x i8> [[TMP0]], <8 x i8> <i8 7, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison>)
+// LLVM: [[TMP1:%.*]] = extractelement <8 x i8> [[VQSHLB_N_U8]], i64 0
+// LLVM: ret i8 [[TMP1]]
+  return (uint8_t)vqshlb_n_u8(a, 7);
+}
+
+// ALL-LABEL: @test_vqshlh_n_u16(
+uint16_t test_vqshlh_n_u16(uint16_t a) {
+// CIR: [[IMM:%.*]] = cir.const #cir.int<15> : !u16i
+// CIR: [[AMT:%.*]] = cir.vec.insert [[IMM]], {{.*}} : !cir.vector<4 x !u16i>
+// CIR: {{%.*}} = cir.call_llvm_intrinsic "aarch64.neon.uqshl" {{%.*}}, [[AMT]] : (!cir.vector<4 x !u16i>, !cir.vector<4 x !u16i>) -> !cir.vector<4 x !u16i>
+
+// LLVM-SAME: i16 {{.*}}[[A:%.*]])
+// LLVM: [[TMP0:%.*]] = insertelement <4 x i16> poison, i16 [[A]], i64 0
+// LLVM: [[VQSHLH_N_U16:%.*]] = call <4 x i16> @llvm.aarch64.neon.uqshl.v4i16(<4 x i16> [[TMP0]], <4 x i16> <i16 15, i16 poison, i16 poison, i16 poison>)
+// LLVM: [[TMP1:%.*]] = extractelement <4 x i16> [[VQSHLH_N_U16]], i64 0
+// LLVM: ret i16 [[TMP1]]
+  return (uint16_t)vqshlh_n_u16(a, 15);
+}
+
+// ALL-LABEL: @test_vqshls_n_u32(
+uint32_t test_vqshls_n_u32(uint32_t a) {
+// CIR: [[AMT:%.*]] = cir.const #cir.int<31> : !s32i
+// CIR: {{%.*}} = cir.call_llvm_intrinsic "aarch64.neon.uqshl" {{%.*}}, [[AMT]] : (!u32i, !s32i) -> !u32i
+
+// LLVM-SAME: i32 {{.*}}[[A:%.*]])
+// LLVM: [[VQSHLS_N_U32:%.*]] = call i32 @llvm.aarch64.neon.uqshl.i32(i32 [[A]], i32 31)
+// LLVM: ret i32 [[VQSHLS_N_U32]]
+  return (uint32_t)vqshls_n_u32(a, 31);
+}
+
+// ALL-LABEL: @test_vqshl_n_u64(
+uint64x1_t test_vqshl_n_u64(uint64x1_t a) {
+// CIR: [[IMM:%.*]] = cir.const #cir.int<1> : !u64i
+// CIR: [[AMT:%.*]] = cir.vec.splat [[IMM]] : !u64i, !cir.vector<1 x !u64i>
+// CIR: {{%.*}} = cir.call_llvm_intrinsic "aarch64.neon.uqshl" {{%.*}}, [[AMT]] : (!cir.vector<1 x !u64i>, !cir.vector<1 x !u64i>) -> !cir.vector<1 x !u64i>
+
+// LLVM-SAME: <1 x i64> {{.*}}[[A:%.*]])
+// LLVM: [[TMP0:%.*]] = bitcast <1 x i64> [[A]] to <8 x i8>
+// LLVM: [[VQSHL_N:%.*]] = bitcast <8 x i8> [[TMP0]] to <1 x i64>
+// LLVM: [[VQSHL_N1:%.*]] = call <1 x i64> @llvm.aarch64.neon.uqshl.v1i64(<1 x i64> [[VQSHL_N]], <1 x i64> splat (i64 1))
+// LLVM: ret <1 x i64> [[VQSHL_N1]]
+  return vqshl_n_u64(a, 1);
+}
+
+// ALL-LABEL: @test_vqshlub_n_s8(
+int8_t test_vqshlub_n_s8(int8_t a) {
+// CIR: [[IMM:%.*]] = cir.const #cir.int<7> : !s8i
+// CIR: [[AMT:%.*]] = cir.vec.insert [[IMM]], {{.*}} : !cir.vector<8 x !s8i>
+// CIR: {{%.*}} = cir.call_llvm_intrinsic "aarch64.neon.sqshlu" {{%.*}}, [[AMT]] : (!cir.vector<8 x !s8i>, !cir.vector<8 x !s8i>) -> !cir.vector<8 x !s8i>
+
+// LLVM-SAME: i8 {{.*}}[[A:%.*]])
+// LLVM: [[TMP0:%.*]] = insertelement <8 x i8> poison, i8 [[A]], i64 0
+// LLVM: [[VQSHLUB_N_S8:%.*]] = call <8 x i8> @llvm.aarch64.neon.sqshlu.v8i8(<8 x i8> [[TMP0]], <8 x i8> <i8 7, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison, i8 poison>)
+// LLVM: [[TMP1:%.*]] = extractelement <8 x i8> [[VQSHLUB_N_S8]], i64 0
+// LLVM: ret i8 [[TMP1]]
+  return (int8_t)vqshlub_n_s8(a, 7);
+}
+
+// ALL-LABEL: @test_vqshluh_n_s16(
+int16_t test_vqshluh_n_s16(int16_t a) {
+// CIR: [[IMM:%.*]] = cir.const #cir.int<15> : !s16i
+// CIR: [[AMT:%.*]] = cir.vec.insert [[IMM]], {{.*}} : !cir.vector<4 x !s16i>
+// CIR: {{%.*}} = cir.call_llvm_intrinsic "aarch64.neon.sqshlu" {{%.*}}, [[AMT]] : (!cir.vector<4 x !s16i>, !cir.vector<4 x !s16i>) -> !cir.vector<4 x !s16i>
+
+// LLVM-SAME: i16 {{.*}}[[A:%.*]])
+// LLVM: [[TMP0:%.*]] = insertelement <4 x i16> poison, i16 [[A]], i64 0
+// LLVM: [[VQSHLUH_N_S16:%.*]] = call <4 x i16> @llvm.aarch64.neon.sqshlu.v4i16(<4 x i16> [[TMP0]], <4 x i16> <i16 15, i16 poison, i16 poison, i16 poison>)
+// LLVM: [[TMP1:%.*]] = extractelement <4 x i16> [[VQSHLUH_N_S16]], i64 0
+// LLVM: ret i16 [[TMP1]]
+  return (int16_t)vqshluh_n_s16(a, 15);
+}
+
+// ALL-LABEL: @test_vqshlus_n_s32(
+int32_t test_vqshlus_n_s32(int32_t a) {
+// CIR: [[AMT:%.*]] = cir.const #cir.int<31> : !s32i
+// CIR: {{%.*}} = cir.call_llvm_intrinsic "aarch64.neon.sqshlu" {{%.*}}, [[AMT]] : (!s32i, !s32i) -> !s32i
+
+// LLVM-SAME: i32 {{.*}}[[A:%.*]])
+// LLVM: [[VQSHLUS_N_S32:%.*]] = call i32 @llvm.aarch64.neon.sqshlu.i32(i32 [[A]], i32 31)
+// LLVM: ret i32 [[VQSHLUS_N_S32]]
+  return (int32_t)vqshlus_n_s32(a, 31);
+}
+
+// ALL-LABEL: @test_vqshlu_n_s64(
+uint64x1_t test_vqshlu_n_s64(int64x1_t a) {
+// CIR: [[IMM:%.*]] = cir.const #cir.int<1> : !u64i
+// CIR: [[AMT:%.*]] = cir.vec.splat [[IMM]] : !u64i, !cir.vector<1 x !u64i>
+// CIR: {{%.*}} = cir.call_llvm_intrinsic "aarch64.neon.sqshlu" {{%.*}}, [[AMT]] : (!cir.vector<1 x !u64i>, !cir.vector<1 x !u64i>) -> !cir.vector<1 x !u64i>
+
+// LLVM-SAME: <1 x i64> {{.*}}[[A:%.*]])
+// LLVM: [[TMP0:%.*]] = bitcast <1 x i64> [[A]] to <8 x i8>
+// LLVM: [[VQSHLU_N:%.*]] = bitcast <8 x i8> [[TMP0]] to <1 x i64>
+// LLVM: [[VQSHLU_N1:%.*]] = call <1 x i64> @llvm.aarch64.neon.sqshlu.v1i64(<1 x i64> [[VQSHLU_N]], <1 x i64> splat (i64 1))
+// LLVM: ret <1 x i64> [[VQSHLU_N1]]
+  return vqshlu_n_s64(a, 1);
+}
+
 //===------------------------------------------------------===//
 // 2.1.7.1 Negate
 // https://arm-software.github.io/acle/neon_intrinsics/advsimd.html#negate



More information about the cfe-commits mailing list