https://github.com/iamvickynguyen updated 
https://github.com/llvm/llvm-project/pull/196776

>From bff26630dc8da45f62bdbea1772e8d532ac1e950 Mon Sep 17 00:00:00 2001
From: Vicky Nguyen <[email protected]>
Date: Mon, 4 May 2026 22:23:49 -0700
Subject: [PATCH 1/2] [CIR][AArch64] Upstream vector-shift-right-and-insert
 NEON builtins

---
 .../lib/CIR/CodeGen/CIRGenBuiltinAArch64.cpp  |  23 +-
 clang/test/CodeGen/AArch64/neon-intrinsics.c  | 281 ---------------
 clang/test/CodeGen/AArch64/neon/intrinsics.c  | 339 ++++++++++++++++++
 clang/test/CodeGen/AArch64/poly64.c           |  28 --
 4 files changed, 361 insertions(+), 310 deletions(-)

diff --git a/clang/lib/CIR/CodeGen/CIRGenBuiltinAArch64.cpp 
b/clang/lib/CIR/CodeGen/CIRGenBuiltinAArch64.cpp
index 16018d9b080ad..865cf86589443 100644
--- a/clang/lib/CIR/CodeGen/CIRGenBuiltinAArch64.cpp
+++ b/clang/lib/CIR/CodeGen/CIRGenBuiltinAArch64.cpp
@@ -346,6 +346,22 @@ static mlir::Value emitCommonNeonSISDBuiltinExpr(
     return emitNeonCall(cgf.cgm, cgf.getBuilder(),
                         {cgf.convertType(expr->getArg(0)->getType())}, ops,
                         llvmIntrName, cgf.convertType(expr->getType()), loc);
+  case NEON::BI__builtin_neon_vsrid_n_s64:
+  case NEON::BI__builtin_neon_vsrid_n_u64: {
+    CIRGenBuilderTy &B = cgf.getBuilder();
+    mlir::Type scalarTy = cgf.convertType(expr->getType());
+    auto v1Ty = cir::VectorType::get(scalarTy, /*size=*/1);
+    mlir::Value a = B.createBitcast(ops[0], v1Ty);
+    mlir::Value b = B.createBitcast(ops[1], v1Ty);
+    llvm::SmallVector<mlir::Value, 4> vsriArgs{
+        a, b, B.createIntCast(ops[2], B.getUInt32Ty())};
+    mlir::Value r =
+        emitNeonCall(cgf.cgm, B,
+                     /*argTypes=*/{v1Ty, v1Ty, B.getUInt32Ty()}, vsriArgs,
+                     /*intrinsicName=*/"aarch64.neon.vsri",
+                     /*funcResTy=*/v1Ty, loc);
+    return B.createBitcast(r, scalarTy);
+  }
   }
 
   return nullptr;
@@ -2801,7 +2817,12 @@ CIRGenFunction::emitAArch64BuiltinExpr(unsigned 
builtinID, const CallExpr *expr,
     return result;
   }
   case NEON::BI__builtin_neon_vsri_n_v:
-  case NEON::BI__builtin_neon_vsriq_n_v:
+  case NEON::BI__builtin_neon_vsriq_n_v: {
+    llvm::SmallVector<mlir::Value, 4> vsriArgs = {
+        ops[0], ops[1], builder.createIntCast(ops[2], builder.getUInt32Ty())};
+    return emitNeonCall(cgm, builder, {ty, ty, builder.getUInt32Ty()}, 
vsriArgs,
+                        "aarch64.neon.vsri", ty, loc);
+  }
   case NEON::BI__builtin_neon_vsli_n_v:
   case NEON::BI__builtin_neon_vsliq_n_v:
   case NEON::BI__builtin_neon_vsra_n_v:
diff --git a/clang/test/CodeGen/AArch64/neon-intrinsics.c 
b/clang/test/CodeGen/AArch64/neon-intrinsics.c
index 5a1cbc492cd85..8a57e12154c17 100644
--- a/clang/test/CodeGen/AArch64/neon-intrinsics.c
+++ b/clang/test/CodeGen/AArch64/neon-intrinsics.c
@@ -5601,234 +5601,6 @@ uint64x2_t test_vsraq_n_u64(uint64x2_t a, uint64x2_t b) 
{
   return vsraq_n_u64(a, b, 3);
 }
 
-// CHECK-LABEL: define dso_local <8 x i8> @test_vsri_n_s8(
-// CHECK-SAME: <8 x i8> noundef [[A:%.*]], <8 x i8> noundef [[B:%.*]]) 
#[[ATTR0]] {
-// CHECK-NEXT:  [[ENTRY:.*:]]
-// CHECK-NEXT:    [[VSRI_N:%.*]] = call <8 x i8> 
@llvm.aarch64.neon.vsri.v8i8(<8 x i8> [[A]], <8 x i8> [[B]], i32 3)
-// CHECK-NEXT:    ret <8 x i8> [[VSRI_N]]
-//
-int8x8_t test_vsri_n_s8(int8x8_t a, int8x8_t b) {
-  return vsri_n_s8(a, b, 3);
-}
-
-// CHECK-LABEL: define dso_local <4 x i16> @test_vsri_n_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:    [[VSRI_N:%.*]] = bitcast <8 x i8> [[TMP0]] to <4 x i16>
-// CHECK-NEXT:    [[VSRI_N1:%.*]] = bitcast <8 x i8> [[TMP1]] to <4 x i16>
-// CHECK-NEXT:    [[VSRI_N2:%.*]] = call <4 x i16> 
@llvm.aarch64.neon.vsri.v4i16(<4 x i16> [[VSRI_N]], <4 x i16> [[VSRI_N1]], i32 
3)
-// CHECK-NEXT:    ret <4 x i16> [[VSRI_N2]]
-//
-int16x4_t test_vsri_n_s16(int16x4_t a, int16x4_t b) {
-  return vsri_n_s16(a, b, 3);
-}
-
-// CHECK-LABEL: define dso_local <2 x i32> @test_vsri_n_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:    [[VSRI_N:%.*]] = bitcast <8 x i8> [[TMP0]] to <2 x i32>
-// CHECK-NEXT:    [[VSRI_N1:%.*]] = bitcast <8 x i8> [[TMP1]] to <2 x i32>
-// CHECK-NEXT:    [[VSRI_N2:%.*]] = call <2 x i32> 
@llvm.aarch64.neon.vsri.v2i32(<2 x i32> [[VSRI_N]], <2 x i32> [[VSRI_N1]], i32 
3)
-// CHECK-NEXT:    ret <2 x i32> [[VSRI_N2]]
-//
-int32x2_t test_vsri_n_s32(int32x2_t a, int32x2_t b) {
-  return vsri_n_s32(a, b, 3);
-}
-
-// CHECK-LABEL: define dso_local <16 x i8> @test_vsriq_n_s8(
-// CHECK-SAME: <16 x i8> noundef [[A:%.*]], <16 x i8> noundef [[B:%.*]]) 
#[[ATTR0]] {
-// CHECK-NEXT:  [[ENTRY:.*:]]
-// CHECK-NEXT:    [[VSRI_N:%.*]] = call <16 x i8> 
@llvm.aarch64.neon.vsri.v16i8(<16 x i8> [[A]], <16 x i8> [[B]], i32 3)
-// CHECK-NEXT:    ret <16 x i8> [[VSRI_N]]
-//
-int8x16_t test_vsriq_n_s8(int8x16_t a, int8x16_t b) {
-  return vsriq_n_s8(a, b, 3);
-}
-
-// CHECK-LABEL: define dso_local <8 x i16> @test_vsriq_n_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:    [[VSRI_N:%.*]] = bitcast <16 x i8> [[TMP0]] to <8 x i16>
-// CHECK-NEXT:    [[VSRI_N1:%.*]] = bitcast <16 x i8> [[TMP1]] to <8 x i16>
-// CHECK-NEXT:    [[VSRI_N2:%.*]] = call <8 x i16> 
@llvm.aarch64.neon.vsri.v8i16(<8 x i16> [[VSRI_N]], <8 x i16> [[VSRI_N1]], i32 
3)
-// CHECK-NEXT:    ret <8 x i16> [[VSRI_N2]]
-//
-int16x8_t test_vsriq_n_s16(int16x8_t a, int16x8_t b) {
-  return vsriq_n_s16(a, b, 3);
-}
-
-// CHECK-LABEL: define dso_local <4 x i32> @test_vsriq_n_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:    [[VSRI_N:%.*]] = bitcast <16 x i8> [[TMP0]] to <4 x i32>
-// CHECK-NEXT:    [[VSRI_N1:%.*]] = bitcast <16 x i8> [[TMP1]] to <4 x i32>
-// CHECK-NEXT:    [[VSRI_N2:%.*]] = call <4 x i32> 
@llvm.aarch64.neon.vsri.v4i32(<4 x i32> [[VSRI_N]], <4 x i32> [[VSRI_N1]], i32 
3)
-// CHECK-NEXT:    ret <4 x i32> [[VSRI_N2]]
-//
-int32x4_t test_vsriq_n_s32(int32x4_t a, int32x4_t b) {
-  return vsriq_n_s32(a, b, 3);
-}
-
-// CHECK-LABEL: define dso_local <2 x i64> @test_vsriq_n_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:    [[VSRI_N:%.*]] = bitcast <16 x i8> [[TMP0]] to <2 x i64>
-// CHECK-NEXT:    [[VSRI_N1:%.*]] = bitcast <16 x i8> [[TMP1]] to <2 x i64>
-// CHECK-NEXT:    [[VSRI_N2:%.*]] = call <2 x i64> 
@llvm.aarch64.neon.vsri.v2i64(<2 x i64> [[VSRI_N]], <2 x i64> [[VSRI_N1]], i32 
3)
-// CHECK-NEXT:    ret <2 x i64> [[VSRI_N2]]
-//
-int64x2_t test_vsriq_n_s64(int64x2_t a, int64x2_t b) {
-  return vsriq_n_s64(a, b, 3);
-}
-
-// CHECK-LABEL: define dso_local <8 x i8> @test_vsri_n_u8(
-// CHECK-SAME: <8 x i8> noundef [[A:%.*]], <8 x i8> noundef [[B:%.*]]) 
#[[ATTR0]] {
-// CHECK-NEXT:  [[ENTRY:.*:]]
-// CHECK-NEXT:    [[VSRI_N:%.*]] = call <8 x i8> 
@llvm.aarch64.neon.vsri.v8i8(<8 x i8> [[A]], <8 x i8> [[B]], i32 3)
-// CHECK-NEXT:    ret <8 x i8> [[VSRI_N]]
-//
-uint8x8_t test_vsri_n_u8(uint8x8_t a, uint8x8_t b) {
-  return vsri_n_u8(a, b, 3);
-}
-
-// CHECK-LABEL: define dso_local <4 x i16> @test_vsri_n_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:    [[VSRI_N:%.*]] = bitcast <8 x i8> [[TMP0]] to <4 x i16>
-// CHECK-NEXT:    [[VSRI_N1:%.*]] = bitcast <8 x i8> [[TMP1]] to <4 x i16>
-// CHECK-NEXT:    [[VSRI_N2:%.*]] = call <4 x i16> 
@llvm.aarch64.neon.vsri.v4i16(<4 x i16> [[VSRI_N]], <4 x i16> [[VSRI_N1]], i32 
3)
-// CHECK-NEXT:    ret <4 x i16> [[VSRI_N2]]
-//
-uint16x4_t test_vsri_n_u16(uint16x4_t a, uint16x4_t b) {
-  return vsri_n_u16(a, b, 3);
-}
-
-// CHECK-LABEL: define dso_local <2 x i32> @test_vsri_n_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:    [[VSRI_N:%.*]] = bitcast <8 x i8> [[TMP0]] to <2 x i32>
-// CHECK-NEXT:    [[VSRI_N1:%.*]] = bitcast <8 x i8> [[TMP1]] to <2 x i32>
-// CHECK-NEXT:    [[VSRI_N2:%.*]] = call <2 x i32> 
@llvm.aarch64.neon.vsri.v2i32(<2 x i32> [[VSRI_N]], <2 x i32> [[VSRI_N1]], i32 
3)
-// CHECK-NEXT:    ret <2 x i32> [[VSRI_N2]]
-//
-uint32x2_t test_vsri_n_u32(uint32x2_t a, uint32x2_t b) {
-  return vsri_n_u32(a, b, 3);
-}
-
-// CHECK-LABEL: define dso_local <16 x i8> @test_vsriq_n_u8(
-// CHECK-SAME: <16 x i8> noundef [[A:%.*]], <16 x i8> noundef [[B:%.*]]) 
#[[ATTR0]] {
-// CHECK-NEXT:  [[ENTRY:.*:]]
-// CHECK-NEXT:    [[VSRI_N:%.*]] = call <16 x i8> 
@llvm.aarch64.neon.vsri.v16i8(<16 x i8> [[A]], <16 x i8> [[B]], i32 3)
-// CHECK-NEXT:    ret <16 x i8> [[VSRI_N]]
-//
-uint8x16_t test_vsriq_n_u8(uint8x16_t a, uint8x16_t b) {
-  return vsriq_n_u8(a, b, 3);
-}
-
-// CHECK-LABEL: define dso_local <8 x i16> @test_vsriq_n_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:    [[VSRI_N:%.*]] = bitcast <16 x i8> [[TMP0]] to <8 x i16>
-// CHECK-NEXT:    [[VSRI_N1:%.*]] = bitcast <16 x i8> [[TMP1]] to <8 x i16>
-// CHECK-NEXT:    [[VSRI_N2:%.*]] = call <8 x i16> 
@llvm.aarch64.neon.vsri.v8i16(<8 x i16> [[VSRI_N]], <8 x i16> [[VSRI_N1]], i32 
3)
-// CHECK-NEXT:    ret <8 x i16> [[VSRI_N2]]
-//
-uint16x8_t test_vsriq_n_u16(uint16x8_t a, uint16x8_t b) {
-  return vsriq_n_u16(a, b, 3);
-}
-
-// CHECK-LABEL: define dso_local <4 x i32> @test_vsriq_n_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:    [[VSRI_N:%.*]] = bitcast <16 x i8> [[TMP0]] to <4 x i32>
-// CHECK-NEXT:    [[VSRI_N1:%.*]] = bitcast <16 x i8> [[TMP1]] to <4 x i32>
-// CHECK-NEXT:    [[VSRI_N2:%.*]] = call <4 x i32> 
@llvm.aarch64.neon.vsri.v4i32(<4 x i32> [[VSRI_N]], <4 x i32> [[VSRI_N1]], i32 
3)
-// CHECK-NEXT:    ret <4 x i32> [[VSRI_N2]]
-//
-uint32x4_t test_vsriq_n_u32(uint32x4_t a, uint32x4_t b) {
-  return vsriq_n_u32(a, b, 3);
-}
-
-// CHECK-LABEL: define dso_local <2 x i64> @test_vsriq_n_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:    [[VSRI_N:%.*]] = bitcast <16 x i8> [[TMP0]] to <2 x i64>
-// CHECK-NEXT:    [[VSRI_N1:%.*]] = bitcast <16 x i8> [[TMP1]] to <2 x i64>
-// CHECK-NEXT:    [[VSRI_N2:%.*]] = call <2 x i64> 
@llvm.aarch64.neon.vsri.v2i64(<2 x i64> [[VSRI_N]], <2 x i64> [[VSRI_N1]], i32 
3)
-// CHECK-NEXT:    ret <2 x i64> [[VSRI_N2]]
-//
-uint64x2_t test_vsriq_n_u64(uint64x2_t a, uint64x2_t b) {
-  return vsriq_n_u64(a, b, 3);
-}
-
-// CHECK-LABEL: define dso_local <8 x i8> @test_vsri_n_p8(
-// CHECK-SAME: <8 x i8> noundef [[A:%.*]], <8 x i8> noundef [[B:%.*]]) 
#[[ATTR0]] {
-// CHECK-NEXT:  [[ENTRY:.*:]]
-// CHECK-NEXT:    [[VSRI_N:%.*]] = call <8 x i8> 
@llvm.aarch64.neon.vsri.v8i8(<8 x i8> [[A]], <8 x i8> [[B]], i32 3)
-// CHECK-NEXT:    ret <8 x i8> [[VSRI_N]]
-//
-poly8x8_t test_vsri_n_p8(poly8x8_t a, poly8x8_t b) {
-  return vsri_n_p8(a, b, 3);
-}
-
-// CHECK-LABEL: define dso_local <4 x i16> @test_vsri_n_p16(
-// 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:    [[VSRI_N:%.*]] = bitcast <8 x i8> [[TMP0]] to <4 x i16>
-// CHECK-NEXT:    [[VSRI_N1:%.*]] = bitcast <8 x i8> [[TMP1]] to <4 x i16>
-// CHECK-NEXT:    [[VSRI_N2:%.*]] = call <4 x i16> 
@llvm.aarch64.neon.vsri.v4i16(<4 x i16> [[VSRI_N]], <4 x i16> [[VSRI_N1]], i32 
15)
-// CHECK-NEXT:    ret <4 x i16> [[VSRI_N2]]
-//
-poly16x4_t test_vsri_n_p16(poly16x4_t a, poly16x4_t b) {
-  return vsri_n_p16(a, b, 15);
-}
-
-// CHECK-LABEL: define dso_local <16 x i8> @test_vsriq_n_p8(
-// CHECK-SAME: <16 x i8> noundef [[A:%.*]], <16 x i8> noundef [[B:%.*]]) 
#[[ATTR0]] {
-// CHECK-NEXT:  [[ENTRY:.*:]]
-// CHECK-NEXT:    [[VSRI_N:%.*]] = call <16 x i8> 
@llvm.aarch64.neon.vsri.v16i8(<16 x i8> [[A]], <16 x i8> [[B]], i32 3)
-// CHECK-NEXT:    ret <16 x i8> [[VSRI_N]]
-//
-poly8x16_t test_vsriq_n_p8(poly8x16_t a, poly8x16_t b) {
-  return vsriq_n_p8(a, b, 3);
-}
-
-// CHECK-LABEL: define dso_local <8 x i16> @test_vsriq_n_p16(
-// 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:    [[VSRI_N:%.*]] = bitcast <16 x i8> [[TMP0]] to <8 x i16>
-// CHECK-NEXT:    [[VSRI_N1:%.*]] = bitcast <16 x i8> [[TMP1]] to <8 x i16>
-// CHECK-NEXT:    [[VSRI_N2:%.*]] = call <8 x i16> 
@llvm.aarch64.neon.vsri.v8i16(<8 x i16> [[VSRI_N]], <8 x i16> [[VSRI_N1]], i32 
15)
-// CHECK-NEXT:    ret <8 x i16> [[VSRI_N2]]
-//
-poly16x8_t test_vsriq_n_p16(poly16x8_t a, poly16x8_t b) {
-  return vsriq_n_p16(a, b, 15);
-}
-
 // CHECK-LABEL: define dso_local <8 x i8> @test_vsli_n_s8(
 // CHECK-SAME: <8 x i8> noundef [[A:%.*]], <8 x i8> noundef [[B:%.*]]) 
#[[ATTR0]] {
 // CHECK-NEXT:  [[ENTRY:.*:]]
@@ -15763,59 +15535,6 @@ uint64x1_t test_vqshlu_n_s64(int64x1_t a) {
   return vqshlu_n_s64(a, 1);
 }
 
-// CHECK-LABEL: define dso_local i64 @test_vsrid_n_s64(
-// CHECK-SAME: i64 noundef [[A:%.*]], i64 noundef [[B:%.*]]) #[[ATTR0]] {
-// CHECK-NEXT:  [[ENTRY:.*:]]
-// CHECK-NEXT:    [[VSRID_N_S64:%.*]] = bitcast i64 [[A]] to <1 x i64>
-// CHECK-NEXT:    [[VSRID_N_S641:%.*]] = bitcast i64 [[B]] to <1 x i64>
-// CHECK-NEXT:    [[VSRID_N_S642:%.*]] = call <1 x i64> 
@llvm.aarch64.neon.vsri.v1i64(<1 x i64> [[VSRID_N_S64]], <1 x i64> 
[[VSRID_N_S641]], i32 63)
-// CHECK-NEXT:    [[VSRID_N_S643:%.*]] = bitcast <1 x i64> [[VSRID_N_S642]] to 
i64
-// CHECK-NEXT:    ret i64 [[VSRID_N_S643]]
-//
-int64_t test_vsrid_n_s64(int64_t a, int64_t b) {
-  return (int64_t)vsrid_n_s64(a, b, 63);
-}
-
-// CHECK-LABEL: define dso_local <1 x i64> @test_vsri_n_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:    [[VSRI_N:%.*]] = bitcast <8 x i8> [[TMP0]] to <1 x i64>
-// CHECK-NEXT:    [[VSRI_N1:%.*]] = bitcast <8 x i8> [[TMP1]] to <1 x i64>
-// CHECK-NEXT:    [[VSRI_N2:%.*]] = call <1 x i64> 
@llvm.aarch64.neon.vsri.v1i64(<1 x i64> [[VSRI_N]], <1 x i64> [[VSRI_N1]], i32 
1)
-// CHECK-NEXT:    ret <1 x i64> [[VSRI_N2]]
-//
-int64x1_t test_vsri_n_s64(int64x1_t a, int64x1_t b) {
-  return vsri_n_s64(a, b, 1);
-}
-
-// CHECK-LABEL: define dso_local i64 @test_vsrid_n_u64(
-// CHECK-SAME: i64 noundef [[A:%.*]], i64 noundef [[B:%.*]]) #[[ATTR0]] {
-// CHECK-NEXT:  [[ENTRY:.*:]]
-// CHECK-NEXT:    [[VSRID_N_U64:%.*]] = bitcast i64 [[A]] to <1 x i64>
-// CHECK-NEXT:    [[VSRID_N_U641:%.*]] = bitcast i64 [[B]] to <1 x i64>
-// CHECK-NEXT:    [[VSRID_N_U642:%.*]] = call <1 x i64> 
@llvm.aarch64.neon.vsri.v1i64(<1 x i64> [[VSRID_N_U64]], <1 x i64> 
[[VSRID_N_U641]], i32 63)
-// CHECK-NEXT:    [[VSRID_N_U643:%.*]] = bitcast <1 x i64> [[VSRID_N_U642]] to 
i64
-// CHECK-NEXT:    ret i64 [[VSRID_N_U643]]
-//
-uint64_t test_vsrid_n_u64(uint64_t a, uint64_t b) {
-  return (uint64_t)vsrid_n_u64(a, b, 63);
-}
-
-// CHECK-LABEL: define dso_local <1 x i64> @test_vsri_n_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:    [[VSRI_N:%.*]] = bitcast <8 x i8> [[TMP0]] to <1 x i64>
-// CHECK-NEXT:    [[VSRI_N1:%.*]] = bitcast <8 x i8> [[TMP1]] to <1 x i64>
-// CHECK-NEXT:    [[VSRI_N2:%.*]] = call <1 x i64> 
@llvm.aarch64.neon.vsri.v1i64(<1 x i64> [[VSRI_N]], <1 x i64> [[VSRI_N1]], i32 
1)
-// CHECK-NEXT:    ret <1 x i64> [[VSRI_N2]]
-//
-uint64x1_t test_vsri_n_u64(uint64x1_t a, uint64x1_t b) {
-  return vsri_n_u64(a, b, 1);
-}
 
 // CHECK-LABEL: define dso_local i64 @test_vslid_n_s64(
 // CHECK-SAME: i64 noundef [[A:%.*]], i64 noundef [[B:%.*]]) #[[ATTR0]] {
diff --git a/clang/test/CodeGen/AArch64/neon/intrinsics.c 
b/clang/test/CodeGen/AArch64/neon/intrinsics.c
index 7e71ea4c00422..26cea888f84d2 100644
--- a/clang/test/CodeGen/AArch64/neon/intrinsics.c
+++ b/clang/test/CodeGen/AArch64/neon/intrinsics.c
@@ -4842,3 +4842,342 @@ uint64_t test_vaddlvq_u32(uint32x4_t a) {
 // LLVM-NEXT:    ret i64 [[VADDLVQ_U32_I]]
   return vaddlvq_u32(a);
 }
+
+//===------------------------------------------------------===//
+// 2.1.3.2.9 Vector Shift Right and Insert
+// 
https://arm-software.github.io/acle/neon_intrinsics/advsimd.html#vector-shift-right-and-insert
+//===------------------------------------------------------===//
+
+// LLVM-LABEL: @test_vsri_n_s8(
+// CIR-LABEL: @test_vsri_n_s8(
+int8x8_t test_vsri_n_s8(int8x8_t a, int8x8_t b) {
+// CIR: cir.call_llvm_intrinsic "aarch64.neon.vsri"
+
+// LLVM-SAME: <8 x i8> {{.*}}[[A:%.*]], <8 x i8> {{.*}}[[B:%.*]]) {{.*}} {
+// LLVM:    [[VSRI_N:%.*]] = call <8 x i8> @llvm.aarch64.neon.vsri.v8i8(<8 x 
i8> [[A]], <8 x i8> [[B]], i32 3)
+// LLVM:    ret <8 x i8> [[VSRI_N]]
+  return vsri_n_s8(a, b, 3);
+}
+
+// LLVM-LABEL: @test_vsri_n_s16(
+// CIR-LABEL: @test_vsri_n_s16(
+int16x4_t test_vsri_n_s16(int16x4_t a, int16x4_t b) {
+// CIR: cir.call_llvm_intrinsic "aarch64.neon.vsri"
+
+// LLVM-SAME: <4 x i16> {{.*}}[[A:%.*]], <4 x i16> {{.*}}[[B:%.*]]) {{.*}} {
+// LLVM:    [[TMP0:%.*]] = bitcast <4 x i16> [[A]] to <8 x i8>
+// LLVM-NEXT:    [[TMP1:%.*]] = bitcast <4 x i16> [[B]] to <8 x i8>
+// LLVM-NEXT:    [[VSRI_N:%.*]] = bitcast <8 x i8> [[TMP0]] to <4 x i16>
+// LLVM-NEXT:    [[VSRI_N1:%.*]] = bitcast <8 x i8> [[TMP1]] to <4 x i16>
+// LLVM-NEXT:    [[VSRI_N2:%.*]] = call <4 x i16> 
@llvm.aarch64.neon.vsri.v4i16(<4 x i16> [[VSRI_N]], <4 x i16> [[VSRI_N1]], i32 
3)
+// LLVM:    ret <4 x i16> [[VSRI_N2]]
+  return vsri_n_s16(a, b, 3);
+}
+
+// LLVM-LABEL: @test_vsri_n_s32(
+// CIR-LABEL: @test_vsri_n_s32(
+int32x2_t test_vsri_n_s32(int32x2_t a, int32x2_t b) {
+// CIR: cir.call_llvm_intrinsic "aarch64.neon.vsri"
+
+// LLVM-SAME: <2 x i32> {{.*}}[[A:%.*]], <2 x i32> {{.*}}[[B:%.*]]) {{.*}} {
+// LLVM:    [[TMP0:%.*]] = bitcast <2 x i32> [[A]] to <8 x i8>
+// LLVM-NEXT:    [[TMP1:%.*]] = bitcast <2 x i32> [[B]] to <8 x i8>
+// LLVM-NEXT:    [[VSRI_N:%.*]] = bitcast <8 x i8> [[TMP0]] to <2 x i32>
+// LLVM-NEXT:    [[VSRI_N1:%.*]] = bitcast <8 x i8> [[TMP1]] to <2 x i32>
+// LLVM-NEXT:    [[VSRI_N2:%.*]] = call <2 x i32> 
@llvm.aarch64.neon.vsri.v2i32(<2 x i32> [[VSRI_N]], <2 x i32> [[VSRI_N1]], i32 
3)
+// LLVM:    ret <2 x i32> [[VSRI_N2]]
+  return vsri_n_s32(a, b, 3);
+}
+
+// LLVM-LABEL: @test_vsriq_n_s8(
+// CIR-LABEL: @test_vsriq_n_s8(
+int8x16_t test_vsriq_n_s8(int8x16_t a, int8x16_t b) {
+// CIR: cir.call_llvm_intrinsic "aarch64.neon.vsri"
+
+// LLVM-SAME: <16 x i8> {{.*}}[[A:%.*]], <16 x i8> {{.*}}[[B:%.*]]) {{.*}} {
+// LLVM:    [[VSRI_N:%.*]] = call <16 x i8> @llvm.aarch64.neon.vsri.v16i8(<16 
x i8> [[A]], <16 x i8> [[B]], i32 3)
+// LLVM:    ret <16 x i8> [[VSRI_N]]
+  return vsriq_n_s8(a, b, 3);
+}
+
+// LLVM-LABEL: @test_vsriq_n_s16(
+// CIR-LABEL: @test_vsriq_n_s16(
+int16x8_t test_vsriq_n_s16(int16x8_t a, int16x8_t b) {
+// CIR: cir.call_llvm_intrinsic "aarch64.neon.vsri"
+
+// LLVM-SAME: <8 x i16> {{.*}}[[A:%.*]], <8 x i16> {{.*}}[[B:%.*]]) {{.*}} {
+// LLVM:    [[TMP0:%.*]] = bitcast <8 x i16> [[A]] to <16 x i8>
+// LLVM-NEXT:    [[TMP1:%.*]] = bitcast <8 x i16> [[B]] to <16 x i8>
+// LLVM-NEXT:    [[VSRI_N:%.*]] = bitcast <16 x i8> [[TMP0]] to <8 x i16>
+// LLVM-NEXT:    [[VSRI_N1:%.*]] = bitcast <16 x i8> [[TMP1]] to <8 x i16>
+// LLVM-NEXT:    [[VSRI_N2:%.*]] = call <8 x i16> 
@llvm.aarch64.neon.vsri.v8i16(<8 x i16> [[VSRI_N]], <8 x i16> [[VSRI_N1]], i32 
3)
+// LLVM:    ret <8 x i16> [[VSRI_N2]]
+  return vsriq_n_s16(a, b, 3);
+}
+
+// LLVM-LABEL: @test_vsriq_n_s32(
+// CIR-LABEL: @test_vsriq_n_s32(
+int32x4_t test_vsriq_n_s32(int32x4_t a, int32x4_t b) {
+// CIR: cir.call_llvm_intrinsic "aarch64.neon.vsri"
+
+// LLVM-SAME: <4 x i32> {{.*}}[[A:%.*]], <4 x i32> {{.*}}[[B:%.*]]) {{.*}} {
+// LLVM:    [[TMP0:%.*]] = bitcast <4 x i32> [[A]] to <16 x i8>
+// LLVM-NEXT:    [[TMP1:%.*]] = bitcast <4 x i32> [[B]] to <16 x i8>
+// LLVM-NEXT:    [[VSRI_N:%.*]] = bitcast <16 x i8> [[TMP0]] to <4 x i32>
+// LLVM-NEXT:    [[VSRI_N1:%.*]] = bitcast <16 x i8> [[TMP1]] to <4 x i32>
+// LLVM-NEXT:    [[VSRI_N2:%.*]] = call <4 x i32> 
@llvm.aarch64.neon.vsri.v4i32(<4 x i32> [[VSRI_N]], <4 x i32> [[VSRI_N1]], i32 
3)
+// LLVM:    ret <4 x i32> [[VSRI_N2]]
+  return vsriq_n_s32(a, b, 3);
+}
+
+// LLVM-LABEL: @test_vsriq_n_s64(
+// CIR-LABEL: @test_vsriq_n_s64(
+int64x2_t test_vsriq_n_s64(int64x2_t a, int64x2_t b) {
+// CIR: cir.call_llvm_intrinsic "aarch64.neon.vsri"
+
+// LLVM-SAME: <2 x i64> {{.*}}[[A:%.*]], <2 x i64> {{.*}}[[B:%.*]]) {{.*}} {
+// LLVM:    [[TMP0:%.*]] = bitcast <2 x i64> [[A]] to <16 x i8>
+// LLVM-NEXT:    [[TMP1:%.*]] = bitcast <2 x i64> [[B]] to <16 x i8>
+// LLVM-NEXT:    [[VSRI_N:%.*]] = bitcast <16 x i8> [[TMP0]] to <2 x i64>
+// LLVM-NEXT:    [[VSRI_N1:%.*]] = bitcast <16 x i8> [[TMP1]] to <2 x i64>
+// LLVM-NEXT:    [[VSRI_N2:%.*]] = call <2 x i64> 
@llvm.aarch64.neon.vsri.v2i64(<2 x i64> [[VSRI_N]], <2 x i64> [[VSRI_N1]], i32 
3)
+// LLVM:    ret <2 x i64> [[VSRI_N2]]
+  return vsriq_n_s64(a, b, 3);
+}
+
+// LLVM-LABEL: @test_vsri_n_u8(
+// CIR-LABEL: @test_vsri_n_u8(
+uint8x8_t test_vsri_n_u8(uint8x8_t a, uint8x8_t b) {
+// CIR: cir.call_llvm_intrinsic "aarch64.neon.vsri"
+
+// LLVM-SAME: <8 x i8> {{.*}}[[A:%.*]], <8 x i8> {{.*}}[[B:%.*]]) {{.*}} {
+// LLVM:    [[VSRI_N:%.*]] = call <8 x i8> @llvm.aarch64.neon.vsri.v8i8(<8 x 
i8> [[A]], <8 x i8> [[B]], i32 3)
+// LLVM:    ret <8 x i8> [[VSRI_N]]
+  return vsri_n_u8(a, b, 3);
+}
+
+// LLVM-LABEL: @test_vsri_n_u16(
+// CIR-LABEL: @test_vsri_n_u16(
+uint16x4_t test_vsri_n_u16(uint16x4_t a, uint16x4_t b) {
+// CIR: cir.call_llvm_intrinsic "aarch64.neon.vsri"
+
+// LLVM-SAME: <4 x i16> {{.*}}[[A:%.*]], <4 x i16> {{.*}}[[B:%.*]]) {{.*}} {
+// LLVM:    [[TMP0:%.*]] = bitcast <4 x i16> [[A]] to <8 x i8>
+// LLVM-NEXT:    [[TMP1:%.*]] = bitcast <4 x i16> [[B]] to <8 x i8>
+// LLVM-NEXT:    [[VSRI_N:%.*]] = bitcast <8 x i8> [[TMP0]] to <4 x i16>
+// LLVM-NEXT:    [[VSRI_N1:%.*]] = bitcast <8 x i8> [[TMP1]] to <4 x i16>
+// LLVM-NEXT:    [[VSRI_N2:%.*]] = call <4 x i16> 
@llvm.aarch64.neon.vsri.v4i16(<4 x i16> [[VSRI_N]], <4 x i16> [[VSRI_N1]], i32 
3)
+// LLVM:    ret <4 x i16> [[VSRI_N2]]
+  return vsri_n_u16(a, b, 3);
+}
+
+// LLVM-LABEL: @test_vsri_n_u32(
+// CIR-LABEL: @test_vsri_n_u32(
+uint32x2_t test_vsri_n_u32(uint32x2_t a, uint32x2_t b) {
+// CIR: cir.call_llvm_intrinsic "aarch64.neon.vsri"
+
+// LLVM-SAME: <2 x i32> {{.*}}[[A:%.*]], <2 x i32> {{.*}}[[B:%.*]]) {{.*}} {
+// LLVM:    [[TMP0:%.*]] = bitcast <2 x i32> [[A]] to <8 x i8>
+// LLVM-NEXT:    [[TMP1:%.*]] = bitcast <2 x i32> [[B]] to <8 x i8>
+// LLVM-NEXT:    [[VSRI_N:%.*]] = bitcast <8 x i8> [[TMP0]] to <2 x i32>
+// LLVM-NEXT:    [[VSRI_N1:%.*]] = bitcast <8 x i8> [[TMP1]] to <2 x i32>
+// LLVM-NEXT:    [[VSRI_N2:%.*]] = call <2 x i32> 
@llvm.aarch64.neon.vsri.v2i32(<2 x i32> [[VSRI_N]], <2 x i32> [[VSRI_N1]], i32 
3)
+// LLVM:    ret <2 x i32> [[VSRI_N2]]
+  return vsri_n_u32(a, b, 3);
+}
+
+// LLVM-LABEL: @test_vsriq_n_u8(
+// CIR-LABEL: @test_vsriq_n_u8(
+uint8x16_t test_vsriq_n_u8(uint8x16_t a, uint8x16_t b) {
+// CIR: cir.call_llvm_intrinsic "aarch64.neon.vsri"
+
+// LLVM-SAME: <16 x i8> {{.*}}[[A:%.*]], <16 x i8> {{.*}}[[B:%.*]]) {{.*}} {
+// LLVM:    [[VSRI_N:%.*]] = call <16 x i8> @llvm.aarch64.neon.vsri.v16i8(<16 
x i8> [[A]], <16 x i8> [[B]], i32 3)
+// LLVM:    ret <16 x i8> [[VSRI_N]]
+  return vsriq_n_u8(a, b, 3);
+}
+
+// LLVM-LABEL: @test_vsriq_n_u16(
+// CIR-LABEL: @test_vsriq_n_u16(
+uint16x8_t test_vsriq_n_u16(uint16x8_t a, uint16x8_t b) {
+// CIR: cir.call_llvm_intrinsic "aarch64.neon.vsri"
+
+// LLVM-SAME: <8 x i16> {{.*}}[[A:%.*]], <8 x i16> {{.*}}[[B:%.*]]) {{.*}} {
+// LLVM:    [[TMP0:%.*]] = bitcast <8 x i16> [[A]] to <16 x i8>
+// LLVM-NEXT:    [[TMP1:%.*]] = bitcast <8 x i16> [[B]] to <16 x i8>
+// LLVM-NEXT:    [[VSRI_N:%.*]] = bitcast <16 x i8> [[TMP0]] to <8 x i16>
+// LLVM-NEXT:    [[VSRI_N1:%.*]] = bitcast <16 x i8> [[TMP1]] to <8 x i16>
+// LLVM-NEXT:    [[VSRI_N2:%.*]] = call <8 x i16> 
@llvm.aarch64.neon.vsri.v8i16(<8 x i16> [[VSRI_N]], <8 x i16> [[VSRI_N1]], i32 
3)
+// LLVM:    ret <8 x i16> [[VSRI_N2]]
+  return vsriq_n_u16(a, b, 3);
+}
+
+// LLVM-LABEL: @test_vsriq_n_u32(
+// CIR-LABEL: @test_vsriq_n_u32(
+uint32x4_t test_vsriq_n_u32(uint32x4_t a, uint32x4_t b) {
+// CIR: cir.call_llvm_intrinsic "aarch64.neon.vsri"
+
+// LLVM-SAME: <4 x i32> {{.*}}[[A:%.*]], <4 x i32> {{.*}}[[B:%.*]]) {{.*}} {
+// LLVM:    [[TMP0:%.*]] = bitcast <4 x i32> [[A]] to <16 x i8>
+// LLVM-NEXT:    [[TMP1:%.*]] = bitcast <4 x i32> [[B]] to <16 x i8>
+// LLVM-NEXT:    [[VSRI_N:%.*]] = bitcast <16 x i8> [[TMP0]] to <4 x i32>
+// LLVM-NEXT:    [[VSRI_N1:%.*]] = bitcast <16 x i8> [[TMP1]] to <4 x i32>
+// LLVM-NEXT:    [[VSRI_N2:%.*]] = call <4 x i32> 
@llvm.aarch64.neon.vsri.v4i32(<4 x i32> [[VSRI_N]], <4 x i32> [[VSRI_N1]], i32 
3)
+// LLVM:    ret <4 x i32> [[VSRI_N2]]
+  return vsriq_n_u32(a, b, 3);
+}
+
+// LLVM-LABEL: @test_vsriq_n_u64(
+// CIR-LABEL: @test_vsriq_n_u64(
+uint64x2_t test_vsriq_n_u64(uint64x2_t a, uint64x2_t b) {
+// CIR: cir.call_llvm_intrinsic "aarch64.neon.vsri"
+
+// LLVM-SAME: <2 x i64> {{.*}}[[A:%.*]], <2 x i64> {{.*}}[[B:%.*]]) {{.*}} {
+// LLVM:    [[TMP0:%.*]] = bitcast <2 x i64> [[A]] to <16 x i8>
+// LLVM-NEXT:    [[TMP1:%.*]] = bitcast <2 x i64> [[B]] to <16 x i8>
+// LLVM-NEXT:    [[VSRI_N:%.*]] = bitcast <16 x i8> [[TMP0]] to <2 x i64>
+// LLVM-NEXT:    [[VSRI_N1:%.*]] = bitcast <16 x i8> [[TMP1]] to <2 x i64>
+// LLVM-NEXT:    [[VSRI_N2:%.*]] = call <2 x i64> 
@llvm.aarch64.neon.vsri.v2i64(<2 x i64> [[VSRI_N]], <2 x i64> [[VSRI_N1]], i32 
3)
+// LLVM:    ret <2 x i64> [[VSRI_N2]]
+  return vsriq_n_u64(a, b, 3);
+}
+
+// LLVM-LABEL: @test_vsri_n_p8(
+// CIR-LABEL: @test_vsri_n_p8(
+poly8x8_t test_vsri_n_p8(poly8x8_t a, poly8x8_t b) {
+// CIR: cir.call_llvm_intrinsic "aarch64.neon.vsri"
+
+// LLVM-SAME: <8 x i8> {{.*}}[[A:%.*]], <8 x i8> {{.*}}[[B:%.*]]) {{.*}} {
+// LLVM:    [[VSRI_N:%.*]] = call <8 x i8> @llvm.aarch64.neon.vsri.v8i8(<8 x 
i8> [[A]], <8 x i8> [[B]], i32 3)
+// LLVM:    ret <8 x i8> [[VSRI_N]]
+  return vsri_n_p8(a, b, 3);
+}
+
+// LLVM-LABEL: @test_vsri_n_p16(
+// CIR-LABEL: @test_vsri_n_p16(
+poly16x4_t test_vsri_n_p16(poly16x4_t a, poly16x4_t b) {
+// CIR: cir.call_llvm_intrinsic "aarch64.neon.vsri"
+
+// LLVM-SAME: <4 x i16> {{.*}}[[A:%.*]], <4 x i16> {{.*}}[[B:%.*]]) {{.*}} {
+// LLVM:    [[TMP0:%.*]] = bitcast <4 x i16> [[A]] to <8 x i8>
+// LLVM-NEXT:    [[TMP1:%.*]] = bitcast <4 x i16> [[B]] to <8 x i8>
+// LLVM-NEXT:    [[VSRI_N:%.*]] = bitcast <8 x i8> [[TMP0]] to <4 x i16>
+// LLVM-NEXT:    [[VSRI_N1:%.*]] = bitcast <8 x i8> [[TMP1]] to <4 x i16>
+// LLVM-NEXT:    [[VSRI_N2:%.*]] = call <4 x i16> 
@llvm.aarch64.neon.vsri.v4i16(<4 x i16> [[VSRI_N]], <4 x i16> [[VSRI_N1]], i32 
15)
+// LLVM:    ret <4 x i16> [[VSRI_N2]]
+  return vsri_n_p16(a, b, 15);
+}
+
+// LLVM-LABEL: @test_vsriq_n_p8(
+// CIR-LABEL: @test_vsriq_n_p8(
+poly8x16_t test_vsriq_n_p8(poly8x16_t a, poly8x16_t b) {
+// CIR: cir.call_llvm_intrinsic "aarch64.neon.vsri"
+
+// LLVM-SAME: <16 x i8> {{.*}}[[A:%.*]], <16 x i8> {{.*}}[[B:%.*]]) {{.*}} {
+// LLVM:    [[VSRI_N:%.*]] = call <16 x i8> @llvm.aarch64.neon.vsri.v16i8(<16 
x i8> [[A]], <16 x i8> [[B]], i32 3)
+// LLVM:    ret <16 x i8> [[VSRI_N]]
+  return vsriq_n_p8(a, b, 3);
+}
+
+// LLVM-LABEL: @test_vsriq_n_p16(
+// CIR-LABEL: @test_vsriq_n_p16(
+poly16x8_t test_vsriq_n_p16(poly16x8_t a, poly16x8_t b) {
+// CIR: cir.call_llvm_intrinsic "aarch64.neon.vsri"
+
+// LLVM-SAME: <8 x i16> {{.*}}[[A:%.*]], <8 x i16> {{.*}}[[B:%.*]]) {{.*}} {
+// LLVM:    [[TMP0:%.*]] = bitcast <8 x i16> [[A]] to <16 x i8>
+// LLVM-NEXT:    [[TMP1:%.*]] = bitcast <8 x i16> [[B]] to <16 x i8>
+// LLVM-NEXT:    [[VSRI_N:%.*]] = bitcast <16 x i8> [[TMP0]] to <8 x i16>
+// LLVM-NEXT:    [[VSRI_N1:%.*]] = bitcast <16 x i8> [[TMP1]] to <8 x i16>
+// LLVM-NEXT:    [[VSRI_N2:%.*]] = call <8 x i16> 
@llvm.aarch64.neon.vsri.v8i16(<8 x i16> [[VSRI_N]], <8 x i16> [[VSRI_N1]], i32 
15)
+// LLVM:    ret <8 x i16> [[VSRI_N2]]
+  return vsriq_n_p16(a, b, 15);
+}
+
+// LLVM-LABEL: @test_vsrid_n_s64(
+// CIR-LABEL: @test_vsrid_n_s64(
+int64_t test_vsrid_n_s64(int64_t a, int64_t b) {
+// CIR: cir.call_llvm_intrinsic "aarch64.neon.vsri"
+
+// LLVM-SAME: i64 {{.*}}[[A:%.*]], i64 {{.*}}[[B:%.*]]) {{.*}} {
+// LLVM:    [[T0:%.*]] = bitcast i64 [[A]] to <1 x i64>
+// LLVM-NEXT:    [[T1:%.*]] = bitcast i64 [[B]] to <1 x i64>
+// LLVM-NEXT:    [[T2:%.*]] = call <1 x i64> @llvm.aarch64.neon.vsri.v1i64(<1 
x i64> [[T0]], <1 x i64> [[T1]], i32 63)
+// LLVM-NEXT:    [[T3:%.*]] = bitcast <1 x i64> [[T2]] to i64
+// LLVM:    ret i64 [[T3]]
+  return (int64_t)vsrid_n_s64(a, b, 63);
+}
+
+// LLVM-LABEL: @test_vsri_n_s64(
+// CIR-LABEL: @test_vsri_n_s64(
+int64x1_t test_vsri_n_s64(int64x1_t a, int64x1_t b) {
+// CIR: cir.call_llvm_intrinsic "aarch64.neon.vsri"
+
+// LLVM-SAME: <1 x i64> {{.*}}[[A:%.*]], <1 x i64> {{.*}}[[B:%.*]]) {{.*}} {
+// LLVM:    [[TMP0:%.*]] = bitcast <1 x i64> [[A]] to <8 x i8>
+// LLVM-NEXT:    [[TMP1:%.*]] = bitcast <1 x i64> [[B]] to <8 x i8>
+// LLVM-NEXT:    [[VSRI_N:%.*]] = bitcast <8 x i8> [[TMP0]] to <1 x i64>
+// LLVM-NEXT:    [[VSRI_N1:%.*]] = bitcast <8 x i8> [[TMP1]] to <1 x i64>
+// LLVM-NEXT:    [[VSRI_N2:%.*]] = call <1 x i64> 
@llvm.aarch64.neon.vsri.v1i64(<1 x i64> [[VSRI_N]], <1 x i64> [[VSRI_N1]], i32 
1)
+// LLVM:    ret <1 x i64> [[VSRI_N2]]
+  return vsri_n_s64(a, b, 1);
+}
+
+// LLVM-LABEL: @test_vsrid_n_u64(
+// CIR-LABEL: @test_vsrid_n_u64(
+uint64_t test_vsrid_n_u64(uint64_t a, uint64_t b) {
+// CIR: cir.call_llvm_intrinsic "aarch64.neon.vsri"
+
+// LLVM-SAME: i64 {{.*}}[[A:%.*]], i64 {{.*}}[[B:%.*]]) {{.*}} {
+// LLVM:    [[T0:%.*]] = bitcast i64 [[A]] to <1 x i64>
+// LLVM-NEXT:    [[T1:%.*]] = bitcast i64 [[B]] to <1 x i64>
+// LLVM-NEXT:    [[T2:%.*]] = call <1 x i64> @llvm.aarch64.neon.vsri.v1i64(<1 
x i64> [[T0]], <1 x i64> [[T1]], i32 63)
+// LLVM-NEXT:    [[T3:%.*]] = bitcast <1 x i64> [[T2]] to i64
+// LLVM:    ret i64 [[T3]]
+  return (uint64_t)vsrid_n_u64(a, b, 63);
+}
+
+// LLVM-LABEL: @test_vsri_n_u64(
+// CIR-LABEL: @test_vsri_n_u64(
+uint64x1_t test_vsri_n_u64(uint64x1_t a, uint64x1_t b) {
+// CIR: cir.call_llvm_intrinsic "aarch64.neon.vsri"
+
+// LLVM-SAME: <1 x i64> {{.*}}[[A:%.*]], <1 x i64> {{.*}}[[B:%.*]]) {{.*}} {
+// LLVM:    [[TMP0:%.*]] = bitcast <1 x i64> [[A]] to <8 x i8>
+// LLVM-NEXT:    [[TMP1:%.*]] = bitcast <1 x i64> [[B]] to <8 x i8>
+// LLVM-NEXT:    [[VSRI_N:%.*]] = bitcast <8 x i8> [[TMP0]] to <1 x i64>
+// LLVM-NEXT:    [[VSRI_N1:%.*]] = bitcast <8 x i8> [[TMP1]] to <1 x i64>
+// LLVM-NEXT:    [[VSRI_N2:%.*]] = call <1 x i64> 
@llvm.aarch64.neon.vsri.v1i64(<1 x i64> [[VSRI_N]], <1 x i64> [[VSRI_N1]], i32 
1)
+// LLVM:    ret <1 x i64> [[VSRI_N2]]
+  return vsri_n_u64(a, b, 1);
+}
+
+// LLVM-LABEL: @test_vsri_n_p64(
+// CIR-LABEL: @test_vsri_n_p64(
+poly64x1_t test_vsri_n_p64(poly64x1_t a, poly64x1_t b) {
+// CIR: cir.call_llvm_intrinsic "aarch64.neon.vsri"
+
+// LLVM-SAME: <1 x i64> {{.*}}[[A:%.*]], <1 x i64> {{.*}}[[B:%.*]]) #[[ATTR0]] 
{
+// LLVM:    [[TMP0:%.*]] = bitcast <1 x i64> [[A]] to <8 x i8>
+// LLVM-NEXT:    [[TMP1:%.*]] = bitcast <1 x i64> [[B]] to <8 x i8>
+// LLVM-NEXT:    [[VSRI_N:%.*]] = bitcast <8 x i8> [[TMP0]] to <1 x i64>
+// LLVM-NEXT:    [[VSRI_N1:%.*]] = bitcast <8 x i8> [[TMP1]] to <1 x i64>
+// LLVM-NEXT:    [[VSRI_N2:%.*]] = call <1 x i64> 
@llvm.aarch64.neon.vsri.v1i64(<1 x i64> [[VSRI_N]], <1 x i64> [[VSRI_N1]], i32 
33)
+// LLVM:    ret <1 x i64> [[VSRI_N2]]
+  return vsri_n_p64(a, b, 33);
+}
+
+// LLVM-LABEL: @test_vsriq_n_p64(
+// CIR-LABEL: @test_vsriq_n_p64(
+poly64x2_t test_vsriq_n_p64(poly64x2_t a, poly64x2_t b) {
+// CIR: cir.call_llvm_intrinsic "aarch64.neon.vsri"
+
+// LLVM-SAME: <2 x i64> {{.*}}[[A:%.*]], <2 x i64> {{.*}}[[B:%.*]]) #[[ATTR0]] 
{
+// LLVM:    [[TMP0:%.*]] = bitcast <2 x i64> [[A]] to <16 x i8>
+// LLVM-NEXT:    [[TMP1:%.*]] = bitcast <2 x i64> [[B]] to <16 x i8>
+// LLVM-NEXT:    [[VSRI_N:%.*]] = bitcast <16 x i8> [[TMP0]] to <2 x i64>
+// LLVM-NEXT:    [[VSRI_N1:%.*]] = bitcast <16 x i8> [[TMP1]] to <2 x i64>
+// LLVM-NEXT:    [[VSRI_N2:%.*]] = call <2 x i64> 
@llvm.aarch64.neon.vsri.v2i64(<2 x i64> [[VSRI_N]], <2 x i64> [[VSRI_N1]], i32 
64)
+// LLVM:    ret <2 x i64> [[VSRI_N2]]
+  return vsriq_n_p64(a, b, 64);
+}
diff --git a/clang/test/CodeGen/AArch64/poly64.c 
b/clang/test/CodeGen/AArch64/poly64.c
index 5e808f07a6f56..4f5f6c95c2981 100644
--- a/clang/test/CodeGen/AArch64/poly64.c
+++ b/clang/test/CodeGen/AArch64/poly64.c
@@ -554,31 +554,3 @@ poly64x2_t test_vtrn1q_p64(poly64x2_t a, poly64x2_t b) {
 poly64x2_t test_vtrn2q_p64(poly64x2_t a, poly64x2_t b) {
   return vtrn2q_u64(a, b);
 }
-
-// CHECK-LABEL: define dso_local <1 x i64> @test_vsri_n_p64(
-// 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:    [[VSRI_N:%.*]] = bitcast <8 x i8> [[TMP0]] to <1 x i64>
-// CHECK-NEXT:    [[VSRI_N1:%.*]] = bitcast <8 x i8> [[TMP1]] to <1 x i64>
-// CHECK-NEXT:    [[VSRI_N2:%.*]] = call <1 x i64> 
@llvm.aarch64.neon.vsri.v1i64(<1 x i64> [[VSRI_N]], <1 x i64> [[VSRI_N1]], i32 
33)
-// CHECK-NEXT:    ret <1 x i64> [[VSRI_N2]]
-//
-poly64x1_t test_vsri_n_p64(poly64x1_t a, poly64x1_t b) {
-  return vsri_n_p64(a, b, 33);
-}
-
-// CHECK-LABEL: define dso_local <2 x i64> @test_vsriq_n_p64(
-// 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:    [[VSRI_N:%.*]] = bitcast <16 x i8> [[TMP0]] to <2 x i64>
-// CHECK-NEXT:    [[VSRI_N1:%.*]] = bitcast <16 x i8> [[TMP1]] to <2 x i64>
-// CHECK-NEXT:    [[VSRI_N2:%.*]] = call <2 x i64> 
@llvm.aarch64.neon.vsri.v2i64(<2 x i64> [[VSRI_N]], <2 x i64> [[VSRI_N1]], i32 
64)
-// CHECK-NEXT:    ret <2 x i64> [[VSRI_N2]]
-//
-poly64x2_t test_vsriq_n_p64(poly64x2_t a, poly64x2_t b) {
-  return vsriq_n_p64(a, b, 64);
-}

>From 499596b30159bbbda9d5321ce567e8814058ea81 Mon Sep 17 00:00:00 2001
From: Vicky Nguyen <[email protected]>
Date: Wed, 13 May 2026 21:16:51 -0700
Subject: [PATCH 2/2] [CIR][AArch64] Generalize NEON SISD builtin emission via
 TypeModifier

Refactor emitCommonNeonSISDBuiltinExpr to derive argument and result types
from TypeModifier flags, mirroring the ARM.cpp flow in
EmitCommonNeonSISDBuiltinExpr + LookupNeonLLVMIntrinsic.
This keeps SISD lowering in one generic path and removes builtin-specific
type-shaping logic.
---
 .../lib/CIR/CodeGen/CIRGenBuiltinAArch64.cpp  | 104 +++++++++++++++---
 1 file changed, 86 insertions(+), 18 deletions(-)

diff --git a/clang/lib/CIR/CodeGen/CIRGenBuiltinAArch64.cpp 
b/clang/lib/CIR/CodeGen/CIRGenBuiltinAArch64.cpp
index 865cf86589443..830345dea525e 100644
--- a/clang/lib/CIR/CodeGen/CIRGenBuiltinAArch64.cpp
+++ b/clang/lib/CIR/CodeGen/CIRGenBuiltinAArch64.cpp
@@ -343,28 +343,96 @@ static mlir::Value emitCommonNeonSISDBuiltinExpr(
   case NEON::BI__builtin_neon_vrecpxh_f16:
   case NEON::BI__builtin_neon_vrsqrteh_f16:
   case NEON::BI__builtin_neon_vrsqrtsh_f16:
-    return emitNeonCall(cgf.cgm, cgf.getBuilder(),
-                        {cgf.convertType(expr->getArg(0)->getType())}, ops,
-                        llvmIntrName, cgf.convertType(expr->getType()), loc);
   case NEON::BI__builtin_neon_vsrid_n_s64:
-  case NEON::BI__builtin_neon_vsrid_n_u64: {
-    CIRGenBuilderTy &B = cgf.getBuilder();
-    mlir::Type scalarTy = cgf.convertType(expr->getType());
-    auto v1Ty = cir::VectorType::get(scalarTy, /*size=*/1);
-    mlir::Value a = B.createBitcast(ops[0], v1Ty);
-    mlir::Value b = B.createBitcast(ops[1], v1Ty);
-    llvm::SmallVector<mlir::Value, 4> vsriArgs{
-        a, b, B.createIntCast(ops[2], B.getUInt32Ty())};
-    mlir::Value r =
-        emitNeonCall(cgf.cgm, B,
-                     /*argTypes=*/{v1Ty, v1Ty, B.getUInt32Ty()}, vsriArgs,
-                     /*intrinsicName=*/"aarch64.neon.vsri",
-                     /*funcResTy=*/v1Ty, loc);
-    return B.createBitcast(r, scalarTy);
+  case NEON::BI__builtin_neon_vsrid_n_u64:
+    break;
   }
+
+  // Generic handling based on TypeModifier flags, mirroring
+  // EmitCommonNeonSISDBuiltinExpr + LookupNeonLLVMIntrinsic in ARM.cpp.
+  //
+  // The TypeModifier encodes how the intrinsic's argument and return types
+  // relate to the builtin's scalar types. For SISD builtins the key flags
+  // are:
+  //   - VectorizeArgTypes: wrap each arg type into a fixed-width vector
+  //   - Use64BitVectors / Use128BitVectors: choose the vector width
+  //     (when neither is set the vector has 1 element)
+  //   - AddRetType / VectorizeRetType: analogous flags for the return type
+  //
+  // ARM.cpp doesn't need to know about specific builtins like
+  // `vsrid_n_{s,u}64` because it lets LLVM resolve the intrinsic's
+  // signature (via `CGM.getIntrinsic`) and then walks the resolved
+  // Function* formal parameter types. CIR has no LLVMContext here, so
+  // we derive the same argument/result types directly from the Clang
+  // operand types.
+
+  unsigned modifier = info.TypeModifier;
+  CIRGenBuilderTy &builder = cgf.getBuilder();
+  mlir::Type argTy = cgf.convertType(expr->getArg(0)->getType());
+  mlir::Type resultTy = cgf.convertType(expr->getType());
+
+  int vectorSize = 0;
+  if (modifier & Use64BitVectors)
+    vectorSize = 64;
+  else if (modifier & Use128BitVectors)
+    vectorSize = 128;
+
+  auto wrapAsVector = [&](mlir::Type ty) -> cir::VectorType {
+    unsigned bits = cgf.cgm.getDataLayout().getTypeSizeInBits(ty);
+    unsigned elts = vectorSize ? vectorSize / bits : 1;
+    return cir::VectorType::get(ty, elts);
+  };
+
+  // Determine the vectorized data type.
+  cir::VectorType vecArgTy;
+  if (modifier & VectorizeArgTypes)
+    vecArgTy = wrapAsVector(argTy);
+
+  // Determine the intrinsic result type: `VectorizeRetType` returns a
+  // vector; otherwise, if data args are vectorized and `AddRetType` is
+  // unset, use a vector return with the same shape as those args.
+  mlir::Type funcResTy = resultTy;
+  if (modifier & VectorizeRetType)
+    funcResTy = wrapAsVector(resultTy);
+  else if (vecArgTy && !(modifier & AddRetType))
+    funcResTy = wrapAsVector(resultTy);
+
+  // Build the arg types for `emitNeonCall`.
+  llvm::SmallVector<mlir::Type> argTypes;
+  argTypes.reserve(ops.size());
+  for (mlir::Value op : ops) {
+    if (vecArgTy && op.getType() == argTy)
+      argTypes.push_back(vecArgTy);
+    else
+      argTypes.push_back(op.getType());
+  }
+
+  // Promote each scalar operand to a vector if the intrinsic expects one.
+  for (unsigned j = 0; j < ops.size(); ++j) {
+    mlir::Type opTy = ops[j].getType();
+    mlir::Type expectedTy = argTypes[j];
+
+    if (auto expectedVecTy = mlir::dyn_cast<cir::VectorType>(expectedTy)) {
+      if (!mlir::isa<cir::VectorType>(opTy)) {
+        // The constant argument to an _n_ intrinsic always has i32 type, so
+        // truncate/cast it to the vector element type before forming the 
vector
+        // value.
+        mlir::Type eltTy = expectedVecTy.getElementType();
+        if (opTy != eltTy)
+          ops[j] = builder.createIntCast(ops[j], eltTy);
+        ops[j] = builder.createBitcast(ops[j], expectedVecTy);
+      }
+    }
   }
 
-  return nullptr;
+  mlir::Value result = emitNeonCall(cgf.cgm, builder, std::move(argTypes), ops,
+                                    llvmIntrName, funcResTy, loc);
+
+  // If the intrinsic returned a vector but the builtin is scalar, cast back.
+  if (result.getType() != resultTy)
+    result = builder.createBitcast(result, resultTy);
+
+  return result;
 }
 
 
//===----------------------------------------------------------------------===//

_______________________________________________
cfe-commits mailing list
[email protected]
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits

Reply via email to