[AArch64] Add support for legacy AArch32 NEON scalar shift right by immediate

and accumulate instructions.

llvm-svn: 194732

2 files changed, 35 insertions, 0 deletions

diff --git a/clang/lib/CodeGen/CGBuiltin.cpp b/clang/lib/CodeGen/CGBuiltin.cpp
index a2669fe8fe4..363caedf314 100644
--- a/clang/lib/CodeGen/CGBuiltin.cpp
+++ b/clang/lib/CodeGen/CGBuiltin.cpp
@@ -2913,10 +2913,21 @@ Value *CodeGenFunction::EmitAArch64BuiltinExpr(unsigned BuiltinID,
                : Intrinsic::aarch64_neon_vsrshr;
     return EmitNeonCall(CGM.getIntrinsic(Int, Ty), Ops, "vrshr_n");
   case AArch64::BI__builtin_neon_vsra_n_v:
+    if (VTy->getElementType()->isIntegerTy(64)) {
+      Int = usgn ? Intrinsic::aarch64_neon_vsradu_n
+                 : Intrinsic::aarch64_neon_vsrads_n;
+      return EmitNeonCall(CGM.getIntrinsic(Int), Ops, "vsra_n");
+    }
     return EmitARMBuiltinExpr(ARM::BI__builtin_neon_vsra_n_v, E);
   case AArch64::BI__builtin_neon_vsraq_n_v:
     return EmitARMBuiltinExpr(ARM::BI__builtin_neon_vsraq_n_v, E);
   case AArch64::BI__builtin_neon_vrsra_n_v:
+    if (VTy->getElementType()->isIntegerTy(64)) {
+      Int = usgn ? Intrinsic::aarch64_neon_vrsradu_n
+                 : Intrinsic::aarch64_neon_vrsrads_n;
+      return EmitNeonCall(CGM.getIntrinsic(Int), Ops, "vrsra_n");
+    }
+    // fall through
   case AArch64::BI__builtin_neon_vrsraq_n_v: {
     Ops[0] = Builder.CreateBitCast(Ops[0], Ty);
     Ops[1] = Builder.CreateBitCast(Ops[1], Ty);
diff --git a/clang/test/CodeGen/aarch64-neon-intrinsics.c b/clang/test/CodeGen/aarch64-neon-intrinsics.c
index c73e87d6ac6..4c2f3cc1016 100644
--- a/clang/test/CodeGen/aarch64-neon-intrinsics.c
+++ b/clang/test/CodeGen/aarch64-neon-intrinsics.c
@@ -7538,24 +7538,48 @@ int64_t test_vsrad_n_s64(int64_t a, int64_t b) {
   return (int64_t)vsrad_n_s64(a, b, 63);
 }
 
+int64x1_t test_vsra_n_s64(int64x1_t a, int64x1_t b) {
+// CHECK: test_vsra_n_s64
+// CHECK: ssra d{{[0-9]+}}, d{{[0-9]+}}, #1
+  return vsra_n_s64(a, b, 1);
+}
+
 uint64_t test_vsrad_n_u64(uint64_t a, uint64_t b) {
 // CHECK-LABEL: test_vsrad_n_u64
 // CHECK: usra {{d[0-9]+}}, {{d[0-9]+}}, #63
   return (uint64_t)vsrad_n_u64(a, b, 63);
 }
 
+uint64x1_t test_vsra_n_u64(uint64x1_t a, uint64x1_t b) {
+// CHECK: test_vsra_n_u64
+// CHECK: usra d{{[0-9]+}}, d{{[0-9]+}}, #1
+  return vsra_n_u64(a, b, 1);
+}
+
 int64_t test_vrsrad_n_s64(int64_t a, int64_t b) {
 // CHECK-LABEL: test_vrsrad_n_s64
 // CHECK: srsra {{d[0-9]+}}, {{d[0-9]+}}, #63
   return (int64_t)vrsrad_n_s64(a, b, 63);
 }
 
+int64x1_t test_vrsra_n_s64(int64x1_t a, int64x1_t b) {
+// CHECK: test_vrsra_n_s64
+// CHECK: srsra d{{[0-9]+}}, d{{[0-9]+}}, #1
+  return vrsra_n_s64(a, b, 1);
+}
+
 uint64_t test_vrsrad_n_u64(uint64_t a, uint64_t b) {
 // CHECK-LABEL: test_vrsrad_n_u64
 // CHECK: ursra {{d[0-9]+}}, {{d[0-9]+}}, #63
   return (uint64_t)vrsrad_n_u64(a, b, 63);
 }
 
+uint64x1_t test_vrsra_n_u64(uint64x1_t a, uint64x1_t b) {
+// CHECK: test_vrsra_n_u64
+// CHECK: ursra d{{[0-9]+}}, d{{[0-9]+}}, #1
+  return vrsra_n_u64(a, b, 1);
+}
+
 int64_t test_vshld_n_s64(int64_t a) {
 // CHECK-LABEL: test_vshld_n_s64
 // CHECK: shl {{d[0-9]+}}, {{d[0-9]+}}, #0