[CIR][CIRGen][Builtin][Neon] Lower neon_vaddvq_s64/s32, neon_vaddv_s32/u32, (#1240)

Co-authored-by: Bruno Cardoso Lopes <[email protected]>

Author: ghehg

clang/lib/CIR/CodeGen/CIRGenBuiltinAArch64.cpp

@@ -2678,19 +2678,17 @@ static mlir::Value emitCommonNeonSISDBuiltinExpr(
   case NEON::BI__builtin_neon_vaddlvq_u32:
     return emitNeonCall(builder, {argTy}, ops, "aarch64.neon.uaddlv", resultTy,
                         loc);
-  case NEON::BI__builtin_neon_vaddv_s32:
-    llvm_unreachable(" neon_vaddv_s32 NYI ");
-  case NEON::BI__builtin_neon_vaddv_u32:
-    llvm_unreachable(" neon_vaddv_u32 NYI ");
   case NEON::BI__builtin_neon_vaddv_f32:
   case NEON::BI__builtin_neon_vaddvq_f32:
   case NEON::BI__builtin_neon_vaddvq_f64:
     return emitNeonCall(builder, {argTy}, ops, "aarch64.neon.faddv", resultTy,
                         loc);
+  case NEON::BI__builtin_neon_vaddv_s32:
   case NEON::BI__builtin_neon_vaddvq_s32:
-    llvm_unreachable(" neon_vaddvq_s32 NYI ");
   case NEON::BI__builtin_neon_vaddvq_s64:
-    llvm_unreachable(" neon_vaddvq_s64 NYI ");
+    return emitNeonCall(builder, {argTy}, ops, "aarch64.neon.saddv", resultTy,
+                        loc);
+  case NEON::BI__builtin_neon_vaddv_u32:
   case NEON::BI__builtin_neon_vaddvq_u32:
   case NEON::BI__builtin_neon_vaddvq_u64:
     return emitNeonCall(builder, {argTy}, ops, "aarch64.neon.uaddv", resultTy,

clang/test/CIR/CodeGen/AArch64/neon-arith.c

@@ -952,3 +952,27 @@ uint64_t test_vaddvq_u64(uint64x2_t a) {
   // LLVM:   [[VADDVQ_U64_I:%.*]] = call i64 @llvm.aarch64.neon.uaddv.i64.v2i64(<2 x i64> {{%.*}})
   // LLVM:   ret i64 [[VADDVQ_U64_I]]
 }
+
+int32_t test_vaddvq_s32(int32x4_t a) {
+  return vaddvq_s32(a);
+
+  // CIR-LABEL: vaddvq_s32
+  // CIR: cir.llvm.intrinsic "aarch64.neon.saddv" {{%.*}} : (!cir.vector<!s32i x 4>) -> !s32i
+
+  // LLVM-LABEL: test_vaddvq_s32
+  // LLVM-SAME: (<4 x i32> [[a:%.*]])
+  // LLVM:   [[VADDVQ_S32_I:%.*]] = call i32 @llvm.aarch64.neon.saddv.i32.v4i32(<4 x i32> [[a]])
+  // LLVM:   ret i32 [[VADDVQ_S32_I]]
+}
+
+int64_t test_vaddvq_s64(int64x2_t a) {
+  return vaddvq_s64(a);
+
+  // CIR-LABEL: vaddvq_s64
+  // CIR: cir.llvm.intrinsic "aarch64.neon.saddv" {{%.*}} : (!cir.vector<!s64i x 2>) -> !s64i
+
+  // LLVM-LABEL: test_vaddvq_s64
+  // LLVM-SAME: (<2 x i64> [[a:%.*]])
+  // LLVM:   [[VADDVQ_S64_I:%.*]] = call i64 @llvm.aarch64.neon.saddv.i64.v2i64(<2 x i64> [[a]])
+  // LLVM:   ret i64 [[VADDVQ_S64_I]]
+}

clang/test/CIR/CodeGen/AArch64/neon.c

@@ -18967,19 +18967,29 @@ float64x1_t test_vmax_f64(float64x1_t a, float64x1_t b) {
 //   return vmaxv_u32(a);
 // }
 
-// NYI-LABEL: @test_vaddv_s32(
-// NYI:   [[VADDV_S32_I:%.*]] = call i32 @llvm.aarch64.neon.saddv.i32.v2i32(<2 x i32> %a)
-// NYI:   ret i32 [[VADDV_S32_I]]
-// int32_t test_vaddv_s32(int32x2_t a) {
-//   return vaddv_s32(a);
-// }
+int32_t test_vaddv_s32(int32x2_t a) {
+  return vaddv_s32(a);
 
-// NYI-LABEL: @test_vaddv_u32(
-// NYI:   [[VADDV_U32_I:%.*]] = call i32 @llvm.aarch64.neon.uaddv.i32.v2i32(<2 x i32> %a)
-// NYI:   ret i32 [[VADDV_U32_I]]
-// uint32_t test_vaddv_u32(uint32x2_t a) {
-//   return vaddv_u32(a);
-// }
+  // CIR-LABEL: vaddv_s32
+  // CIR: cir.llvm.intrinsic "aarch64.neon.saddv" {{%.*}} : (!cir.vector<!s32i x 2>) -> !s32i
+
+  // LLVM-LABEL: test_vaddv_s32
+  // LLVM-SAME: (<2 x i32> [[a:%.*]])
+  // LLVM:   [[VADDV_S32_I:%.*]] = call i32 @llvm.aarch64.neon.saddv.i32.v2i32(<2 x i32> [[a]])
+  // LLVM:   ret i32 [[VADDV_S32_I]]
+}
+
+uint32_t test_vaddv_u32(uint32x2_t a) {
+  return vaddv_u32(a);
+
+  // CIR-LABEL: vaddv_u32
+  // CIR: cir.llvm.intrinsic "aarch64.neon.uaddv" {{%.*}} : (!cir.vector<!u32i x 2>) -> !u32i
+
+  // LLVM-LABEL: test_vaddv_u32
+  // LLVM-SAME: (<2 x i32> [[a:%.*]])
+  // LLVM:   [[VADDV_U32_I:%.*]] = call i32 @llvm.aarch64.neon.uaddv.i32.v2i32(<2 x i32> [[a]])
+  // LLVM:   ret i32 [[VADDV_U32_I]]
+}
 
 // NYI-LABEL: @test_vaddlv_s32(
 // NYI:   [[VADDLV_S32_I:%.*]] = call i64 @llvm.aarch64.neon.saddlv.i64.v2i32(<2 x i32> %a)