1; RUN: llc < %s -mtriple=arm64-eabi -aarch64-neon-syntax=apple | FileCheck %s 2 3define <8 x i8> @vsras8(<8 x i8>* %A, <8 x i8>* %B) nounwind { 4;CHECK-LABEL: vsras8: 5;CHECK: ssra.8b 6 %tmp1 = load <8 x i8>, <8 x i8>* %A 7 %tmp2 = load <8 x i8>, <8 x i8>* %B 8 %tmp3 = ashr <8 x i8> %tmp2, < i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7 > 9 %tmp4 = add <8 x i8> %tmp1, %tmp3 10 ret <8 x i8> %tmp4 11} 12 13define <4 x i16> @vsras16(<4 x i16>* %A, <4 x i16>* %B) nounwind { 14;CHECK-LABEL: vsras16: 15;CHECK: ssra.4h 16 %tmp1 = load <4 x i16>, <4 x i16>* %A 17 %tmp2 = load <4 x i16>, <4 x i16>* %B 18 %tmp3 = ashr <4 x i16> %tmp2, < i16 15, i16 15, i16 15, i16 15 > 19 %tmp4 = add <4 x i16> %tmp1, %tmp3 20 ret <4 x i16> %tmp4 21} 22 23define <2 x i32> @vsras32(<2 x i32>* %A, <2 x i32>* %B) nounwind { 24;CHECK-LABEL: vsras32: 25;CHECK: ssra.2s 26 %tmp1 = load <2 x i32>, <2 x i32>* %A 27 %tmp2 = load <2 x i32>, <2 x i32>* %B 28 %tmp3 = ashr <2 x i32> %tmp2, < i32 31, i32 31 > 29 %tmp4 = add <2 x i32> %tmp1, %tmp3 30 ret <2 x i32> %tmp4 31} 32 33define <16 x i8> @vsraQs8(<16 x i8>* %A, <16 x i8>* %B) nounwind { 34;CHECK-LABEL: vsraQs8: 35;CHECK: ssra.16b 36 %tmp1 = load <16 x i8>, <16 x i8>* %A 37 %tmp2 = load <16 x i8>, <16 x i8>* %B 38 %tmp3 = ashr <16 x i8> %tmp2, < i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7 > 39 %tmp4 = add <16 x i8> %tmp1, %tmp3 40 ret <16 x i8> %tmp4 41} 42 43define <8 x i16> @vsraQs16(<8 x i16>* %A, <8 x i16>* %B) nounwind { 44;CHECK-LABEL: vsraQs16: 45;CHECK: ssra.8h 46 %tmp1 = load <8 x i16>, <8 x i16>* %A 47 %tmp2 = load <8 x i16>, <8 x i16>* %B 48 %tmp3 = ashr <8 x i16> %tmp2, < i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15 > 49 %tmp4 = add <8 x i16> %tmp1, %tmp3 50 ret <8 x i16> %tmp4 51} 52 53define <4 x i32> @vsraQs32(<4 x i32>* %A, <4 x i32>* %B) nounwind { 54;CHECK-LABEL: vsraQs32: 55;CHECK: ssra.4s 56 %tmp1 = load <4 x i32>, <4 x i32>* %A 57 %tmp2 = load <4 x i32>, <4 x i32>* %B 58 %tmp3 = ashr <4 x i32> %tmp2, < i32 31, i32 31, i32 31, i32 31 > 59 %tmp4 = add <4 x i32> %tmp1, %tmp3 60 ret <4 x i32> %tmp4 61} 62 63define <2 x i64> @vsraQs64(<2 x i64>* %A, <2 x i64>* %B) nounwind { 64;CHECK-LABEL: vsraQs64: 65;CHECK: ssra.2d 66 %tmp1 = load <2 x i64>, <2 x i64>* %A 67 %tmp2 = load <2 x i64>, <2 x i64>* %B 68 %tmp3 = ashr <2 x i64> %tmp2, < i64 63, i64 63 > 69 %tmp4 = add <2 x i64> %tmp1, %tmp3 70 ret <2 x i64> %tmp4 71} 72 73define <8 x i8> @vsrau8(<8 x i8>* %A, <8 x i8>* %B) nounwind { 74;CHECK-LABEL: vsrau8: 75;CHECK: usra.8b 76 %tmp1 = load <8 x i8>, <8 x i8>* %A 77 %tmp2 = load <8 x i8>, <8 x i8>* %B 78 %tmp3 = lshr <8 x i8> %tmp2, < i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7 > 79 %tmp4 = add <8 x i8> %tmp1, %tmp3 80 ret <8 x i8> %tmp4 81} 82 83define <4 x i16> @vsrau16(<4 x i16>* %A, <4 x i16>* %B) nounwind { 84;CHECK-LABEL: vsrau16: 85;CHECK: usra.4h 86 %tmp1 = load <4 x i16>, <4 x i16>* %A 87 %tmp2 = load <4 x i16>, <4 x i16>* %B 88 %tmp3 = lshr <4 x i16> %tmp2, < i16 15, i16 15, i16 15, i16 15 > 89 %tmp4 = add <4 x i16> %tmp1, %tmp3 90 ret <4 x i16> %tmp4 91} 92 93define <2 x i32> @vsrau32(<2 x i32>* %A, <2 x i32>* %B) nounwind { 94;CHECK-LABEL: vsrau32: 95;CHECK: usra.2s 96 %tmp1 = load <2 x i32>, <2 x i32>* %A 97 %tmp2 = load <2 x i32>, <2 x i32>* %B 98 %tmp3 = lshr <2 x i32> %tmp2, < i32 31, i32 31 > 99 %tmp4 = add <2 x i32> %tmp1, %tmp3 100 ret <2 x i32> %tmp4 101} 102 103 104define <16 x i8> @vsraQu8(<16 x i8>* %A, <16 x i8>* %B) nounwind { 105;CHECK-LABEL: vsraQu8: 106;CHECK: usra.16b 107 %tmp1 = load <16 x i8>, <16 x i8>* %A 108 %tmp2 = load <16 x i8>, <16 x i8>* %B 109 %tmp3 = lshr <16 x i8> %tmp2, < i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7 > 110 %tmp4 = add <16 x i8> %tmp1, %tmp3 111 ret <16 x i8> %tmp4 112} 113 114define <8 x i16> @vsraQu16(<8 x i16>* %A, <8 x i16>* %B) nounwind { 115;CHECK-LABEL: vsraQu16: 116;CHECK: usra.8h 117 %tmp1 = load <8 x i16>, <8 x i16>* %A 118 %tmp2 = load <8 x i16>, <8 x i16>* %B 119 %tmp3 = lshr <8 x i16> %tmp2, < i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15 > 120 %tmp4 = add <8 x i16> %tmp1, %tmp3 121 ret <8 x i16> %tmp4 122} 123 124define <4 x i32> @vsraQu32(<4 x i32>* %A, <4 x i32>* %B) nounwind { 125;CHECK-LABEL: vsraQu32: 126;CHECK: usra.4s 127 %tmp1 = load <4 x i32>, <4 x i32>* %A 128 %tmp2 = load <4 x i32>, <4 x i32>* %B 129 %tmp3 = lshr <4 x i32> %tmp2, < i32 31, i32 31, i32 31, i32 31 > 130 %tmp4 = add <4 x i32> %tmp1, %tmp3 131 ret <4 x i32> %tmp4 132} 133 134define <2 x i64> @vsraQu64(<2 x i64>* %A, <2 x i64>* %B) nounwind { 135;CHECK-LABEL: vsraQu64: 136;CHECK: usra.2d 137 %tmp1 = load <2 x i64>, <2 x i64>* %A 138 %tmp2 = load <2 x i64>, <2 x i64>* %B 139 %tmp3 = lshr <2 x i64> %tmp2, < i64 63, i64 63 > 140 %tmp4 = add <2 x i64> %tmp1, %tmp3 141 ret <2 x i64> %tmp4 142} 143 144define <1 x i64> @vsra_v1i64(<1 x i64> %A, <1 x i64> %B) nounwind { 145; CHECK-LABEL: vsra_v1i64: 146; CHECK: ssra d0, d1, #63 147 %tmp3 = ashr <1 x i64> %B, < i64 63 > 148 %tmp4 = add <1 x i64> %A, %tmp3 149 ret <1 x i64> %tmp4 150} 151