1 ; RUN: llc < %s -march=arm -mattr=+neon | FileCheck %s 2 3 define void @vst2i8(i8* %A, <8 x i8>* %B) nounwind { 4 ;CHECK: vst2i8: 5 ;Check the alignment value. Max for this instruction is 128 bits: 6 ;CHECK: vst2.8 {d16, d17}, [r0:64] 7 %tmp1 = load <8 x i8>* %B 8 call void @llvm.arm.neon.vst2.v8i8(i8* %A, <8 x i8> %tmp1, <8 x i8> %tmp1, i32 8) 9 ret void 10 } 11 12 ;Check for a post-increment updating store with register increment. 13 define void @vst2i8_update(i8** %ptr, <8 x i8>* %B, i32 %inc) nounwind { 14 ;CHECK: vst2i8_update: 15 ;CHECK: vst2.8 {d16, d17}, [r1], r2 16 %A = load i8** %ptr 17 %tmp1 = load <8 x i8>* %B 18 call void @llvm.arm.neon.vst2.v8i8(i8* %A, <8 x i8> %tmp1, <8 x i8> %tmp1, i32 4) 19 %tmp2 = getelementptr i8* %A, i32 %inc 20 store i8* %tmp2, i8** %ptr 21 ret void 22 } 23 24 define void @vst2i16(i16* %A, <4 x i16>* %B) nounwind { 25 ;CHECK: vst2i16: 26 ;Check the alignment value. Max for this instruction is 128 bits: 27 ;CHECK: vst2.16 {d16, d17}, [r0:128] 28 %tmp0 = bitcast i16* %A to i8* 29 %tmp1 = load <4 x i16>* %B 30 call void @llvm.arm.neon.vst2.v4i16(i8* %tmp0, <4 x i16> %tmp1, <4 x i16> %tmp1, i32 32) 31 ret void 32 } 33 34 define void @vst2i32(i32* %A, <2 x i32>* %B) nounwind { 35 ;CHECK: vst2i32: 36 ;CHECK: vst2.32 37 %tmp0 = bitcast i32* %A to i8* 38 %tmp1 = load <2 x i32>* %B 39 call void @llvm.arm.neon.vst2.v2i32(i8* %tmp0, <2 x i32> %tmp1, <2 x i32> %tmp1, i32 1) 40 ret void 41 } 42 43 define void @vst2f(float* %A, <2 x float>* %B) nounwind { 44 ;CHECK: vst2f: 45 ;CHECK: vst2.32 46 %tmp0 = bitcast float* %A to i8* 47 %tmp1 = load <2 x float>* %B 48 call void @llvm.arm.neon.vst2.v2f32(i8* %tmp0, <2 x float> %tmp1, <2 x float> %tmp1, i32 1) 49 ret void 50 } 51 52 define void @vst2i64(i64* %A, <1 x i64>* %B) nounwind { 53 ;CHECK: vst2i64: 54 ;Check the alignment value. Max for this instruction is 128 bits: 55 ;CHECK: vst1.64 {d16, d17}, [r0:128] 56 %tmp0 = bitcast i64* %A to i8* 57 %tmp1 = load <1 x i64>* %B 58 call void @llvm.arm.neon.vst2.v1i64(i8* %tmp0, <1 x i64> %tmp1, <1 x i64> %tmp1, i32 32) 59 ret void 60 } 61 62 ;Check for a post-increment updating store. 63 define void @vst2i64_update(i64** %ptr, <1 x i64>* %B) nounwind { 64 ;CHECK: vst2i64_update: 65 ;CHECK: vst1.64 {d16, d17}, [r1:64]! 66 %A = load i64** %ptr 67 %tmp0 = bitcast i64* %A to i8* 68 %tmp1 = load <1 x i64>* %B 69 call void @llvm.arm.neon.vst2.v1i64(i8* %tmp0, <1 x i64> %tmp1, <1 x i64> %tmp1, i32 8) 70 %tmp2 = getelementptr i64* %A, i32 2 71 store i64* %tmp2, i64** %ptr 72 ret void 73 } 74 75 define void @vst2Qi8(i8* %A, <16 x i8>* %B) nounwind { 76 ;CHECK: vst2Qi8: 77 ;Check the alignment value. Max for this instruction is 256 bits: 78 ;CHECK: vst2.8 {d16, d17, d18, d19}, [r0:64] 79 %tmp1 = load <16 x i8>* %B 80 call void @llvm.arm.neon.vst2.v16i8(i8* %A, <16 x i8> %tmp1, <16 x i8> %tmp1, i32 8) 81 ret void 82 } 83 84 define void @vst2Qi16(i16* %A, <8 x i16>* %B) nounwind { 85 ;CHECK: vst2Qi16: 86 ;Check the alignment value. Max for this instruction is 256 bits: 87 ;CHECK: vst2.16 {d16, d17, d18, d19}, [r0:128] 88 %tmp0 = bitcast i16* %A to i8* 89 %tmp1 = load <8 x i16>* %B 90 call void @llvm.arm.neon.vst2.v8i16(i8* %tmp0, <8 x i16> %tmp1, <8 x i16> %tmp1, i32 16) 91 ret void 92 } 93 94 define void @vst2Qi32(i32* %A, <4 x i32>* %B) nounwind { 95 ;CHECK: vst2Qi32: 96 ;Check the alignment value. Max for this instruction is 256 bits: 97 ;CHECK: vst2.32 {d16, d17, d18, d19}, [r0:256] 98 %tmp0 = bitcast i32* %A to i8* 99 %tmp1 = load <4 x i32>* %B 100 call void @llvm.arm.neon.vst2.v4i32(i8* %tmp0, <4 x i32> %tmp1, <4 x i32> %tmp1, i32 64) 101 ret void 102 } 103 104 define void @vst2Qf(float* %A, <4 x float>* %B) nounwind { 105 ;CHECK: vst2Qf: 106 ;CHECK: vst2.32 107 %tmp0 = bitcast float* %A to i8* 108 %tmp1 = load <4 x float>* %B 109 call void @llvm.arm.neon.vst2.v4f32(i8* %tmp0, <4 x float> %tmp1, <4 x float> %tmp1, i32 1) 110 ret void 111 } 112 113 define i8* @vst2update(i8* %out, <4 x i16>* %B) nounwind { 114 ;CHECK: vst2update 115 ;CHECK: vst2.16 {d16, d17}, [r0]! 116 %tmp1 = load <4 x i16>* %B 117 tail call void @llvm.arm.neon.vst2.v4i16(i8* %out, <4 x i16> %tmp1, <4 x i16> %tmp1, i32 2) 118 %t5 = getelementptr inbounds i8* %out, i32 16 119 ret i8* %t5 120 } 121 122 define i8* @vst2update2(i8 * %out, <4 x float> * %this) nounwind optsize ssp align 2 { 123 ;CHECK: vst2update2 124 ;CHECK: vst2.32 {d16, d17, d18, d19}, [r0]! 125 %tmp1 = load <4 x float>* %this 126 call void @llvm.arm.neon.vst2.v4f32(i8* %out, <4 x float> %tmp1, <4 x float> %tmp1, i32 4) nounwind 127 %tmp2 = getelementptr inbounds i8* %out, i32 32 128 ret i8* %tmp2 129 } 130 131 declare void @llvm.arm.neon.vst2.v8i8(i8*, <8 x i8>, <8 x i8>, i32) nounwind 132 declare void @llvm.arm.neon.vst2.v4i16(i8*, <4 x i16>, <4 x i16>, i32) nounwind 133 declare void @llvm.arm.neon.vst2.v2i32(i8*, <2 x i32>, <2 x i32>, i32) nounwind 134 declare void @llvm.arm.neon.vst2.v2f32(i8*, <2 x float>, <2 x float>, i32) nounwind 135 declare void @llvm.arm.neon.vst2.v1i64(i8*, <1 x i64>, <1 x i64>, i32) nounwind 136 137 declare void @llvm.arm.neon.vst2.v16i8(i8*, <16 x i8>, <16 x i8>, i32) nounwind 138 declare void @llvm.arm.neon.vst2.v8i16(i8*, <8 x i16>, <8 x i16>, i32) nounwind 139 declare void @llvm.arm.neon.vst2.v4i32(i8*, <4 x i32>, <4 x i32>, i32) nounwind 140 declare void @llvm.arm.neon.vst2.v4f32(i8*, <4 x float>, <4 x float>, i32) nounwind 141