1 ; RUN: llc < %s -march=arm -mattr=+neon | FileCheck %s
3 %struct.__neon_int8x8x4_t = type { <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8> }
4 %struct.__neon_int16x4x4_t = type { <4 x i16>, <4 x i16>, <4 x i16>, <4 x i16> }
5 %struct.__neon_int32x2x4_t = type { <2 x i32>, <2 x i32>, <2 x i32>, <2 x i32> }
6 %struct.__neon_float32x2x4_t = type { <2 x float>, <2 x float>, <2 x float>, <2 x float> }
7 %struct.__neon_int64x1x4_t = type { <1 x i64>, <1 x i64>, <1 x i64>, <1 x i64> }
9 %struct.__neon_int8x16x4_t = type { <16 x i8>, <16 x i8>, <16 x i8>, <16 x i8> }
10 %struct.__neon_int16x8x4_t = type { <8 x i16>, <8 x i16>, <8 x i16>, <8 x i16> }
11 %struct.__neon_int32x4x4_t = type { <4 x i32>, <4 x i32>, <4 x i32>, <4 x i32> }
12 %struct.__neon_float32x4x4_t = type { <4 x float>, <4 x float>, <4 x float>, <4 x float> }
14 define <8 x i8> @vld4i8(i8* %A) nounwind {
16 ;Check the alignment value. Max for this instruction is 256 bits:
17 ;CHECK: vld4.8 {d16, d17, d18, d19}, [r0, :64]
18 %tmp1 = call %struct.__neon_int8x8x4_t @llvm.arm.neon.vld4.v8i8(i8* %A, i32 8)
19 %tmp2 = extractvalue %struct.__neon_int8x8x4_t %tmp1, 0
20 %tmp3 = extractvalue %struct.__neon_int8x8x4_t %tmp1, 2
21 %tmp4 = add <8 x i8> %tmp2, %tmp3
25 define <4 x i16> @vld4i16(i16* %A) nounwind {
27 ;Check the alignment value. Max for this instruction is 256 bits:
28 ;CHECK: vld4.16 {d16, d17, d18, d19}, [r0, :128]
29 %tmp0 = bitcast i16* %A to i8*
30 %tmp1 = call %struct.__neon_int16x4x4_t @llvm.arm.neon.vld4.v4i16(i8* %tmp0, i32 16)
31 %tmp2 = extractvalue %struct.__neon_int16x4x4_t %tmp1, 0
32 %tmp3 = extractvalue %struct.__neon_int16x4x4_t %tmp1, 2
33 %tmp4 = add <4 x i16> %tmp2, %tmp3
37 define <2 x i32> @vld4i32(i32* %A) nounwind {
39 ;Check the alignment value. Max for this instruction is 256 bits:
40 ;CHECK: vld4.32 {d16, d17, d18, d19}, [r0, :256]
41 %tmp0 = bitcast i32* %A to i8*
42 %tmp1 = call %struct.__neon_int32x2x4_t @llvm.arm.neon.vld4.v2i32(i8* %tmp0, i32 32)
43 %tmp2 = extractvalue %struct.__neon_int32x2x4_t %tmp1, 0
44 %tmp3 = extractvalue %struct.__neon_int32x2x4_t %tmp1, 2
45 %tmp4 = add <2 x i32> %tmp2, %tmp3
49 define <2 x float> @vld4f(float* %A) nounwind {
52 %tmp0 = bitcast float* %A to i8*
53 %tmp1 = call %struct.__neon_float32x2x4_t @llvm.arm.neon.vld4.v2f32(i8* %tmp0, i32 1)
54 %tmp2 = extractvalue %struct.__neon_float32x2x4_t %tmp1, 0
55 %tmp3 = extractvalue %struct.__neon_float32x2x4_t %tmp1, 2
56 %tmp4 = fadd <2 x float> %tmp2, %tmp3
60 define <1 x i64> @vld4i64(i64* %A) nounwind {
62 ;Check the alignment value. Max for this instruction is 256 bits:
63 ;CHECK: vld1.64 {d16, d17, d18, d19}, [r0, :256]
64 %tmp0 = bitcast i64* %A to i8*
65 %tmp1 = call %struct.__neon_int64x1x4_t @llvm.arm.neon.vld4.v1i64(i8* %tmp0, i32 64)
66 %tmp2 = extractvalue %struct.__neon_int64x1x4_t %tmp1, 0
67 %tmp3 = extractvalue %struct.__neon_int64x1x4_t %tmp1, 2
68 %tmp4 = add <1 x i64> %tmp2, %tmp3
72 define <16 x i8> @vld4Qi8(i8* %A) nounwind {
74 ;Check the alignment value. Max for this instruction is 256 bits:
75 ;CHECK: vld4.8 {d16, d18, d20, d22}, [r0, :256]!
76 ;CHECK: vld4.8 {d17, d19, d21, d23}, [r0, :256]
77 %tmp1 = call %struct.__neon_int8x16x4_t @llvm.arm.neon.vld4.v16i8(i8* %A, i32 64)
78 %tmp2 = extractvalue %struct.__neon_int8x16x4_t %tmp1, 0
79 %tmp3 = extractvalue %struct.__neon_int8x16x4_t %tmp1, 2
80 %tmp4 = add <16 x i8> %tmp2, %tmp3
84 define <8 x i16> @vld4Qi16(i16* %A) nounwind {
86 ;Check for no alignment specifier.
87 ;CHECK: vld4.16 {d16, d18, d20, d22}, [r0]!
88 ;CHECK: vld4.16 {d17, d19, d21, d23}, [r0]
89 %tmp0 = bitcast i16* %A to i8*
90 %tmp1 = call %struct.__neon_int16x8x4_t @llvm.arm.neon.vld4.v8i16(i8* %tmp0, i32 1)
91 %tmp2 = extractvalue %struct.__neon_int16x8x4_t %tmp1, 0
92 %tmp3 = extractvalue %struct.__neon_int16x8x4_t %tmp1, 2
93 %tmp4 = add <8 x i16> %tmp2, %tmp3
97 define <4 x i32> @vld4Qi32(i32* %A) nounwind {
101 %tmp0 = bitcast i32* %A to i8*
102 %tmp1 = call %struct.__neon_int32x4x4_t @llvm.arm.neon.vld4.v4i32(i8* %tmp0, i32 1)
103 %tmp2 = extractvalue %struct.__neon_int32x4x4_t %tmp1, 0
104 %tmp3 = extractvalue %struct.__neon_int32x4x4_t %tmp1, 2
105 %tmp4 = add <4 x i32> %tmp2, %tmp3
109 define <4 x float> @vld4Qf(float* %A) nounwind {
113 %tmp0 = bitcast float* %A to i8*
114 %tmp1 = call %struct.__neon_float32x4x4_t @llvm.arm.neon.vld4.v4f32(i8* %tmp0, i32 1)
115 %tmp2 = extractvalue %struct.__neon_float32x4x4_t %tmp1, 0
116 %tmp3 = extractvalue %struct.__neon_float32x4x4_t %tmp1, 2
117 %tmp4 = fadd <4 x float> %tmp2, %tmp3
118 ret <4 x float> %tmp4
121 declare %struct.__neon_int8x8x4_t @llvm.arm.neon.vld4.v8i8(i8*, i32) nounwind readonly
122 declare %struct.__neon_int16x4x4_t @llvm.arm.neon.vld4.v4i16(i8*, i32) nounwind readonly
123 declare %struct.__neon_int32x2x4_t @llvm.arm.neon.vld4.v2i32(i8*, i32) nounwind readonly
124 declare %struct.__neon_float32x2x4_t @llvm.arm.neon.vld4.v2f32(i8*, i32) nounwind readonly
125 declare %struct.__neon_int64x1x4_t @llvm.arm.neon.vld4.v1i64(i8*, i32) nounwind readonly
127 declare %struct.__neon_int8x16x4_t @llvm.arm.neon.vld4.v16i8(i8*, i32) nounwind readonly
128 declare %struct.__neon_int16x8x4_t @llvm.arm.neon.vld4.v8i16(i8*, i32) nounwind readonly
129 declare %struct.__neon_int32x4x4_t @llvm.arm.neon.vld4.v4i32(i8*, i32) nounwind readonly
130 declare %struct.__neon_float32x4x4_t @llvm.arm.neon.vld4.v4f32(i8*, i32) nounwind readonly