1 ; RUN: llc < %s -march=arm -mattr=+neon | FileCheck %s
3 define <8 x i8> @vld1dupi8(i8* %A) nounwind {
5 ;Check the (default) alignment value.
6 ;CHECK: vld1.8 {d16[]}, [r0]
7 %tmp1 = load i8* %A, align 8
8 %tmp2 = insertelement <8 x i8> undef, i8 %tmp1, i32 0
9 %tmp3 = shufflevector <8 x i8> %tmp2, <8 x i8> undef, <8 x i32> zeroinitializer
13 define <4 x i16> @vld1dupi16(i16* %A) nounwind {
15 ;Check the alignment value. Max for this instruction is 16 bits:
16 ;CHECK: vld1.16 {d16[]}, [r0, :16]
17 %tmp1 = load i16* %A, align 8
18 %tmp2 = insertelement <4 x i16> undef, i16 %tmp1, i32 0
19 %tmp3 = shufflevector <4 x i16> %tmp2, <4 x i16> undef, <4 x i32> zeroinitializer
23 define <2 x i32> @vld1dupi32(i32* %A) nounwind {
25 ;Check the alignment value. Max for this instruction is 32 bits:
26 ;CHECK: vld1.32 {d16[]}, [r0, :32]
27 %tmp1 = load i32* %A, align 8
28 %tmp2 = insertelement <2 x i32> undef, i32 %tmp1, i32 0
29 %tmp3 = shufflevector <2 x i32> %tmp2, <2 x i32> undef, <2 x i32> zeroinitializer
33 define <2 x float> @vld1dupf(float* %A) nounwind {
35 ;CHECK: vld1.32 {d16[]}, [r0]
36 %tmp0 = load float* %A
37 %tmp1 = insertelement <2 x float> undef, float %tmp0, i32 0
38 %tmp2 = shufflevector <2 x float> %tmp1, <2 x float> undef, <2 x i32> zeroinitializer
42 define <16 x i8> @vld1dupQi8(i8* %A) nounwind {
44 ;Check the (default) alignment value.
45 ;CHECK: vld1.8 {d16[], d17[]}, [r0]
46 %tmp1 = load i8* %A, align 8
47 %tmp2 = insertelement <16 x i8> undef, i8 %tmp1, i32 0
48 %tmp3 = shufflevector <16 x i8> %tmp2, <16 x i8> undef, <16 x i32> zeroinitializer
52 define <4 x float> @vld1dupQf(float* %A) nounwind {
54 ;CHECK: vld1.32 {d16[], d17[]}, [r0]
55 %tmp0 = load float* %A
56 %tmp1 = insertelement <4 x float> undef, float %tmp0, i32 0
57 %tmp2 = shufflevector <4 x float> %tmp1, <4 x float> undef, <4 x i32> zeroinitializer
61 %struct.__neon_int8x8x2_t = type { <8 x i8>, <8 x i8> }
62 %struct.__neon_int4x16x2_t = type { <4 x i16>, <4 x i16> }
63 %struct.__neon_int2x32x2_t = type { <2 x i32>, <2 x i32> }
65 define <8 x i8> @vld2dupi8(i8* %A) nounwind {
67 ;Check the (default) alignment value.
68 ;CHECK: vld2.8 {d16[], d17[]}, [r0]
69 %tmp0 = tail call %struct.__neon_int8x8x2_t @llvm.arm.neon.vld2lane.v8i8(i8* %A, <8 x i8> undef, <8 x i8> undef, i32 0, i32 1)
70 %tmp1 = extractvalue %struct.__neon_int8x8x2_t %tmp0, 0
71 %tmp2 = shufflevector <8 x i8> %tmp1, <8 x i8> undef, <8 x i32> zeroinitializer
72 %tmp3 = extractvalue %struct.__neon_int8x8x2_t %tmp0, 1
73 %tmp4 = shufflevector <8 x i8> %tmp3, <8 x i8> undef, <8 x i32> zeroinitializer
74 %tmp5 = add <8 x i8> %tmp2, %tmp4
78 define <4 x i16> @vld2dupi16(i16* %A) nounwind {
80 ;Check that a power-of-two alignment smaller than the total size of the memory
81 ;being loaded is ignored.
82 ;CHECK: vld2.16 {d16[], d17[]}, [r0]
83 %tmp0 = tail call %struct.__neon_int4x16x2_t @llvm.arm.neon.vld2lane.v4i16(i16* %A, <4 x i16> undef, <4 x i16> undef, i32 0, i32 2)
84 %tmp1 = extractvalue %struct.__neon_int4x16x2_t %tmp0, 0
85 %tmp2 = shufflevector <4 x i16> %tmp1, <4 x i16> undef, <4 x i32> zeroinitializer
86 %tmp3 = extractvalue %struct.__neon_int4x16x2_t %tmp0, 1
87 %tmp4 = shufflevector <4 x i16> %tmp3, <4 x i16> undef, <4 x i32> zeroinitializer
88 %tmp5 = add <4 x i16> %tmp2, %tmp4
92 ;Check for a post-increment updating load.
93 define <4 x i16> @vld2dupi16_update(i16** %ptr) nounwind {
94 ;CHECK: vld2dupi16_update:
95 ;CHECK: vld2.16 {d16[], d17[]}, [r1]!
97 %tmp0 = tail call %struct.__neon_int4x16x2_t @llvm.arm.neon.vld2lane.v4i16(i16* %A, <4 x i16> undef, <4 x i16> undef, i32 0, i32 2)
98 %tmp1 = extractvalue %struct.__neon_int4x16x2_t %tmp0, 0
99 %tmp2 = shufflevector <4 x i16> %tmp1, <4 x i16> undef, <4 x i32> zeroinitializer
100 %tmp3 = extractvalue %struct.__neon_int4x16x2_t %tmp0, 1
101 %tmp4 = shufflevector <4 x i16> %tmp3, <4 x i16> undef, <4 x i32> zeroinitializer
102 %tmp5 = add <4 x i16> %tmp2, %tmp4
103 %tmp6 = getelementptr i16* %A, i32 2
104 store i16* %tmp6, i16** %ptr
108 define <2 x i32> @vld2dupi32(i32* %A) nounwind {
110 ;Check the alignment value. Max for this instruction is 64 bits:
111 ;CHECK: vld2.32 {d16[], d17[]}, [r0, :64]
112 %tmp0 = tail call %struct.__neon_int2x32x2_t @llvm.arm.neon.vld2lane.v2i32(i32* %A, <2 x i32> undef, <2 x i32> undef, i32 0, i32 16)
113 %tmp1 = extractvalue %struct.__neon_int2x32x2_t %tmp0, 0
114 %tmp2 = shufflevector <2 x i32> %tmp1, <2 x i32> undef, <2 x i32> zeroinitializer
115 %tmp3 = extractvalue %struct.__neon_int2x32x2_t %tmp0, 1
116 %tmp4 = shufflevector <2 x i32> %tmp3, <2 x i32> undef, <2 x i32> zeroinitializer
117 %tmp5 = add <2 x i32> %tmp2, %tmp4
121 declare %struct.__neon_int8x8x2_t @llvm.arm.neon.vld2lane.v8i8(i8*, <8 x i8>, <8 x i8>, i32, i32) nounwind readonly
122 declare %struct.__neon_int4x16x2_t @llvm.arm.neon.vld2lane.v4i16(i16*, <4 x i16>, <4 x i16>, i32, i32) nounwind readonly
123 declare %struct.__neon_int2x32x2_t @llvm.arm.neon.vld2lane.v2i32(i32*, <2 x i32>, <2 x i32>, i32, i32) nounwind readonly
125 %struct.__neon_int8x8x3_t = type { <8 x i8>, <8 x i8>, <8 x i8> }
126 %struct.__neon_int16x4x3_t = type { <4 x i16>, <4 x i16>, <4 x i16> }
128 ;Check for a post-increment updating load with register increment.
129 define <8 x i8> @vld3dupi8_update(i8** %ptr, i32 %inc) nounwind {
130 ;CHECK: vld3dupi8_update:
131 ;CHECK: vld3.8 {d16[], d17[], d18[]}, [r2], r1
133 %tmp0 = tail call %struct.__neon_int8x8x3_t @llvm.arm.neon.vld3lane.v8i8(i8* %A, <8 x i8> undef, <8 x i8> undef, <8 x i8> undef, i32 0, i32 8)
134 %tmp1 = extractvalue %struct.__neon_int8x8x3_t %tmp0, 0
135 %tmp2 = shufflevector <8 x i8> %tmp1, <8 x i8> undef, <8 x i32> zeroinitializer
136 %tmp3 = extractvalue %struct.__neon_int8x8x3_t %tmp0, 1
137 %tmp4 = shufflevector <8 x i8> %tmp3, <8 x i8> undef, <8 x i32> zeroinitializer
138 %tmp5 = extractvalue %struct.__neon_int8x8x3_t %tmp0, 2
139 %tmp6 = shufflevector <8 x i8> %tmp5, <8 x i8> undef, <8 x i32> zeroinitializer
140 %tmp7 = add <8 x i8> %tmp2, %tmp4
141 %tmp8 = add <8 x i8> %tmp7, %tmp6
142 %tmp9 = getelementptr i8* %A, i32 %inc
143 store i8* %tmp9, i8** %ptr
147 define <4 x i16> @vld3dupi16(i16* %A) nounwind {
149 ;Check the (default) alignment value. VLD3 does not support alignment.
150 ;CHECK: vld3.16 {d16[], d17[], d18[]}, [r0]
151 %tmp0 = tail call %struct.__neon_int16x4x3_t @llvm.arm.neon.vld3lane.v4i16(i16* %A, <4 x i16> undef, <4 x i16> undef, <4 x i16> undef, i32 0, i32 8)
152 %tmp1 = extractvalue %struct.__neon_int16x4x3_t %tmp0, 0
153 %tmp2 = shufflevector <4 x i16> %tmp1, <4 x i16> undef, <4 x i32> zeroinitializer
154 %tmp3 = extractvalue %struct.__neon_int16x4x3_t %tmp0, 1
155 %tmp4 = shufflevector <4 x i16> %tmp3, <4 x i16> undef, <4 x i32> zeroinitializer
156 %tmp5 = extractvalue %struct.__neon_int16x4x3_t %tmp0, 2
157 %tmp6 = shufflevector <4 x i16> %tmp5, <4 x i16> undef, <4 x i32> zeroinitializer
158 %tmp7 = add <4 x i16> %tmp2, %tmp4
159 %tmp8 = add <4 x i16> %tmp7, %tmp6
163 declare %struct.__neon_int8x8x3_t @llvm.arm.neon.vld3lane.v8i8(i8*, <8 x i8>, <8 x i8>, <8 x i8>, i32, i32) nounwind readonly
164 declare %struct.__neon_int16x4x3_t @llvm.arm.neon.vld3lane.v4i16(i16*, <4 x i16>, <4 x i16>, <4 x i16>, i32, i32) nounwind readonly
166 %struct.__neon_int16x4x4_t = type { <4 x i16>, <4 x i16>, <4 x i16>, <4 x i16> }
167 %struct.__neon_int32x2x4_t = type { <2 x i32>, <2 x i32>, <2 x i32>, <2 x i32> }
169 ;Check for a post-increment updating load.
170 define <4 x i16> @vld4dupi16_update(i16** %ptr) nounwind {
171 ;CHECK: vld4dupi16_update:
172 ;CHECK: vld4.16 {d16[], d17[], d18[], d19[]}, [r1]!
174 %tmp0 = tail call %struct.__neon_int16x4x4_t @llvm.arm.neon.vld4lane.v4i16(i16* %A, <4 x i16> undef, <4 x i16> undef, <4 x i16> undef, <4 x i16> undef, i32 0, i32 1)
175 %tmp1 = extractvalue %struct.__neon_int16x4x4_t %tmp0, 0
176 %tmp2 = shufflevector <4 x i16> %tmp1, <4 x i16> undef, <4 x i32> zeroinitializer
177 %tmp3 = extractvalue %struct.__neon_int16x4x4_t %tmp0, 1
178 %tmp4 = shufflevector <4 x i16> %tmp3, <4 x i16> undef, <4 x i32> zeroinitializer
179 %tmp5 = extractvalue %struct.__neon_int16x4x4_t %tmp0, 2
180 %tmp6 = shufflevector <4 x i16> %tmp5, <4 x i16> undef, <4 x i32> zeroinitializer
181 %tmp7 = extractvalue %struct.__neon_int16x4x4_t %tmp0, 3
182 %tmp8 = shufflevector <4 x i16> %tmp7, <4 x i16> undef, <4 x i32> zeroinitializer
183 %tmp9 = add <4 x i16> %tmp2, %tmp4
184 %tmp10 = add <4 x i16> %tmp6, %tmp8
185 %tmp11 = add <4 x i16> %tmp9, %tmp10
186 %tmp12 = getelementptr i16* %A, i32 4
187 store i16* %tmp12, i16** %ptr
191 define <2 x i32> @vld4dupi32(i32* %A) nounwind {
193 ;Check the alignment value. An 8-byte alignment is allowed here even though
194 ;it is smaller than the total size of the memory being loaded.
195 ;CHECK: vld4.32 {d16[], d17[], d18[], d19[]}, [r0, :64]
196 %tmp0 = tail call %struct.__neon_int32x2x4_t @llvm.arm.neon.vld4lane.v2i32(i32* %A, <2 x i32> undef, <2 x i32> undef, <2 x i32> undef, <2 x i32> undef, i32 0, i32 8)
197 %tmp1 = extractvalue %struct.__neon_int32x2x4_t %tmp0, 0
198 %tmp2 = shufflevector <2 x i32> %tmp1, <2 x i32> undef, <2 x i32> zeroinitializer
199 %tmp3 = extractvalue %struct.__neon_int32x2x4_t %tmp0, 1
200 %tmp4 = shufflevector <2 x i32> %tmp3, <2 x i32> undef, <2 x i32> zeroinitializer
201 %tmp5 = extractvalue %struct.__neon_int32x2x4_t %tmp0, 2
202 %tmp6 = shufflevector <2 x i32> %tmp5, <2 x i32> undef, <2 x i32> zeroinitializer
203 %tmp7 = extractvalue %struct.__neon_int32x2x4_t %tmp0, 3
204 %tmp8 = shufflevector <2 x i32> %tmp7, <2 x i32> undef, <2 x i32> zeroinitializer
205 %tmp9 = add <2 x i32> %tmp2, %tmp4
206 %tmp10 = add <2 x i32> %tmp6, %tmp8
207 %tmp11 = add <2 x i32> %tmp9, %tmp10
211 declare %struct.__neon_int16x4x4_t @llvm.arm.neon.vld4lane.v4i16(i16*, <4 x i16>, <4 x i16>, <4 x i16>, <4 x i16>, i32, i32) nounwind readonly
212 declare %struct.__neon_int32x2x4_t @llvm.arm.neon.vld4lane.v2i32(i32*, <2 x i32>, <2 x i32>, <2 x i32>, <2 x i32>, i32, i32) nounwind readonly