1 ; RUN: llc -mtriple=arm-eabi -float-abi=soft -mattr=+neon %s -o - | FileCheck %s
3 ; RUN: llc -mtriple=arm-eabi -float-abi=soft -mattr=+neon -regalloc=basic %s -o - \
6 define <8 x i8> @vld1i8(i8* %A) nounwind {
8 ;Check the alignment value. Max for this instruction is 64 bits:
9 ;CHECK: vld1.8 {d16}, [r0:64]
10 %tmp1 = call <8 x i8> @llvm.arm.neon.vld1.v8i8.p0i8(i8* %A, i32 16)
14 define <4 x i16> @vld1i16(i16* %A) nounwind {
15 ;CHECK-LABEL: vld1i16:
17 %tmp0 = bitcast i16* %A to i8*
18 %tmp1 = call <4 x i16> @llvm.arm.neon.vld1.v4i16.p0i8(i8* %tmp0, i32 1)
22 ;Check for a post-increment updating load.
23 define <4 x i16> @vld1i16_update(i16** %ptr) nounwind {
24 ;CHECK-LABEL: vld1i16_update:
25 ;CHECK: vld1.16 {d16}, [{{r[0-9]+}}]!
26 %A = load i16*, i16** %ptr
27 %tmp0 = bitcast i16* %A to i8*
28 %tmp1 = call <4 x i16> @llvm.arm.neon.vld1.v4i16.p0i8(i8* %tmp0, i32 1)
29 %tmp2 = getelementptr i16, i16* %A, i32 4
30 store i16* %tmp2, i16** %ptr
34 define <2 x i32> @vld1i32(i32* %A) nounwind {
35 ;CHECK-LABEL: vld1i32:
37 %tmp0 = bitcast i32* %A to i8*
38 %tmp1 = call <2 x i32> @llvm.arm.neon.vld1.v2i32.p0i8(i8* %tmp0, i32 1)
42 ;Check for a post-increment updating load with register increment.
43 define <2 x i32> @vld1i32_update(i32** %ptr, i32 %inc) nounwind {
44 ;CHECK-LABEL: vld1i32_update:
45 ;CHECK: vld1.32 {d16}, [{{r[0-9]+}}], {{r[0-9]+}}
46 %A = load i32*, i32** %ptr
47 %tmp0 = bitcast i32* %A to i8*
48 %tmp1 = call <2 x i32> @llvm.arm.neon.vld1.v2i32.p0i8(i8* %tmp0, i32 1)
49 %tmp2 = getelementptr i32, i32* %A, i32 %inc
50 store i32* %tmp2, i32** %ptr
54 define <2 x float> @vld1f(float* %A) nounwind {
57 %tmp0 = bitcast float* %A to i8*
58 %tmp1 = call <2 x float> @llvm.arm.neon.vld1.v2f32.p0i8(i8* %tmp0, i32 1)
62 define <1 x i64> @vld1i64(i64* %A) nounwind {
63 ;CHECK-LABEL: vld1i64:
65 %tmp0 = bitcast i64* %A to i8*
66 %tmp1 = call <1 x i64> @llvm.arm.neon.vld1.v1i64.p0i8(i8* %tmp0, i32 1)
70 define <16 x i8> @vld1Qi8(i8* %A) nounwind {
71 ;CHECK-LABEL: vld1Qi8:
72 ;Check the alignment value. Max for this instruction is 128 bits:
73 ;CHECK: vld1.8 {d16, d17}, [r0:64]
74 %tmp1 = call <16 x i8> @llvm.arm.neon.vld1.v16i8.p0i8(i8* %A, i32 8)
78 ;Check for a post-increment updating load.
79 define <16 x i8> @vld1Qi8_update(i8** %ptr) nounwind {
80 ;CHECK-LABEL: vld1Qi8_update:
81 ;CHECK: vld1.8 {d16, d17}, [{{r[0-9]+|lr}}:64]!
82 %A = load i8*, i8** %ptr
83 %tmp1 = call <16 x i8> @llvm.arm.neon.vld1.v16i8.p0i8(i8* %A, i32 8)
84 %tmp2 = getelementptr i8, i8* %A, i32 16
85 store i8* %tmp2, i8** %ptr
89 define <8 x i16> @vld1Qi16(i16* %A) nounwind {
90 ;CHECK-LABEL: vld1Qi16:
91 ;Check the alignment value. Max for this instruction is 128 bits:
92 ;CHECK: vld1.16 {d16, d17}, [r0:128]
93 %tmp0 = bitcast i16* %A to i8*
94 %tmp1 = call <8 x i16> @llvm.arm.neon.vld1.v8i16.p0i8(i8* %tmp0, i32 32)
98 define <4 x i32> @vld1Qi32(i32* %A) nounwind {
99 ;CHECK-LABEL: vld1Qi32:
101 %tmp0 = bitcast i32* %A to i8*
102 %tmp1 = call <4 x i32> @llvm.arm.neon.vld1.v4i32.p0i8(i8* %tmp0, i32 1)
106 define <4 x float> @vld1Qf(float* %A) nounwind {
107 ;CHECK-LABEL: vld1Qf:
109 %tmp0 = bitcast float* %A to i8*
110 %tmp1 = call <4 x float> @llvm.arm.neon.vld1.v4f32.p0i8(i8* %tmp0, i32 1)
111 ret <4 x float> %tmp1
114 define <2 x i64> @vld1Qi64(i64* %A) nounwind {
115 ;CHECK-LABEL: vld1Qi64:
117 %tmp0 = bitcast i64* %A to i8*
118 %tmp1 = call <2 x i64> @llvm.arm.neon.vld1.v2i64.p0i8(i8* %tmp0, i32 1)
122 define <2 x double> @vld1Qf64(double* %A) nounwind {
123 ;CHECK-LABEL: vld1Qf64:
125 %tmp0 = bitcast double* %A to i8*
126 %tmp1 = call <2 x double> @llvm.arm.neon.vld1.v2f64.p0i8(i8* %tmp0, i32 1)
127 ret <2 x double> %tmp1
130 declare <8 x i8> @llvm.arm.neon.vld1.v8i8.p0i8(i8*, i32) nounwind readonly
131 declare <4 x i16> @llvm.arm.neon.vld1.v4i16.p0i8(i8*, i32) nounwind readonly
132 declare <2 x i32> @llvm.arm.neon.vld1.v2i32.p0i8(i8*, i32) nounwind readonly
133 declare <2 x float> @llvm.arm.neon.vld1.v2f32.p0i8(i8*, i32) nounwind readonly
134 declare <1 x i64> @llvm.arm.neon.vld1.v1i64.p0i8(i8*, i32) nounwind readonly
136 declare <16 x i8> @llvm.arm.neon.vld1.v16i8.p0i8(i8*, i32) nounwind readonly
137 declare <8 x i16> @llvm.arm.neon.vld1.v8i16.p0i8(i8*, i32) nounwind readonly
138 declare <4 x i32> @llvm.arm.neon.vld1.v4i32.p0i8(i8*, i32) nounwind readonly
139 declare <4 x float> @llvm.arm.neon.vld1.v4f32.p0i8(i8*, i32) nounwind readonly
140 declare <2 x i64> @llvm.arm.neon.vld1.v2i64.p0i8(i8*, i32) nounwind readonly
141 declare <2 x double> @llvm.arm.neon.vld1.v2f64.p0i8(i8*, i32) nounwind readonly
144 ; Do not crash if the vld1 result is not used.
145 define void @unused_vld1_result() {
147 %0 = call <4 x float> @llvm.arm.neon.vld1.v4f32.p0i8(i8* undef, i32 1)
148 call void @llvm.trap()
152 declare void @llvm.trap() nounwind