|
150
|
1 // RUN: %clang_cc1 -triple arm64-apple-ios7 -target-feature +neon -ffreestanding -fallow-half-arguments-and-returns -S -o - -disable-O0-optnone -emit-llvm %s | opt -S -mem2reg | FileCheck %s
|
|
|
2
|
|
|
3 #include <arm_neon.h>
|
|
|
4
|
|
|
5 // vdupq_n_f64 -> dup.2d v0, v0[0]
|
|
|
6 //
|
|
207
|
7 // CHECK-LABEL: define{{.*}} <2 x double> @test_vdupq_n_f64(double %w) #0 {
|
|
150
|
8 // CHECK: [[VECINIT_I:%.*]] = insertelement <2 x double> undef, double %w, i32 0
|
|
|
9 // CHECK: [[VECINIT1_I:%.*]] = insertelement <2 x double> [[VECINIT_I]], double %w, i32 1
|
|
|
10 // CHECK: ret <2 x double> [[VECINIT1_I]]
|
|
|
11 float64x2_t test_vdupq_n_f64(float64_t w) {
|
|
|
12 return vdupq_n_f64(w);
|
|
|
13 }
|
|
|
14
|
|
|
15 // might as well test this while we're here
|
|
|
16 // vdupq_n_f32 -> dup.4s v0, v0[0]
|
|
207
|
17 // CHECK-LABEL: define{{.*}} <4 x float> @test_vdupq_n_f32(float %w) #0 {
|
|
150
|
18 // CHECK: [[VECINIT_I:%.*]] = insertelement <4 x float> undef, float %w, i32 0
|
|
|
19 // CHECK: [[VECINIT1_I:%.*]] = insertelement <4 x float> [[VECINIT_I]], float %w, i32 1
|
|
|
20 // CHECK: [[VECINIT2_I:%.*]] = insertelement <4 x float> [[VECINIT1_I]], float %w, i32 2
|
|
|
21 // CHECK: [[VECINIT3_I:%.*]] = insertelement <4 x float> [[VECINIT2_I]], float %w, i32 3
|
|
|
22 // CHECK: ret <4 x float> [[VECINIT3_I]]
|
|
|
23 float32x4_t test_vdupq_n_f32(float32_t w) {
|
|
|
24 return vdupq_n_f32(w);
|
|
|
25 }
|
|
|
26
|
|
|
27 // vdupq_lane_f64 -> dup.2d v0, v0[0]
|
|
|
28 // this was in <rdar://problem/11778405>, but had already been implemented,
|
|
|
29 // test anyway
|
|
207
|
30 // CHECK-LABEL: define{{.*}} <2 x double> @test_vdupq_lane_f64(<1 x double> %V) #0 {
|
|
173
|
31 // CHECK: [[TMP0:%.*]] = bitcast <1 x double> %V to <8 x i8>
|
|
|
32 // CHECK: [[TMP1:%.*]] = bitcast <8 x i8> [[TMP0]] to <1 x double>
|
|
|
33 // CHECK: [[SHUFFLE:%.*]] = shufflevector <1 x double> [[TMP1]], <1 x double> [[TMP1]], <2 x i32> zeroinitializer
|
|
150
|
34 // CHECK: ret <2 x double> [[SHUFFLE]]
|
|
|
35 float64x2_t test_vdupq_lane_f64(float64x1_t V) {
|
|
|
36 return vdupq_lane_f64(V, 0);
|
|
|
37 }
|
|
|
38
|
|
|
39 // vmovq_n_f64 -> dup Vd.2d,X0
|
|
|
40 // this wasn't in <rdar://problem/11778405>, but it was between the vdups
|
|
207
|
41 // CHECK-LABEL: define{{.*}} <2 x double> @test_vmovq_n_f64(double %w) #0 {
|
|
150
|
42 // CHECK: [[VECINIT_I:%.*]] = insertelement <2 x double> undef, double %w, i32 0
|
|
|
43 // CHECK: [[VECINIT1_I:%.*]] = insertelement <2 x double> [[VECINIT_I]], double %w, i32 1
|
|
|
44 // CHECK: ret <2 x double> [[VECINIT1_I]]
|
|
|
45 float64x2_t test_vmovq_n_f64(float64_t w) {
|
|
|
46 return vmovq_n_f64(w);
|
|
|
47 }
|
|
|
48
|
|
207
|
49 // CHECK-LABEL: define{{.*}} <4 x half> @test_vmov_n_f16(half* %a1) #1 {
|
|
150
|
50 // CHECK: [[TMP0:%.*]] = load half, half* %a1, align 2
|
|
|
51 // CHECK: [[VECINIT:%.*]] = insertelement <4 x half> undef, half [[TMP0]], i32 0
|
|
|
52 // CHECK: [[VECINIT1:%.*]] = insertelement <4 x half> [[VECINIT]], half [[TMP0]], i32 1
|
|
|
53 // CHECK: [[VECINIT2:%.*]] = insertelement <4 x half> [[VECINIT1]], half [[TMP0]], i32 2
|
|
|
54 // CHECK: [[VECINIT3:%.*]] = insertelement <4 x half> [[VECINIT2]], half [[TMP0]], i32 3
|
|
|
55 // CHECK: ret <4 x half> [[VECINIT3]]
|
|
|
56 float16x4_t test_vmov_n_f16(float16_t *a1) {
|
|
|
57 return vmov_n_f16(*a1);
|
|
|
58 }
|
|
|
59
|
|
|
60 /*
|
|
|
61 float64x1_t test_vmov_n_f64(float64_t a1) {
|
|
|
62 return vmov_n_f64(a1);
|
|
|
63 }
|
|
|
64 */
|
|
|
65
|
|
207
|
66 // CHECK-LABEL: define{{.*}} <8 x half> @test_vmovq_n_f16(half* %a1) #0 {
|
|
150
|
67 // CHECK: [[TMP0:%.*]] = load half, half* %a1, align 2
|
|
|
68 // CHECK: [[VECINIT:%.*]] = insertelement <8 x half> undef, half [[TMP0]], i32 0
|
|
|
69 // CHECK: [[VECINIT1:%.*]] = insertelement <8 x half> [[VECINIT]], half [[TMP0]], i32 1
|
|
|
70 // CHECK: [[VECINIT2:%.*]] = insertelement <8 x half> [[VECINIT1]], half [[TMP0]], i32 2
|
|
|
71 // CHECK: [[VECINIT3:%.*]] = insertelement <8 x half> [[VECINIT2]], half [[TMP0]], i32 3
|
|
|
72 // CHECK: [[VECINIT4:%.*]] = insertelement <8 x half> [[VECINIT3]], half [[TMP0]], i32 4
|
|
|
73 // CHECK: [[VECINIT5:%.*]] = insertelement <8 x half> [[VECINIT4]], half [[TMP0]], i32 5
|
|
|
74 // CHECK: [[VECINIT6:%.*]] = insertelement <8 x half> [[VECINIT5]], half [[TMP0]], i32 6
|
|
|
75 // CHECK: [[VECINIT7:%.*]] = insertelement <8 x half> [[VECINIT6]], half [[TMP0]], i32 7
|
|
|
76 // CHECK: ret <8 x half> [[VECINIT7]]
|
|
|
77 float16x8_t test_vmovq_n_f16(float16_t *a1) {
|
|
|
78 return vmovq_n_f16(*a1);
|
|
|
79 }
|
|
|
80
|
|
|
81 // CHECK: attributes #0 ={{.*}}"min-legal-vector-width"="128"
|
|
|
82 // CHECK: attributes #1 ={{.*}}"min-legal-vector-width"="64"
|
