1 ; RUN: llc < %s -mtriple=armv8 -mattr=+v8.1a | FileCheck %s
3 ;-----------------------------------------------------------------------------
6 declare <4 x i16> @llvm.arm.neon.vqrdmulh.v4i16(<4 x i16>, <4 x i16>)
7 declare <8 x i16> @llvm.arm.neon.vqrdmulh.v8i16(<8 x i16>, <8 x i16>)
8 declare <2 x i32> @llvm.arm.neon.vqrdmulh.v2i32(<2 x i32>, <2 x i32>)
9 declare <4 x i32> @llvm.arm.neon.vqrdmulh.v4i32(<4 x i32>, <4 x i32>)
11 declare <4 x i16> @llvm.arm.neon.vqadds.v4i16(<4 x i16>, <4 x i16>)
12 declare <8 x i16> @llvm.arm.neon.vqadds.v8i16(<8 x i16>, <8 x i16>)
13 declare <2 x i32> @llvm.arm.neon.vqadds.v2i32(<2 x i32>, <2 x i32>)
14 declare <4 x i32> @llvm.arm.neon.vqadds.v4i32(<4 x i32>, <4 x i32>)
16 declare <4 x i16> @llvm.arm.neon.vqsubs.v4i16(<4 x i16>, <4 x i16>)
17 declare <8 x i16> @llvm.arm.neon.vqsubs.v8i16(<8 x i16>, <8 x i16>)
18 declare <2 x i32> @llvm.arm.neon.vqsubs.v2i32(<2 x i32>, <2 x i32>)
19 declare <4 x i32> @llvm.arm.neon.vqsubs.v4i32(<4 x i32>, <4 x i32>)
21 define <4 x i16> @test_vqrdmlah_v4i16(<4 x i16> %acc, <4 x i16> %mhs, <4 x i16> %rhs) {
22 ; CHECK-LABEL: test_vqrdmlah_v4i16:
23 %prod = call <4 x i16> @llvm.arm.neon.vqrdmulh.v4i16(<4 x i16> %mhs, <4 x i16> %rhs)
24 %retval = call <4 x i16> @llvm.arm.neon.vqadds.v4i16(<4 x i16> %acc, <4 x i16> %prod)
25 ; CHECK: vqrdmlah.s16 {{d[0-9]+}}, {{d[0-9]+}}, {{d[0-9]+}}
29 define <8 x i16> @test_vqrdmlah_v8i16(<8 x i16> %acc, <8 x i16> %mhs, <8 x i16> %rhs) {
30 ; CHECK-LABEL: test_vqrdmlah_v8i16:
31 %prod = call <8 x i16> @llvm.arm.neon.vqrdmulh.v8i16(<8 x i16> %mhs, <8 x i16> %rhs)
32 %retval = call <8 x i16> @llvm.arm.neon.vqadds.v8i16(<8 x i16> %acc, <8 x i16> %prod)
33 ; CHECK: vqrdmlah.s16 {{q[0-9]+}}, {{q[0-9]+}}, {{q[0-9]+}}
37 define <2 x i32> @test_vqrdmlah_v2i32(<2 x i32> %acc, <2 x i32> %mhs, <2 x i32> %rhs) {
38 ; CHECK-LABEL: test_vqrdmlah_v2i32:
39 %prod = call <2 x i32> @llvm.arm.neon.vqrdmulh.v2i32(<2 x i32> %mhs, <2 x i32> %rhs)
40 %retval = call <2 x i32> @llvm.arm.neon.vqadds.v2i32(<2 x i32> %acc, <2 x i32> %prod)
41 ; CHECK: vqrdmlah.s32 {{d[0-9]+}}, {{d[0-9]+}}, {{d[0-9]+}}
45 define <4 x i32> @test_vqrdmlah_v4i32(<4 x i32> %acc, <4 x i32> %mhs, <4 x i32> %rhs) {
46 ; CHECK-LABEL: test_vqrdmlah_v4i32:
47 %prod = call <4 x i32> @llvm.arm.neon.vqrdmulh.v4i32(<4 x i32> %mhs, <4 x i32> %rhs)
48 %retval = call <4 x i32> @llvm.arm.neon.vqadds.v4i32(<4 x i32> %acc, <4 x i32> %prod)
49 ; CHECK: vqrdmlah.s32 {{q[0-9]+}}, {{q[0-9]+}}, {{q[0-9]+}}
53 define <4 x i16> @test_vqrdmlsh_v4i16(<4 x i16> %acc, <4 x i16> %mhs, <4 x i16> %rhs) {
54 ; CHECK-LABEL: test_vqrdmlsh_v4i16:
55 %prod = call <4 x i16> @llvm.arm.neon.vqrdmulh.v4i16(<4 x i16> %mhs, <4 x i16> %rhs)
56 %retval = call <4 x i16> @llvm.arm.neon.vqsubs.v4i16(<4 x i16> %acc, <4 x i16> %prod)
57 ; CHECK: vqrdmlsh.s16 {{d[0-9]+}}, {{d[0-9]+}}, {{d[0-9]+}}
61 define <8 x i16> @test_vqrdmlsh_v8i16(<8 x i16> %acc, <8 x i16> %mhs, <8 x i16> %rhs) {
62 ; CHECK-LABEL: test_vqrdmlsh_v8i16:
63 %prod = call <8 x i16> @llvm.arm.neon.vqrdmulh.v8i16(<8 x i16> %mhs, <8 x i16> %rhs)
64 %retval = call <8 x i16> @llvm.arm.neon.vqsubs.v8i16(<8 x i16> %acc, <8 x i16> %prod)
65 ; CHECK: vqrdmlsh.s16 {{q[0-9]+}}, {{q[0-9]+}}, {{q[0-9]+}}
69 define <2 x i32> @test_vqrdmlsh_v2i32(<2 x i32> %acc, <2 x i32> %mhs, <2 x i32> %rhs) {
70 ; CHECK-LABEL: test_vqrdmlsh_v2i32:
71 %prod = call <2 x i32> @llvm.arm.neon.vqrdmulh.v2i32(<2 x i32> %mhs, <2 x i32> %rhs)
72 %retval = call <2 x i32> @llvm.arm.neon.vqsubs.v2i32(<2 x i32> %acc, <2 x i32> %prod)
73 ; CHECK: vqrdmlsh.s32 {{d[0-9]+}}, {{d[0-9]+}}, {{d[0-9]+}}
77 define <4 x i32> @test_vqrdmlsh_v4i32(<4 x i32> %acc, <4 x i32> %mhs, <4 x i32> %rhs) {
78 ; CHECK-LABEL: test_vqrdmlsh_v4i32:
79 %prod = call <4 x i32> @llvm.arm.neon.vqrdmulh.v4i32(<4 x i32> %mhs, <4 x i32> %rhs)
80 %retval = call <4 x i32> @llvm.arm.neon.vqsubs.v4i32(<4 x i32> %acc, <4 x i32> %prod)
81 ; CHECK: vqrdmlsh.s32 {{q[0-9]+}}, {{q[0-9]+}}, {{q[0-9]+}}
85 ;-----------------------------------------------------------------------------
88 define <4 x i16> @test_vqrdmlah_lane_s16(<4 x i16> %acc, <4 x i16> %x, <4 x i16> %v) {
89 ; CHECK-LABEL: test_vqrdmlah_lane_s16:
91 %shuffle = shufflevector <4 x i16> %v, <4 x i16> undef, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
92 %prod = call <4 x i16> @llvm.arm.neon.vqrdmulh.v4i16(<4 x i16> %x, <4 x i16> %shuffle)
93 %retval = call <4 x i16> @llvm.arm.neon.vqadds.v4i16(<4 x i16> %acc, <4 x i16> %prod)
94 ; CHECK: vqrdmlah.s16 {{d[0-9]+}}, {{d[0-9]+}}, {{d[0-9]+}}[3]
98 define <8 x i16> @test_vqrdmlahq_lane_s16(<8 x i16> %acc, <8 x i16> %x, <4 x i16> %v) {
99 ; CHECK-LABEL: test_vqrdmlahq_lane_s16:
101 %shuffle = shufflevector <4 x i16> %v, <4 x i16> undef, <8 x i32> <i32 2, i32 2, i32 2, i32 2, i32 2, i32 2, i32 2, i32 2>
102 %prod = call <8 x i16> @llvm.arm.neon.vqrdmulh.v8i16(<8 x i16> %x, <8 x i16> %shuffle)
103 %retval = call <8 x i16> @llvm.arm.neon.vqadds.v8i16(<8 x i16> %acc, <8 x i16> %prod)
104 ; CHECK: vqrdmlah.s16 {{q[0-9]+}}, {{q[0-9]+}}, {{d[0-9]+}}[2]
105 ret <8 x i16> %retval
108 define <2 x i32> @test_vqrdmlah_lane_s32(<2 x i32> %acc, <2 x i32> %x, <2 x i32> %v) {
109 ; CHECK-LABEL: test_vqrdmlah_lane_s32:
111 %shuffle = shufflevector <2 x i32> %v, <2 x i32> undef, <2 x i32> <i32 1, i32 1>
112 %prod = tail call <2 x i32> @llvm.arm.neon.vqrdmulh.v2i32(<2 x i32> %x, <2 x i32> %shuffle)
113 %retval = call <2 x i32> @llvm.arm.neon.vqadds.v2i32(<2 x i32> %acc, <2 x i32> %prod)
114 ; CHECK: vqrdmlah.s32 {{d[0-9]+}}, {{d[0-9]+}}, {{d[0-9]+}}[1]
115 ret <2 x i32> %retval
118 define <4 x i32> @test_vqrdmlahq_lane_s32(<4 x i32> %acc,<4 x i32> %x, <2 x i32> %v) {
119 ; CHECK-LABEL: test_vqrdmlahq_lane_s32:
121 %shuffle = shufflevector <2 x i32> %v, <2 x i32> undef, <4 x i32> zeroinitializer
122 %prod = tail call <4 x i32> @llvm.arm.neon.vqrdmulh.v4i32(<4 x i32> %x, <4 x i32> %shuffle)
123 %retval = call <4 x i32> @llvm.arm.neon.vqadds.v4i32(<4 x i32> %acc, <4 x i32> %prod)
124 ; CHECK: vqrdmlah.s32 {{q[0-9]+}}, {{q[0-9]+}}, {{d[0-9]+}}[0]
125 ret <4 x i32> %retval
128 define <4 x i16> @test_vqrdmlsh_lane_s16(<4 x i16> %acc, <4 x i16> %x, <4 x i16> %v) {
129 ; CHECK-LABEL: test_vqrdmlsh_lane_s16:
131 %shuffle = shufflevector <4 x i16> %v, <4 x i16> undef, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
132 %prod = call <4 x i16> @llvm.arm.neon.vqrdmulh.v4i16(<4 x i16> %x, <4 x i16> %shuffle)
133 %retval = call <4 x i16> @llvm.arm.neon.vqsubs.v4i16(<4 x i16> %acc, <4 x i16> %prod)
134 ; CHECK: vqrdmlsh.s16 {{d[0-9]+}}, {{d[0-9]+}}, {{d[0-9]+}}[3]
135 ret <4 x i16> %retval
138 define <8 x i16> @test_vqrdmlshq_lane_s16(<8 x i16> %acc, <8 x i16> %x, <4 x i16> %v) {
139 ; CHECK-LABEL: test_vqrdmlshq_lane_s16:
141 %shuffle = shufflevector <4 x i16> %v, <4 x i16> undef, <8 x i32> <i32 2, i32 2, i32 2, i32 2, i32 2, i32 2, i32 2, i32 2>
142 %prod = call <8 x i16> @llvm.arm.neon.vqrdmulh.v8i16(<8 x i16> %x, <8 x i16> %shuffle)
143 %retval = call <8 x i16> @llvm.arm.neon.vqsubs.v8i16(<8 x i16> %acc, <8 x i16> %prod)
144 ; CHECK: vqrdmlsh.s16 {{q[0-9]+}}, {{q[0-9]+}}, {{d[0-9]+}}[2]
145 ret <8 x i16> %retval
148 define <2 x i32> @test_vqrdmlsh_lane_s32(<2 x i32> %acc, <2 x i32> %x, <2 x i32> %v) {
149 ; CHECK-LABEL: test_vqrdmlsh_lane_s32:
151 %shuffle = shufflevector <2 x i32> %v, <2 x i32> undef, <2 x i32> <i32 1, i32 1>
152 %prod = tail call <2 x i32> @llvm.arm.neon.vqrdmulh.v2i32(<2 x i32> %x, <2 x i32> %shuffle)
153 %retval = call <2 x i32> @llvm.arm.neon.vqsubs.v2i32(<2 x i32> %acc, <2 x i32> %prod)
154 ; CHECK: vqrdmlsh.s32 {{d[0-9]+}}, {{d[0-9]+}}, {{d[0-9]+}}[1]
155 ret <2 x i32> %retval
158 define <4 x i32> @test_vqrdmlshq_lane_s32(<4 x i32> %acc,<4 x i32> %x, <2 x i32> %v) {
159 ; CHECK-LABEL: test_vqrdmlshq_lane_s32:
161 %shuffle = shufflevector <2 x i32> %v, <2 x i32> undef, <4 x i32> zeroinitializer
162 %prod = tail call <4 x i32> @llvm.arm.neon.vqrdmulh.v4i32(<4 x i32> %x, <4 x i32> %shuffle)
163 %retval = call <4 x i32> @llvm.arm.neon.vqsubs.v4i32(<4 x i32> %acc, <4 x i32> %prod)
164 ; CHECK: vqrdmlsh.s32 {{q[0-9]+}}, {{q[0-9]+}}, {{d[0-9]+}}[0]
165 ret <4 x i32> %retval