1 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
2 ; RUN: llc -mtriple=thumbv8.1m.main -mattr=+mve.fp -verify-machineinstrs -o - %s | FileCheck %s
4 define arm_aapcs_vfpcc <8 x half> @test_vrndaq_m_f16(<8 x half> %inactive, <8 x half> %a, i16 zeroext %p) {
5 ; CHECK-LABEL: test_vrndaq_m_f16:
6 ; CHECK: @ %bb.0: @ %entry
7 ; CHECK-NEXT: vmsr p0, r0
9 ; CHECK-NEXT: vrintat.f16 q0, q1
12 %0 = zext i16 %p to i32
13 %1 = tail call <8 x i1> @llvm.arm.mve.pred.i2v.v8i1(i32 %0)
14 %2 = tail call <8 x half> @llvm.arm.mve.vrinta.predicated.v8f16.v8i1(<8 x half> %a, <8 x i1> %1, <8 x half> %inactive)
18 define arm_aapcs_vfpcc <4 x float> @test_vrndaq_m_f32(<4 x float> %inactive, <4 x float> %a, i16 zeroext %p) {
19 ; CHECK-LABEL: test_vrndaq_m_f32:
20 ; CHECK: @ %bb.0: @ %entry
21 ; CHECK-NEXT: vmsr p0, r0
23 ; CHECK-NEXT: vrintat.f32 q0, q1
26 %0 = zext i16 %p to i32
27 %1 = tail call <4 x i1> @llvm.arm.mve.pred.i2v.v4i1(i32 %0)
28 %2 = tail call <4 x float> @llvm.arm.mve.vrinta.predicated.v4f32.v4i1(<4 x float> %a, <4 x i1> %1, <4 x float> %inactive)
32 define arm_aapcs_vfpcc <8 x half> @test_vrndmq_m_f16(<8 x half> %inactive, <8 x half> %a, i16 zeroext %p) {
33 ; CHECK-LABEL: test_vrndmq_m_f16:
34 ; CHECK: @ %bb.0: @ %entry
35 ; CHECK-NEXT: vmsr p0, r0
37 ; CHECK-NEXT: vrintmt.f16 q0, q1
40 %0 = zext i16 %p to i32
41 %1 = tail call <8 x i1> @llvm.arm.mve.pred.i2v.v8i1(i32 %0)
42 %2 = tail call <8 x half> @llvm.arm.mve.vrintm.predicated.v8f16.v8i1(<8 x half> %a, <8 x i1> %1, <8 x half> %inactive)
46 define arm_aapcs_vfpcc <4 x float> @test_vrndmq_m_f32(<4 x float> %inactive, <4 x float> %a, i16 zeroext %p) {
47 ; CHECK-LABEL: test_vrndmq_m_f32:
48 ; CHECK: @ %bb.0: @ %entry
49 ; CHECK-NEXT: vmsr p0, r0
51 ; CHECK-NEXT: vrintmt.f32 q0, q1
54 %0 = zext i16 %p to i32
55 %1 = tail call <4 x i1> @llvm.arm.mve.pred.i2v.v4i1(i32 %0)
56 %2 = tail call <4 x float> @llvm.arm.mve.vrintm.predicated.v4f32.v4i1(<4 x float> %a, <4 x i1> %1, <4 x float> %inactive)
60 define arm_aapcs_vfpcc <8 x half> @test_vrndnq_m_f16(<8 x half> %inactive, <8 x half> %a, i16 zeroext %p) {
61 ; CHECK-LABEL: test_vrndnq_m_f16:
62 ; CHECK: @ %bb.0: @ %entry
63 ; CHECK-NEXT: vmsr p0, r0
65 ; CHECK-NEXT: vrintnt.f16 q0, q1
68 %0 = zext i16 %p to i32
69 %1 = tail call <8 x i1> @llvm.arm.mve.pred.i2v.v8i1(i32 %0)
70 %2 = tail call <8 x half> @llvm.arm.mve.vrintn.predicated.v8f16.v8i1(<8 x half> %a, <8 x i1> %1, <8 x half> %inactive)
74 define arm_aapcs_vfpcc <4 x float> @test_vrndnq_m_f32(<4 x float> %inactive, <4 x float> %a, i16 zeroext %p) {
75 ; CHECK-LABEL: test_vrndnq_m_f32:
76 ; CHECK: @ %bb.0: @ %entry
77 ; CHECK-NEXT: vmsr p0, r0
79 ; CHECK-NEXT: vrintnt.f32 q0, q1
82 %0 = zext i16 %p to i32
83 %1 = tail call <4 x i1> @llvm.arm.mve.pred.i2v.v4i1(i32 %0)
84 %2 = tail call <4 x float> @llvm.arm.mve.vrintn.predicated.v4f32.v4i1(<4 x float> %a, <4 x i1> %1, <4 x float> %inactive)
88 define arm_aapcs_vfpcc <8 x half> @test_vrndpq_m_f16(<8 x half> %inactive, <8 x half> %a, i16 zeroext %p) {
89 ; CHECK-LABEL: test_vrndpq_m_f16:
90 ; CHECK: @ %bb.0: @ %entry
91 ; CHECK-NEXT: vmsr p0, r0
93 ; CHECK-NEXT: vrintpt.f16 q0, q1
96 %0 = zext i16 %p to i32
97 %1 = tail call <8 x i1> @llvm.arm.mve.pred.i2v.v8i1(i32 %0)
98 %2 = tail call <8 x half> @llvm.arm.mve.vrintp.predicated.v8f16.v8i1(<8 x half> %a, <8 x i1> %1, <8 x half> %inactive)
102 define arm_aapcs_vfpcc <4 x float> @test_vrndpq_m_f32(<4 x float> %inactive, <4 x float> %a, i16 zeroext %p) {
103 ; CHECK-LABEL: test_vrndpq_m_f32:
104 ; CHECK: @ %bb.0: @ %entry
105 ; CHECK-NEXT: vmsr p0, r0
107 ; CHECK-NEXT: vrintpt.f32 q0, q1
110 %0 = zext i16 %p to i32
111 %1 = tail call <4 x i1> @llvm.arm.mve.pred.i2v.v4i1(i32 %0)
112 %2 = tail call <4 x float> @llvm.arm.mve.vrintp.predicated.v4f32.v4i1(<4 x float> %a, <4 x i1> %1, <4 x float> %inactive)
116 define arm_aapcs_vfpcc <8 x half> @test_vrndq_m_f16(<8 x half> %inactive, <8 x half> %a, i16 zeroext %p) {
117 ; CHECK-LABEL: test_vrndq_m_f16:
118 ; CHECK: @ %bb.0: @ %entry
119 ; CHECK-NEXT: vmsr p0, r0
121 ; CHECK-NEXT: vrintzt.f16 q0, q1
124 %0 = zext i16 %p to i32
125 %1 = tail call <8 x i1> @llvm.arm.mve.pred.i2v.v8i1(i32 %0)
126 %2 = tail call <8 x half> @llvm.arm.mve.vrintz.predicated.v8f16.v8i1(<8 x half> %a, <8 x i1> %1, <8 x half> %inactive)
130 define arm_aapcs_vfpcc <4 x float> @test_vrndq_m_f32(<4 x float> %inactive, <4 x float> %a, i16 zeroext %p) {
131 ; CHECK-LABEL: test_vrndq_m_f32:
132 ; CHECK: @ %bb.0: @ %entry
133 ; CHECK-NEXT: vmsr p0, r0
135 ; CHECK-NEXT: vrintzt.f32 q0, q1
138 %0 = zext i16 %p to i32
139 %1 = tail call <4 x i1> @llvm.arm.mve.pred.i2v.v4i1(i32 %0)
140 %2 = tail call <4 x float> @llvm.arm.mve.vrintz.predicated.v4f32.v4i1(<4 x float> %a, <4 x i1> %1, <4 x float> %inactive)
144 define arm_aapcs_vfpcc <8 x half> @test_vrndxq_m_f16(<8 x half> %inactive, <8 x half> %a, i16 zeroext %p) {
145 ; CHECK-LABEL: test_vrndxq_m_f16:
146 ; CHECK: @ %bb.0: @ %entry
147 ; CHECK-NEXT: vmsr p0, r0
149 ; CHECK-NEXT: vrintxt.f16 q0, q1
152 %0 = zext i16 %p to i32
153 %1 = tail call <8 x i1> @llvm.arm.mve.pred.i2v.v8i1(i32 %0)
154 %2 = tail call <8 x half> @llvm.arm.mve.vrintx.predicated.v8f16.v8i1(<8 x half> %a, <8 x i1> %1, <8 x half> %inactive)
158 define arm_aapcs_vfpcc <4 x float> @test_vrndxq_m_f32(<4 x float> %inactive, <4 x float> %a, i16 zeroext %p) {
159 ; CHECK-LABEL: test_vrndxq_m_f32:
160 ; CHECK: @ %bb.0: @ %entry
161 ; CHECK-NEXT: vmsr p0, r0
163 ; CHECK-NEXT: vrintxt.f32 q0, q1
166 %0 = zext i16 %p to i32
167 %1 = tail call <4 x i1> @llvm.arm.mve.pred.i2v.v4i1(i32 %0)
168 %2 = tail call <4 x float> @llvm.arm.mve.vrintx.predicated.v4f32.v4i1(<4 x float> %a, <4 x i1> %1, <4 x float> %inactive)
172 declare <8 x i1> @llvm.arm.mve.pred.i2v.v8i1(i32)
173 declare <4 x i1> @llvm.arm.mve.pred.i2v.v4i1(i32)
174 declare <8 x half> @llvm.arm.mve.vrinta.predicated.v8f16.v8i1(<8 x half>, <8 x i1>, <8 x half>)
175 declare <4 x float> @llvm.arm.mve.vrinta.predicated.v4f32.v4i1(<4 x float>, <4 x i1>, <4 x float>)
176 declare <8 x half> @llvm.arm.mve.vrintm.predicated.v8f16.v8i1(<8 x half>, <8 x i1>, <8 x half>)
177 declare <4 x float> @llvm.arm.mve.vrintm.predicated.v4f32.v4i1(<4 x float>, <4 x i1>, <4 x float>)
178 declare <8 x half> @llvm.arm.mve.vrintn.predicated.v8f16.v8i1(<8 x half>, <8 x i1>, <8 x half>)
179 declare <4 x float> @llvm.arm.mve.vrintn.predicated.v4f32.v4i1(<4 x float>, <4 x i1>, <4 x float>)
180 declare <8 x half> @llvm.arm.mve.vrintp.predicated.v8f16.v8i1(<8 x half>, <8 x i1>, <8 x half>)
181 declare <4 x float> @llvm.arm.mve.vrintp.predicated.v4f32.v4i1(<4 x float>, <4 x i1>, <4 x float>)
182 declare <8 x half> @llvm.arm.mve.vrintz.predicated.v8f16.v8i1(<8 x half>, <8 x i1>, <8 x half>)
183 declare <4 x float> @llvm.arm.mve.vrintz.predicated.v4f32.v4i1(<4 x float>, <4 x i1>, <4 x float>)
184 declare <8 x half> @llvm.arm.mve.vrintx.predicated.v8f16.v8i1(<8 x half>, <8 x i1>, <8 x half>)
185 declare <4 x float> @llvm.arm.mve.vrintx.predicated.v4f32.v4i1(<4 x float>, <4 x i1>, <4 x float>)