• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1; RUN: llc < %s -mtriple=aarch64-linux-gnu -mattr=-zcz                    | FileCheck %s -check-prefixes=ALL,NONEGP,NONEFP
2; RUN: llc < %s -mtriple=aarch64-linux-gnu -mattr=+zcz                    | FileCheck %s -check-prefixes=ALL,ZEROGP,ZEROFP
3; RUN: llc < %s -mtriple=aarch64-linux-gnu -mattr=+zcz -mattr=+fullfp16   | FileCheck %s -check-prefixes=ALL,ZEROGP,ZERO16
4; RUN: llc < %s -mtriple=aarch64-linux-gnu -mattr=+zcz-gp                 | FileCheck %s -check-prefixes=ALL,ZEROGP,NONEFP
5; RUN: llc < %s -mtriple=aarch64-linux-gnu -mattr=+zcz-fp                 | FileCheck %s -check-prefixes=ALL,NONEGP,ZEROFP
6; RUN: llc < %s -mtriple=arm64-apple-ios   -mcpu=cyclone                  | FileCheck %s -check-prefixes=ALL,ZEROGP,NONEFP
7; RUN: llc < %s -mtriple=arm64-linux-gnu   -mcpu=apple-a10                | FileCheck %s -check-prefixes=ALL,ZEROGP,ZEROFP
8; RUN: llc < %s -mtriple=arm64-apple-ios   -mcpu=cyclone -mattr=+fullfp16 | FileCheck %s -check-prefixes=ALL,ZEROGP,NONE16
9; RUN: llc < %s -mtriple=aarch64-linux-gnu -mcpu=exynos-m3                | FileCheck %s -check-prefixes=ALL,NONEGP,ZEROFP
10; RUN: llc < %s -mtriple=aarch64-linux-gnu -mcpu=kryo                     | FileCheck %s -check-prefixes=ALL,ZEROGP,ZEROFP
11; RUN: llc < %s -mtriple=aarch64-linux-gnu -mcpu=falkor                   | FileCheck %s -check-prefixes=ALL,ZEROGP,ZEROFP
12
13declare void @bar(half, float, double, <2 x double>)
14declare void @bari(i32, i32)
15declare void @barl(i64, i64)
16declare void @barf(float, float)
17
18define void @t1() nounwind ssp {
19entry:
20; ALL-LABEL: t1:
21; ALL-NOT: fmov
22; NONEFP: ldr h0,{{.*}}
23; NONEFP: fmov s1, wzr
24; NONEFP: fmov d2, xzr
25; NONEFP: movi{{(.16b)?}} v3{{(.2d)?}}, #0
26; NONE16: fmov h0, wzr
27; NONE16: fmov s1, wzr
28; NONE16: fmov d2, xzr
29; NONE16: movi{{(.16b)?}} v3{{(.2d)?}}, #0
30; ZEROFP-DAG: ldr h0,{{.*}}
31; ZEROFP-DAG: movi v{{[0-3]+}}.2d, #0
32; ZEROFP-DAG: movi v{{[0-3]+}}.2d, #0
33; ZEROFP-DAG: movi v{{[0-3]+}}.2d, #0
34; ZERO16: movi v{{[0-3]+}}.2d, #0
35; ZERO16: movi v{{[0-3]+}}.2d, #0
36; ZERO16: movi v{{[0-3]+}}.2d, #0
37; ZERO16: movi v{{[0-3]+}}.2d, #0
38  tail call void @bar(half 0.000000e+00, float 0.000000e+00, double 0.000000e+00, <2 x double> <double 0.000000e+00, double 0.000000e+00>) nounwind
39  ret void
40}
41
42define void @t2() nounwind ssp {
43entry:
44; ALL-LABEL: t2:
45; NONEGP: mov w0, wzr
46; NONEGP: mov w1, wzr
47; ZEROGP: mov w0, #0
48; ZEROGP: mov w1, #0
49  tail call void @bari(i32 0, i32 0) nounwind
50  ret void
51}
52
53define void @t3() nounwind ssp {
54entry:
55; ALL-LABEL: t3:
56; NONEGP: mov x0, xzr
57; NONEGP: mov x1, xzr
58; ZEROGP: mov x0, #0
59; ZEROGP: mov x1, #0
60  tail call void @barl(i64 0, i64 0) nounwind
61  ret void
62}
63
64define void @t4() nounwind ssp {
65; ALL-LABEL: t4:
66; NONEFP: fmov s{{[0-3]+}}, wzr
67; NONEFP: fmov s{{[0-3]+}}, wzr
68; ZEROFP: movi v{{[0-3]+}}.2d, #0
69; ZEROFP: movi v{{[0-3]+}}.2d, #0
70  tail call void @barf(float 0.000000e+00, float 0.000000e+00) nounwind
71  ret void
72}
73
74declare double @sin(double)
75
76; We used to produce spills+reloads for a Q register with zero cycle zeroing
77; enabled.
78; ALL-LABEL: foo:
79; ALL-NOT: str q{{[0-9]+}}
80; ALL-NOT: ldr q{{[0-9]+}}
81define double @foo(i32 %n) {
82entry:
83  br label %for.body
84
85for.body:
86  %phi0 = phi double [ 1.0, %entry ], [ %v0, %for.body ]
87  %i.076 = phi i32 [ 0, %entry ], [ %inc, %for.body ]
88  %conv21 = sitofp i32 %i.076 to double
89  %call = tail call fast double @sin(double %conv21)
90  %cmp.i = fcmp fast olt double %phi0, %call
91  %v0 = select i1 %cmp.i, double %call, double %phi0
92  %inc = add nuw nsw i32 %i.076, 1
93  %cmp = icmp slt i32 %inc, %n
94  br i1 %cmp, label %for.body, label %for.end
95
96for.end:
97  ret double %v0
98}
99
100define <2 x i64> @t6() {
101; ALL-LABEL: t6:
102; ALL: movi{{(.16b)?}} v0{{(.2d)?}}, #0
103  ret <2 x i64> zeroinitializer
104}
105
106define i1 @ti1() {
107entry:
108; ALL-LABEL: ti1:
109; NONEGP: mov w0, wzr
110; ZEROGP: mov w0, #0
111  ret i1 false
112}
113
114define i8 @ti8() {
115entry:
116; ALL-LABEL: ti8:
117; NONEGP: mov w0, wzr
118; ZEROGP: mov w0, #0
119  ret i8 0
120}
121
122define i16 @ti16() {
123entry:
124; ALL-LABEL: ti16:
125; NONEGP: mov w0, wzr
126 ; ZEROGP: mov w0, #0
127  ret i16 0
128}
129
130define i32 @ti32() {
131entry:
132; ALL-LABEL: ti32:
133; NONEGP: mov w0, wzr
134; ZEROGP: mov w0, #0
135  ret i32 0
136}
137
138define i64 @ti64() {
139entry:
140; ALL-LABEL: ti64:
141; NONEGP: mov x0, xzr
142; ZEROGP: mov x0, #0
143  ret i64 0
144}
145
146define float @tf32() {
147entry:
148; ALL-LABEL: tf32:
149; NONEFP: mov s0, wzr
150; ZEROFP: movi v0.2d, #0
151  ret float 0.0
152}
153
154define double @td64() {
155entry:
156; ALL-LABEL: td64:
157; NONEFP: mov d0, xzr
158; ZEROFP: movi v0.2d, #0
159  ret double 0.0
160}
161
162define <8 x i8> @tv8i8() {
163entry:
164; ALL-LABEL: tv8i8:
165; ALL: movi{{(.16b)?}} v0{{(.2d)?}}, #0
166  ret <8 x i8> <i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0>
167}
168
169define <4 x i16> @tv4i16() {
170entry:
171; ALL-LABEL: tv4i16:
172; ALL: movi{{(.16b)?}} v0{{(.2d)?}}, #0
173  ret <4 x i16> <i16 0, i16 0, i16 0, i16 0>
174}
175
176define <2 x i32> @tv2i32() {
177entry:
178; ALL-LABEL: tv2i32:
179; ALL: movi{{(.16b)?}} v0{{(.2d)?}}, #0
180  ret <2 x i32> <i32 0, i32 0>
181}
182
183define <2 x float> @tv2f32() {
184entry:
185; ALL-LABEL: tv2f32:
186; ALL: movi{{(.16b)?}} v0{{(.2d)?}}, #0
187  ret <2 x float> <float 0.0, float 0.0>
188}
189
190define <16 x i8> @tv16i8() {
191entry:
192; ALL-LABEL: tv16i8:
193; ALL: movi{{(.16b)?}} v0{{(.2d)?}}, #0
194  ret <16 x i8> <i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0>
195}
196
197define <8 x i16> @tv8i16() {
198entry:
199; ALL-LABEL: tv8i16:
200; ALL: movi{{(.16b)?}} v0{{(.2d)?}}, #0
201  ret <8 x i16> <i16 0, i16 0, i16 0, i16 0, i16 0, i16 0, i16 0, i16 0>
202}
203
204define <4 x i32> @tv4i32() {
205entry:
206; ALL-LABEL: tv4i32:
207; ALL: movi{{(.16b)?}} v0{{(.2d)?}}, #0
208  ret <4 x i32> <i32 0, i32 0, i32 0, i32 0>
209}
210
211define <2 x i64> @tv2i64() {
212entry:
213; ALL-LABEL: tv2i64:
214; ALL: movi{{(.16b)?}} v0{{(.2d)?}}, #0
215  ret <2 x i64> <i64 0, i64 0>
216}
217
218define <4 x float> @tv4f32() {
219entry:
220; ALL-LABEL: tv4f32:
221; ALL: movi{{(.16b)?}} v0{{(.2d)?}}, #0
222  ret <4 x float> <float 0.0, float 0.0, float 0.0, float 0.0>
223}
224
225define <2 x double> @tv2d64() {
226entry:
227; ALL-LABEL: tv2d64:
228; ALL: movi{{(.16b)?}} v0{{(.2d)?}}, #0
229  ret <2 x double> <double 0.0, double 0.0>
230}
231
232