1; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
2; RUN: llc < %s -mtriple=aarch64-- | FileCheck %s
3
4; PR38527 - https://bugs.llvm.org/show_bug.cgi?id=38527
5
6; Use fsin as the representative test for various data types.
7
8declare <1 x float> @llvm.sin.v1f32(<1 x float>)
9declare <2 x float> @llvm.sin.v2f32(<2 x float>)
10declare <3 x float> @llvm.sin.v3f32(<3 x float>)
11declare <4 x float> @llvm.sin.v4f32(<4 x float>)
12declare <5 x float> @llvm.sin.v5f32(<5 x float>)
13declare <6 x float> @llvm.sin.v6f32(<6 x float>)
14declare <3 x double> @llvm.sin.v3f64(<3 x double>)
15
16; Verify that all of the potential libcall candidates are handled.
17; Some of these have custom lowering, so those cases won't have
18; libcalls.
19
20declare <3 x float> @llvm.fabs.v3f32(<3 x float>)
21declare <3 x float> @llvm.ceil.v3f32(<3 x float>)
22declare <3 x float> @llvm.cos.v3f32(<3 x float>)
23declare <3 x float> @llvm.exp.v3f32(<3 x float>)
24declare <3 x float> @llvm.exp2.v3f32(<3 x float>)
25declare <3 x float> @llvm.floor.v3f32(<3 x float>)
26declare <3 x float> @llvm.log.v3f32(<3 x float>)
27declare <3 x float> @llvm.log10.v3f32(<3 x float>)
28declare <3 x float> @llvm.log2.v3f32(<3 x float>)
29declare <3 x float> @llvm.nearbyint.v3f32(<3 x float>)
30declare <3 x float> @llvm.rint.v3f32(<3 x float>)
31declare <3 x float> @llvm.round.v3f32(<3 x float>)
32declare <3 x float> @llvm.roundeven.v3f32(<3 x float>)
33declare <3 x float> @llvm.sqrt.v3f32(<3 x float>)
34declare <3 x float> @llvm.trunc.v3f32(<3 x float>)
35
36define <1 x float> @sin_v1f32(<1 x float> %x) nounwind {
37; CHECK-LABEL: sin_v1f32:
38; CHECK:       // %bb.0:
39; CHECK-NEXT:    str x30, [sp, #-16]! // 8-byte Folded Spill
40; CHECK-NEXT:    // kill: def $d0 killed $d0 def $q0
41; CHECK-NEXT:    // kill: def $s0 killed $s0 killed $q0
42; CHECK-NEXT:    bl sinf
43; CHECK-NEXT:    // kill: def $s0 killed $s0 def $d0
44; CHECK-NEXT:    ldr x30, [sp], #16 // 8-byte Folded Reload
45; CHECK-NEXT:    ret
46  %r = call <1 x float> @llvm.sin.v1f32(<1 x float> %x)
47  ret <1 x float> %r
48}
49
50define <2 x float> @sin_v2f32(<2 x float> %x) nounwind {
51; CHECK-LABEL: sin_v2f32:
52; CHECK:       // %bb.0:
53; CHECK-NEXT:    sub sp, sp, #48
54; CHECK-NEXT:    // kill: def $d0 killed $d0 def $q0
55; CHECK-NEXT:    str q0, [sp] // 16-byte Folded Spill
56; CHECK-NEXT:    mov s0, v0.s[1]
57; CHECK-NEXT:    str x30, [sp, #32] // 8-byte Folded Spill
58; CHECK-NEXT:    bl sinf
59; CHECK-NEXT:    str d0, [sp, #16] // 16-byte Folded Spill
60; CHECK-NEXT:    ldr q0, [sp] // 16-byte Folded Reload
61; CHECK-NEXT:    // kill: def $s0 killed $s0 killed $q0
62; CHECK-NEXT:    bl sinf
63; CHECK-NEXT:    ldr q1, [sp, #16] // 16-byte Folded Reload
64; CHECK-NEXT:    // kill: def $s0 killed $s0 def $q0
65; CHECK-NEXT:    ldr x30, [sp, #32] // 8-byte Folded Reload
66; CHECK-NEXT:    mov v0.s[1], v1.s[0]
67; CHECK-NEXT:    // kill: def $d0 killed $d0 killed $q0
68; CHECK-NEXT:    add sp, sp, #48
69; CHECK-NEXT:    ret
70  %r = call <2 x float> @llvm.sin.v2f32(<2 x float> %x)
71  ret <2 x float> %r
72}
73
74define <3 x float> @sin_v3f32(<3 x float> %x) nounwind {
75; CHECK-LABEL: sin_v3f32:
76; CHECK:       // %bb.0:
77; CHECK-NEXT:    sub sp, sp, #48
78; CHECK-NEXT:    str q0, [sp, #16] // 16-byte Folded Spill
79; CHECK-NEXT:    mov s0, v0.s[1]
80; CHECK-NEXT:    str x30, [sp, #32] // 8-byte Folded Spill
81; CHECK-NEXT:    bl sinf
82; CHECK-NEXT:    str d0, [sp] // 16-byte Folded Spill
83; CHECK-NEXT:    ldr q0, [sp, #16] // 16-byte Folded Reload
84; CHECK-NEXT:    // kill: def $s0 killed $s0 killed $q0
85; CHECK-NEXT:    bl sinf
86; CHECK-NEXT:    ldr q1, [sp] // 16-byte Folded Reload
87; CHECK-NEXT:    // kill: def $s0 killed $s0 def $q0
88; CHECK-NEXT:    mov v0.s[1], v1.s[0]
89; CHECK-NEXT:    str q0, [sp] // 16-byte Folded Spill
90; CHECK-NEXT:    ldr q0, [sp, #16] // 16-byte Folded Reload
91; CHECK-NEXT:    mov s0, v0.s[2]
92; CHECK-NEXT:    bl sinf
93; CHECK-NEXT:    ldr q1, [sp] // 16-byte Folded Reload
94; CHECK-NEXT:    // kill: def $s0 killed $s0 def $q0
95; CHECK-NEXT:    ldr x30, [sp, #32] // 8-byte Folded Reload
96; CHECK-NEXT:    mov v1.s[2], v0.s[0]
97; CHECK-NEXT:    mov v0.16b, v1.16b
98; CHECK-NEXT:    add sp, sp, #48
99; CHECK-NEXT:    ret
100  %r = call <3 x float> @llvm.sin.v3f32(<3 x float> %x)
101  ret <3 x float> %r
102}
103
104define <4 x float> @sin_v4f32(<4 x float> %x) nounwind {
105; CHECK-LABEL: sin_v4f32:
106; CHECK:       // %bb.0:
107; CHECK-NEXT:    sub sp, sp, #48
108; CHECK-NEXT:    str q0, [sp, #16] // 16-byte Folded Spill
109; CHECK-NEXT:    mov s0, v0.s[1]
110; CHECK-NEXT:    str x30, [sp, #32] // 8-byte Folded Spill
111; CHECK-NEXT:    bl sinf
112; CHECK-NEXT:    str d0, [sp] // 16-byte Folded Spill
113; CHECK-NEXT:    ldr q0, [sp, #16] // 16-byte Folded Reload
114; CHECK-NEXT:    // kill: def $s0 killed $s0 killed $q0
115; CHECK-NEXT:    bl sinf
116; CHECK-NEXT:    ldr q1, [sp] // 16-byte Folded Reload
117; CHECK-NEXT:    // kill: def $s0 killed $s0 def $q0
118; CHECK-NEXT:    mov v0.s[1], v1.s[0]
119; CHECK-NEXT:    str q0, [sp] // 16-byte Folded Spill
120; CHECK-NEXT:    ldr q0, [sp, #16] // 16-byte Folded Reload
121; CHECK-NEXT:    mov s0, v0.s[2]
122; CHECK-NEXT:    bl sinf
123; CHECK-NEXT:    ldr q1, [sp] // 16-byte Folded Reload
124; CHECK-NEXT:    // kill: def $s0 killed $s0 def $q0
125; CHECK-NEXT:    mov v1.s[2], v0.s[0]
126; CHECK-NEXT:    ldr q0, [sp, #16] // 16-byte Folded Reload
127; CHECK-NEXT:    mov s0, v0.s[3]
128; CHECK-NEXT:    str q1, [sp] // 16-byte Folded Spill
129; CHECK-NEXT:    bl sinf
130; CHECK-NEXT:    ldr q1, [sp] // 16-byte Folded Reload
131; CHECK-NEXT:    // kill: def $s0 killed $s0 def $q0
132; CHECK-NEXT:    ldr x30, [sp, #32] // 8-byte Folded Reload
133; CHECK-NEXT:    mov v1.s[3], v0.s[0]
134; CHECK-NEXT:    mov v0.16b, v1.16b
135; CHECK-NEXT:    add sp, sp, #48
136; CHECK-NEXT:    ret
137  %r = call <4 x float> @llvm.sin.v4f32(<4 x float> %x)
138  ret <4 x float> %r
139}
140
141define <5 x float> @sin_v5f32(<5 x float> %x) nounwind {
142; CHECK-LABEL: sin_v5f32:
143; CHECK:       // %bb.0:
144; CHECK-NEXT:    str d12, [sp, #-48]! // 8-byte Folded Spill
145; CHECK-NEXT:    stp d11, d10, [sp, #8] // 16-byte Folded Spill
146; CHECK-NEXT:    fmov s10, s2
147; CHECK-NEXT:    stp d9, d8, [sp, #24] // 16-byte Folded Spill
148; CHECK-NEXT:    fmov s8, s4
149; CHECK-NEXT:    str x30, [sp, #40] // 8-byte Folded Spill
150; CHECK-NEXT:    fmov s9, s3
151; CHECK-NEXT:    fmov s11, s1
152; CHECK-NEXT:    bl sinf
153; CHECK-NEXT:    fmov s12, s0
154; CHECK-NEXT:    fmov s0, s11
155; CHECK-NEXT:    bl sinf
156; CHECK-NEXT:    fmov s11, s0
157; CHECK-NEXT:    fmov s0, s10
158; CHECK-NEXT:    bl sinf
159; CHECK-NEXT:    fmov s10, s0
160; CHECK-NEXT:    fmov s0, s9
161; CHECK-NEXT:    bl sinf
162; CHECK-NEXT:    fmov s9, s0
163; CHECK-NEXT:    fmov s0, s8
164; CHECK-NEXT:    bl sinf
165; CHECK-NEXT:    fmov s1, s11
166; CHECK-NEXT:    fmov s2, s10
167; CHECK-NEXT:    fmov s3, s9
168; CHECK-NEXT:    ldr x30, [sp, #40] // 8-byte Folded Reload
169; CHECK-NEXT:    ldp d9, d8, [sp, #24] // 16-byte Folded Reload
170; CHECK-NEXT:    fmov s4, s0
171; CHECK-NEXT:    fmov s0, s12
172; CHECK-NEXT:    ldp d11, d10, [sp, #8] // 16-byte Folded Reload
173; CHECK-NEXT:    ldr d12, [sp], #48 // 8-byte Folded Reload
174; CHECK-NEXT:    ret
175  %r = call <5 x float> @llvm.sin.v5f32(<5 x float> %x)
176  ret <5 x float> %r
177}
178
179define <6 x float> @sin_v6f32(<6 x float> %x) nounwind {
180; CHECK-LABEL: sin_v6f32:
181; CHECK:       // %bb.0:
182; CHECK-NEXT:    stp d13, d12, [sp, #-64]! // 16-byte Folded Spill
183; CHECK-NEXT:    stp d11, d10, [sp, #16] // 16-byte Folded Spill
184; CHECK-NEXT:    fmov s10, s3
185; CHECK-NEXT:    stp d9, d8, [sp, #32] // 16-byte Folded Spill
186; CHECK-NEXT:    fmov s8, s5
187; CHECK-NEXT:    str x30, [sp, #48] // 8-byte Folded Spill
188; CHECK-NEXT:    fmov s9, s4
189; CHECK-NEXT:    fmov s11, s2
190; CHECK-NEXT:    fmov s12, s1
191; CHECK-NEXT:    bl sinf
192; CHECK-NEXT:    fmov s13, s0
193; CHECK-NEXT:    fmov s0, s12
194; CHECK-NEXT:    bl sinf
195; CHECK-NEXT:    fmov s12, s0
196; CHECK-NEXT:    fmov s0, s11
197; CHECK-NEXT:    bl sinf
198; CHECK-NEXT:    fmov s11, s0
199; CHECK-NEXT:    fmov s0, s10
200; CHECK-NEXT:    bl sinf
201; CHECK-NEXT:    fmov s10, s0
202; CHECK-NEXT:    fmov s0, s9
203; CHECK-NEXT:    bl sinf
204; CHECK-NEXT:    fmov s9, s0
205; CHECK-NEXT:    fmov s0, s8
206; CHECK-NEXT:    bl sinf
207; CHECK-NEXT:    fmov s2, s11
208; CHECK-NEXT:    fmov s3, s10
209; CHECK-NEXT:    fmov s4, s9
210; CHECK-NEXT:    ldr x30, [sp, #48] // 8-byte Folded Reload
211; CHECK-NEXT:    ldp d9, d8, [sp, #32] // 16-byte Folded Reload
212; CHECK-NEXT:    fmov s5, s0
213; CHECK-NEXT:    fmov s0, s13
214; CHECK-NEXT:    ldp d11, d10, [sp, #16] // 16-byte Folded Reload
215; CHECK-NEXT:    fmov s1, s12
216; CHECK-NEXT:    ldp d13, d12, [sp], #64 // 16-byte Folded Reload
217; CHECK-NEXT:    ret
218  %r = call <6 x float> @llvm.sin.v6f32(<6 x float> %x)
219  ret <6 x float> %r
220}
221
222define <3 x double> @sin_v3f64(<3 x double> %x) nounwind {
223; CHECK-LABEL: sin_v3f64:
224; CHECK:       // %bb.0:
225; CHECK-NEXT:    str d10, [sp, #-32]! // 8-byte Folded Spill
226; CHECK-NEXT:    stp d9, d8, [sp, #8] // 16-byte Folded Spill
227; CHECK-NEXT:    fmov d8, d2
228; CHECK-NEXT:    str x30, [sp, #24] // 8-byte Folded Spill
229; CHECK-NEXT:    fmov d9, d1
230; CHECK-NEXT:    bl sin
231; CHECK-NEXT:    fmov d10, d0
232; CHECK-NEXT:    fmov d0, d9
233; CHECK-NEXT:    bl sin
234; CHECK-NEXT:    fmov d9, d0
235; CHECK-NEXT:    fmov d0, d8
236; CHECK-NEXT:    bl sin
237; CHECK-NEXT:    fmov d1, d9
238; CHECK-NEXT:    ldr x30, [sp, #24] // 8-byte Folded Reload
239; CHECK-NEXT:    ldp d9, d8, [sp, #8] // 16-byte Folded Reload
240; CHECK-NEXT:    fmov d2, d0
241; CHECK-NEXT:    fmov d0, d10
242; CHECK-NEXT:    ldr d10, [sp], #32 // 8-byte Folded Reload
243; CHECK-NEXT:    ret
244  %r = call <3 x double> @llvm.sin.v3f64(<3 x double> %x)
245  ret <3 x double> %r
246}
247
248define <3 x float> @fabs_v3f32(<3 x float> %x) nounwind {
249; CHECK-LABEL: fabs_v3f32:
250; CHECK:       // %bb.0:
251; CHECK-NEXT:    fabs v0.4s, v0.4s
252; CHECK-NEXT:    ret
253  %r = call <3 x float> @llvm.fabs.v3f32(<3 x float> %x)
254  ret <3 x float> %r
255}
256
257define <3 x float> @ceil_v3f32(<3 x float> %x) nounwind {
258; CHECK-LABEL: ceil_v3f32:
259; CHECK:       // %bb.0:
260; CHECK-NEXT:    frintp v0.4s, v0.4s
261; CHECK-NEXT:    ret
262  %r = call <3 x float> @llvm.ceil.v3f32(<3 x float> %x)
263  ret <3 x float> %r
264}
265
266define <3 x float> @cos_v3f32(<3 x float> %x) nounwind {
267; CHECK-LABEL: cos_v3f32:
268; CHECK:       // %bb.0:
269; CHECK-NEXT:    sub sp, sp, #48
270; CHECK-NEXT:    str q0, [sp, #16] // 16-byte Folded Spill
271; CHECK-NEXT:    mov s0, v0.s[1]
272; CHECK-NEXT:    str x30, [sp, #32] // 8-byte Folded Spill
273; CHECK-NEXT:    bl cosf
274; CHECK-NEXT:    str d0, [sp] // 16-byte Folded Spill
275; CHECK-NEXT:    ldr q0, [sp, #16] // 16-byte Folded Reload
276; CHECK-NEXT:    // kill: def $s0 killed $s0 killed $q0
277; CHECK-NEXT:    bl cosf
278; CHECK-NEXT:    ldr q1, [sp] // 16-byte Folded Reload
279; CHECK-NEXT:    // kill: def $s0 killed $s0 def $q0
280; CHECK-NEXT:    mov v0.s[1], v1.s[0]
281; CHECK-NEXT:    str q0, [sp] // 16-byte Folded Spill
282; CHECK-NEXT:    ldr q0, [sp, #16] // 16-byte Folded Reload
283; CHECK-NEXT:    mov s0, v0.s[2]
284; CHECK-NEXT:    bl cosf
285; CHECK-NEXT:    ldr q1, [sp] // 16-byte Folded Reload
286; CHECK-NEXT:    // kill: def $s0 killed $s0 def $q0
287; CHECK-NEXT:    ldr x30, [sp, #32] // 8-byte Folded Reload
288; CHECK-NEXT:    mov v1.s[2], v0.s[0]
289; CHECK-NEXT:    mov v0.16b, v1.16b
290; CHECK-NEXT:    add sp, sp, #48
291; CHECK-NEXT:    ret
292  %r = call <3 x float> @llvm.cos.v3f32(<3 x float> %x)
293  ret <3 x float> %r
294}
295
296define <3 x float> @exp_v3f32(<3 x float> %x) nounwind {
297; CHECK-LABEL: exp_v3f32:
298; CHECK:       // %bb.0:
299; CHECK-NEXT:    sub sp, sp, #48
300; CHECK-NEXT:    str q0, [sp, #16] // 16-byte Folded Spill
301; CHECK-NEXT:    mov s0, v0.s[1]
302; CHECK-NEXT:    str x30, [sp, #32] // 8-byte Folded Spill
303; CHECK-NEXT:    bl expf
304; CHECK-NEXT:    str d0, [sp] // 16-byte Folded Spill
305; CHECK-NEXT:    ldr q0, [sp, #16] // 16-byte Folded Reload
306; CHECK-NEXT:    // kill: def $s0 killed $s0 killed $q0
307; CHECK-NEXT:    bl expf
308; CHECK-NEXT:    ldr q1, [sp] // 16-byte Folded Reload
309; CHECK-NEXT:    // kill: def $s0 killed $s0 def $q0
310; CHECK-NEXT:    mov v0.s[1], v1.s[0]
311; CHECK-NEXT:    str q0, [sp] // 16-byte Folded Spill
312; CHECK-NEXT:    ldr q0, [sp, #16] // 16-byte Folded Reload
313; CHECK-NEXT:    mov s0, v0.s[2]
314; CHECK-NEXT:    bl expf
315; CHECK-NEXT:    ldr q1, [sp] // 16-byte Folded Reload
316; CHECK-NEXT:    // kill: def $s0 killed $s0 def $q0
317; CHECK-NEXT:    ldr x30, [sp, #32] // 8-byte Folded Reload
318; CHECK-NEXT:    mov v1.s[2], v0.s[0]
319; CHECK-NEXT:    mov v0.16b, v1.16b
320; CHECK-NEXT:    add sp, sp, #48
321; CHECK-NEXT:    ret
322  %r = call <3 x float> @llvm.exp.v3f32(<3 x float> %x)
323  ret <3 x float> %r
324}
325
326define <3 x float> @exp2_v3f32(<3 x float> %x) nounwind {
327; CHECK-LABEL: exp2_v3f32:
328; CHECK:       // %bb.0:
329; CHECK-NEXT:    sub sp, sp, #48
330; CHECK-NEXT:    str q0, [sp, #16] // 16-byte Folded Spill
331; CHECK-NEXT:    mov s0, v0.s[1]
332; CHECK-NEXT:    str x30, [sp, #32] // 8-byte Folded Spill
333; CHECK-NEXT:    bl exp2f
334; CHECK-NEXT:    str d0, [sp] // 16-byte Folded Spill
335; CHECK-NEXT:    ldr q0, [sp, #16] // 16-byte Folded Reload
336; CHECK-NEXT:    // kill: def $s0 killed $s0 killed $q0
337; CHECK-NEXT:    bl exp2f
338; CHECK-NEXT:    ldr q1, [sp] // 16-byte Folded Reload
339; CHECK-NEXT:    // kill: def $s0 killed $s0 def $q0
340; CHECK-NEXT:    mov v0.s[1], v1.s[0]
341; CHECK-NEXT:    str q0, [sp] // 16-byte Folded Spill
342; CHECK-NEXT:    ldr q0, [sp, #16] // 16-byte Folded Reload
343; CHECK-NEXT:    mov s0, v0.s[2]
344; CHECK-NEXT:    bl exp2f
345; CHECK-NEXT:    ldr q1, [sp] // 16-byte Folded Reload
346; CHECK-NEXT:    // kill: def $s0 killed $s0 def $q0
347; CHECK-NEXT:    ldr x30, [sp, #32] // 8-byte Folded Reload
348; CHECK-NEXT:    mov v1.s[2], v0.s[0]
349; CHECK-NEXT:    mov v0.16b, v1.16b
350; CHECK-NEXT:    add sp, sp, #48
351; CHECK-NEXT:    ret
352  %r = call <3 x float> @llvm.exp2.v3f32(<3 x float> %x)
353  ret <3 x float> %r
354}
355
356define <3 x float> @floor_v3f32(<3 x float> %x) nounwind {
357; CHECK-LABEL: floor_v3f32:
358; CHECK:       // %bb.0:
359; CHECK-NEXT:    frintm v0.4s, v0.4s
360; CHECK-NEXT:    ret
361  %r = call <3 x float> @llvm.floor.v3f32(<3 x float> %x)
362  ret <3 x float> %r
363}
364
365define <3 x float> @log_v3f32(<3 x float> %x) nounwind {
366; CHECK-LABEL: log_v3f32:
367; CHECK:       // %bb.0:
368; CHECK-NEXT:    sub sp, sp, #48
369; CHECK-NEXT:    str q0, [sp, #16] // 16-byte Folded Spill
370; CHECK-NEXT:    mov s0, v0.s[1]
371; CHECK-NEXT:    str x30, [sp, #32] // 8-byte Folded Spill
372; CHECK-NEXT:    bl logf
373; CHECK-NEXT:    str d0, [sp] // 16-byte Folded Spill
374; CHECK-NEXT:    ldr q0, [sp, #16] // 16-byte Folded Reload
375; CHECK-NEXT:    // kill: def $s0 killed $s0 killed $q0
376; CHECK-NEXT:    bl logf
377; CHECK-NEXT:    ldr q1, [sp] // 16-byte Folded Reload
378; CHECK-NEXT:    // kill: def $s0 killed $s0 def $q0
379; CHECK-NEXT:    mov v0.s[1], v1.s[0]
380; CHECK-NEXT:    str q0, [sp] // 16-byte Folded Spill
381; CHECK-NEXT:    ldr q0, [sp, #16] // 16-byte Folded Reload
382; CHECK-NEXT:    mov s0, v0.s[2]
383; CHECK-NEXT:    bl logf
384; CHECK-NEXT:    ldr q1, [sp] // 16-byte Folded Reload
385; CHECK-NEXT:    // kill: def $s0 killed $s0 def $q0
386; CHECK-NEXT:    ldr x30, [sp, #32] // 8-byte Folded Reload
387; CHECK-NEXT:    mov v1.s[2], v0.s[0]
388; CHECK-NEXT:    mov v0.16b, v1.16b
389; CHECK-NEXT:    add sp, sp, #48
390; CHECK-NEXT:    ret
391  %r = call <3 x float> @llvm.log.v3f32(<3 x float> %x)
392  ret <3 x float> %r
393}
394
395define <3 x float> @log10_v3f32(<3 x float> %x) nounwind {
396; CHECK-LABEL: log10_v3f32:
397; CHECK:       // %bb.0:
398; CHECK-NEXT:    sub sp, sp, #48
399; CHECK-NEXT:    str q0, [sp, #16] // 16-byte Folded Spill
400; CHECK-NEXT:    mov s0, v0.s[1]
401; CHECK-NEXT:    str x30, [sp, #32] // 8-byte Folded Spill
402; CHECK-NEXT:    bl log10f
403; CHECK-NEXT:    str d0, [sp] // 16-byte Folded Spill
404; CHECK-NEXT:    ldr q0, [sp, #16] // 16-byte Folded Reload
405; CHECK-NEXT:    // kill: def $s0 killed $s0 killed $q0
406; CHECK-NEXT:    bl log10f
407; CHECK-NEXT:    ldr q1, [sp] // 16-byte Folded Reload
408; CHECK-NEXT:    // kill: def $s0 killed $s0 def $q0
409; CHECK-NEXT:    mov v0.s[1], v1.s[0]
410; CHECK-NEXT:    str q0, [sp] // 16-byte Folded Spill
411; CHECK-NEXT:    ldr q0, [sp, #16] // 16-byte Folded Reload
412; CHECK-NEXT:    mov s0, v0.s[2]
413; CHECK-NEXT:    bl log10f
414; CHECK-NEXT:    ldr q1, [sp] // 16-byte Folded Reload
415; CHECK-NEXT:    // kill: def $s0 killed $s0 def $q0
416; CHECK-NEXT:    ldr x30, [sp, #32] // 8-byte Folded Reload
417; CHECK-NEXT:    mov v1.s[2], v0.s[0]
418; CHECK-NEXT:    mov v0.16b, v1.16b
419; CHECK-NEXT:    add sp, sp, #48
420; CHECK-NEXT:    ret
421  %r = call <3 x float> @llvm.log10.v3f32(<3 x float> %x)
422  ret <3 x float> %r
423}
424
425define <3 x float> @log2_v3f32(<3 x float> %x) nounwind {
426; CHECK-LABEL: log2_v3f32:
427; CHECK:       // %bb.0:
428; CHECK-NEXT:    sub sp, sp, #48
429; CHECK-NEXT:    str q0, [sp, #16] // 16-byte Folded Spill
430; CHECK-NEXT:    mov s0, v0.s[1]
431; CHECK-NEXT:    str x30, [sp, #32] // 8-byte Folded Spill
432; CHECK-NEXT:    bl log2f
433; CHECK-NEXT:    str d0, [sp] // 16-byte Folded Spill
434; CHECK-NEXT:    ldr q0, [sp, #16] // 16-byte Folded Reload
435; CHECK-NEXT:    // kill: def $s0 killed $s0 killed $q0
436; CHECK-NEXT:    bl log2f
437; CHECK-NEXT:    ldr q1, [sp] // 16-byte Folded Reload
438; CHECK-NEXT:    // kill: def $s0 killed $s0 def $q0
439; CHECK-NEXT:    mov v0.s[1], v1.s[0]
440; CHECK-NEXT:    str q0, [sp] // 16-byte Folded Spill
441; CHECK-NEXT:    ldr q0, [sp, #16] // 16-byte Folded Reload
442; CHECK-NEXT:    mov s0, v0.s[2]
443; CHECK-NEXT:    bl log2f
444; CHECK-NEXT:    ldr q1, [sp] // 16-byte Folded Reload
445; CHECK-NEXT:    // kill: def $s0 killed $s0 def $q0
446; CHECK-NEXT:    ldr x30, [sp, #32] // 8-byte Folded Reload
447; CHECK-NEXT:    mov v1.s[2], v0.s[0]
448; CHECK-NEXT:    mov v0.16b, v1.16b
449; CHECK-NEXT:    add sp, sp, #48
450; CHECK-NEXT:    ret
451  %r = call <3 x float> @llvm.log2.v3f32(<3 x float> %x)
452  ret <3 x float> %r
453}
454
455define <3 x float> @nearbyint__v3f32(<3 x float> %x) nounwind {
456; CHECK-LABEL: nearbyint__v3f32:
457; CHECK:       // %bb.0:
458; CHECK-NEXT:    frinti v0.4s, v0.4s
459; CHECK-NEXT:    ret
460  %r = call <3 x float> @llvm.nearbyint.v3f32(<3 x float> %x)
461  ret <3 x float> %r
462}
463
464define <3 x float> @rint_v3f32(<3 x float> %x) nounwind {
465; CHECK-LABEL: rint_v3f32:
466; CHECK:       // %bb.0:
467; CHECK-NEXT:    frintx v0.4s, v0.4s
468; CHECK-NEXT:    ret
469  %r = call <3 x float> @llvm.rint.v3f32(<3 x float> %x)
470  ret <3 x float> %r
471}
472
473define <3 x float> @round_v3f32(<3 x float> %x) nounwind {
474; CHECK-LABEL: round_v3f32:
475; CHECK:       // %bb.0:
476; CHECK-NEXT:    frinta v0.4s, v0.4s
477; CHECK-NEXT:    ret
478  %r = call <3 x float> @llvm.round.v3f32(<3 x float> %x)
479  ret <3 x float> %r
480}
481
482define <3 x float> @roundeven_v3f32(<3 x float> %x) nounwind {
483; CHECK-LABEL: roundeven_v3f32:
484; CHECK:       // %bb.0:
485; CHECK-NEXT:    frintn v0.4s, v0.4s
486; CHECK-NEXT:    ret
487  %r = call <3 x float> @llvm.roundeven.v3f32(<3 x float> %x)
488  ret <3 x float> %r
489}
490
491define <3 x float> @sqrt_v3f32(<3 x float> %x) nounwind {
492; CHECK-LABEL: sqrt_v3f32:
493; CHECK:       // %bb.0:
494; CHECK-NEXT:    fsqrt v0.4s, v0.4s
495; CHECK-NEXT:    ret
496  %r = call <3 x float> @llvm.sqrt.v3f32(<3 x float> %x)
497  ret <3 x float> %r
498}
499
500define <3 x float> @trunc_v3f32(<3 x float> %x) nounwind {
501; CHECK-LABEL: trunc_v3f32:
502; CHECK:       // %bb.0:
503; CHECK-NEXT:    frintz v0.4s, v0.4s
504; CHECK-NEXT:    ret
505  %r = call <3 x float> @llvm.trunc.v3f32(<3 x float> %x)
506  ret <3 x float> %r
507}
508
509