xref: /llvm-project/llvm/test/CodeGen/AArch64/vec-libcalls.ll (revision 61510b51c33464a6bc15e4cf5b1ee07e2e0ec1c9) 
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.tan.v3f32(<3 x float>)
24declare <3 x float> @llvm.asin.v3f32(<3 x float>)
25declare <3 x float> @llvm.acos.v3f32(<3 x float>)
26declare <3 x float> @llvm.atan.v3f32(<3 x float>)
27declare <3 x float> @llvm.atan2.v3f32(<3 x float>, <3 x float>)
28declare <3 x float> @llvm.sinh.v3f32(<3 x float>)
29declare <3 x float> @llvm.cosh.v3f32(<3 x float>)
30declare <3 x float> @llvm.tanh.v3f32(<3 x float>)
31declare <3 x float> @llvm.exp.v3f32(<3 x float>)
32declare <3 x float> @llvm.exp2.v3f32(<3 x float>)
33declare <3 x float> @llvm.floor.v3f32(<3 x float>)
34declare <3 x float> @llvm.log.v3f32(<3 x float>)
35declare <3 x float> @llvm.log10.v3f32(<3 x float>)
36declare <3 x float> @llvm.log2.v3f32(<3 x float>)
37declare <3 x float> @llvm.nearbyint.v3f32(<3 x float>)
38declare <3 x float> @llvm.rint.v3f32(<3 x float>)
39declare <3 x float> @llvm.round.v3f32(<3 x float>)
40declare <3 x float> @llvm.roundeven.v3f32(<3 x float>)
41declare <3 x float> @llvm.sqrt.v3f32(<3 x float>)
42declare <3 x float> @llvm.trunc.v3f32(<3 x float>)
43
44define <1 x float> @sin_v1f32(<1 x float> %x) nounwind {
45; CHECK-LABEL: sin_v1f32:
46; CHECK:       // %bb.0:
47; CHECK-NEXT:    str x30, [sp, #-16]! // 8-byte Folded Spill
48; CHECK-NEXT:    // kill: def $d0 killed $d0 def $q0
49; CHECK-NEXT:    // kill: def $s0 killed $s0 killed $q0
50; CHECK-NEXT:    bl sinf
51; CHECK-NEXT:    // kill: def $s0 killed $s0 def $d0
52; CHECK-NEXT:    ldr x30, [sp], #16 // 8-byte Folded Reload
53; CHECK-NEXT:    ret
54  %r = call <1 x float> @llvm.sin.v1f32(<1 x float> %x)
55  ret <1 x float> %r
56}
57
58define <2 x float> @sin_v2f32(<2 x float> %x) nounwind {
59; CHECK-LABEL: sin_v2f32:
60; CHECK:       // %bb.0:
61; CHECK-NEXT:    sub sp, sp, #48
62; CHECK-NEXT:    // kill: def $d0 killed $d0 def $q0
63; CHECK-NEXT:    str q0, [sp] // 16-byte Folded Spill
64; CHECK-NEXT:    mov s0, v0.s[1]
65; CHECK-NEXT:    str x30, [sp, #32] // 8-byte Folded Spill
66; CHECK-NEXT:    bl sinf
67; CHECK-NEXT:    // kill: def $s0 killed $s0 def $q0
68; CHECK-NEXT:    str q0, [sp, #16] // 16-byte Folded Spill
69; CHECK-NEXT:    ldr q0, [sp] // 16-byte Folded Reload
70; CHECK-NEXT:    // kill: def $s0 killed $s0 killed $q0
71; CHECK-NEXT:    bl sinf
72; CHECK-NEXT:    ldr q1, [sp, #16] // 16-byte Folded Reload
73; CHECK-NEXT:    // kill: def $s0 killed $s0 def $q0
74; CHECK-NEXT:    ldr x30, [sp, #32] // 8-byte Folded Reload
75; CHECK-NEXT:    mov v0.s[1], v1.s[0]
76; CHECK-NEXT:    // kill: def $d0 killed $d0 killed $q0
77; CHECK-NEXT:    add sp, sp, #48
78; CHECK-NEXT:    ret
79  %r = call <2 x float> @llvm.sin.v2f32(<2 x float> %x)
80  ret <2 x float> %r
81}
82
83define <3 x float> @sin_v3f32(<3 x float> %x) nounwind {
84; CHECK-LABEL: sin_v3f32:
85; CHECK:       // %bb.0:
86; CHECK-NEXT:    sub sp, sp, #48
87; CHECK-NEXT:    str q0, [sp, #16] // 16-byte Folded Spill
88; CHECK-NEXT:    mov s0, v0.s[1]
89; CHECK-NEXT:    str x30, [sp, #32] // 8-byte Folded Spill
90; CHECK-NEXT:    bl sinf
91; CHECK-NEXT:    // kill: def $s0 killed $s0 def $q0
92; CHECK-NEXT:    str q0, [sp] // 16-byte Folded Spill
93; CHECK-NEXT:    ldr q0, [sp, #16] // 16-byte Folded Reload
94; CHECK-NEXT:    // kill: def $s0 killed $s0 killed $q0
95; CHECK-NEXT:    bl sinf
96; CHECK-NEXT:    ldr q1, [sp] // 16-byte Folded Reload
97; CHECK-NEXT:    // kill: def $s0 killed $s0 def $q0
98; CHECK-NEXT:    mov v0.s[1], v1.s[0]
99; CHECK-NEXT:    str q0, [sp] // 16-byte Folded Spill
100; CHECK-NEXT:    ldr q0, [sp, #16] // 16-byte Folded Reload
101; CHECK-NEXT:    mov s0, v0.s[2]
102; CHECK-NEXT:    bl sinf
103; CHECK-NEXT:    ldr q1, [sp] // 16-byte Folded Reload
104; CHECK-NEXT:    // kill: def $s0 killed $s0 def $q0
105; CHECK-NEXT:    ldr x30, [sp, #32] // 8-byte Folded Reload
106; CHECK-NEXT:    mov v1.s[2], v0.s[0]
107; CHECK-NEXT:    mov v0.16b, v1.16b
108; CHECK-NEXT:    add sp, sp, #48
109; CHECK-NEXT:    ret
110  %r = call <3 x float> @llvm.sin.v3f32(<3 x float> %x)
111  ret <3 x float> %r
112}
113
114define <4 x float> @sin_v4f32(<4 x float> %x) nounwind {
115; CHECK-LABEL: sin_v4f32:
116; CHECK:       // %bb.0:
117; CHECK-NEXT:    sub sp, sp, #48
118; CHECK-NEXT:    str q0, [sp, #16] // 16-byte Folded Spill
119; CHECK-NEXT:    mov s0, v0.s[1]
120; CHECK-NEXT:    str x30, [sp, #32] // 8-byte Folded Spill
121; CHECK-NEXT:    bl sinf
122; CHECK-NEXT:    // kill: def $s0 killed $s0 def $q0
123; CHECK-NEXT:    str q0, [sp] // 16-byte Folded Spill
124; CHECK-NEXT:    ldr q0, [sp, #16] // 16-byte Folded Reload
125; CHECK-NEXT:    // kill: def $s0 killed $s0 killed $q0
126; CHECK-NEXT:    bl sinf
127; CHECK-NEXT:    ldr q1, [sp] // 16-byte Folded Reload
128; CHECK-NEXT:    // kill: def $s0 killed $s0 def $q0
129; CHECK-NEXT:    mov v0.s[1], v1.s[0]
130; CHECK-NEXT:    str q0, [sp] // 16-byte Folded Spill
131; CHECK-NEXT:    ldr q0, [sp, #16] // 16-byte Folded Reload
132; CHECK-NEXT:    mov s0, v0.s[2]
133; CHECK-NEXT:    bl sinf
134; CHECK-NEXT:    ldr q1, [sp] // 16-byte Folded Reload
135; CHECK-NEXT:    // kill: def $s0 killed $s0 def $q0
136; CHECK-NEXT:    mov v1.s[2], v0.s[0]
137; CHECK-NEXT:    ldr q0, [sp, #16] // 16-byte Folded Reload
138; CHECK-NEXT:    mov s0, v0.s[3]
139; CHECK-NEXT:    str q1, [sp] // 16-byte Folded Spill
140; CHECK-NEXT:    bl sinf
141; CHECK-NEXT:    ldr q1, [sp] // 16-byte Folded Reload
142; CHECK-NEXT:    // kill: def $s0 killed $s0 def $q0
143; CHECK-NEXT:    ldr x30, [sp, #32] // 8-byte Folded Reload
144; CHECK-NEXT:    mov v1.s[3], v0.s[0]
145; CHECK-NEXT:    mov v0.16b, v1.16b
146; CHECK-NEXT:    add sp, sp, #48
147; CHECK-NEXT:    ret
148  %r = call <4 x float> @llvm.sin.v4f32(<4 x float> %x)
149  ret <4 x float> %r
150}
151
152define <5 x float> @sin_v5f32(<5 x float> %x) nounwind {
153; CHECK-LABEL: sin_v5f32:
154; CHECK:       // %bb.0:
155; CHECK-NEXT:    str d12, [sp, #-48]! // 8-byte Folded Spill
156; CHECK-NEXT:    stp d11, d10, [sp, #8] // 16-byte Folded Spill
157; CHECK-NEXT:    fmov s10, s2
158; CHECK-NEXT:    fmov s11, s1
159; CHECK-NEXT:    stp d9, d8, [sp, #24] // 16-byte Folded Spill
160; CHECK-NEXT:    fmov s8, s4
161; CHECK-NEXT:    fmov s9, s3
162; CHECK-NEXT:    str x30, [sp, #40] // 8-byte Folded Spill
163; CHECK-NEXT:    bl sinf
164; CHECK-NEXT:    fmov s12, s0
165; CHECK-NEXT:    fmov s0, s11
166; CHECK-NEXT:    bl sinf
167; CHECK-NEXT:    fmov s11, s0
168; CHECK-NEXT:    fmov s0, s10
169; CHECK-NEXT:    bl sinf
170; CHECK-NEXT:    fmov s10, s0
171; CHECK-NEXT:    fmov s0, s9
172; CHECK-NEXT:    bl sinf
173; CHECK-NEXT:    fmov s9, s0
174; CHECK-NEXT:    fmov s0, s8
175; CHECK-NEXT:    bl sinf
176; CHECK-NEXT:    fmov s1, s11
177; CHECK-NEXT:    fmov s2, s10
178; CHECK-NEXT:    ldr x30, [sp, #40] // 8-byte Folded Reload
179; CHECK-NEXT:    fmov s3, s9
180; CHECK-NEXT:    ldp d9, d8, [sp, #24] // 16-byte Folded Reload
181; CHECK-NEXT:    ldp d11, d10, [sp, #8] // 16-byte Folded Reload
182; CHECK-NEXT:    fmov s4, s0
183; CHECK-NEXT:    fmov s0, s12
184; CHECK-NEXT:    ldr d12, [sp], #48 // 8-byte Folded Reload
185; CHECK-NEXT:    ret
186  %r = call <5 x float> @llvm.sin.v5f32(<5 x float> %x)
187  ret <5 x float> %r
188}
189
190define <6 x float> @sin_v6f32(<6 x float> %x) nounwind {
191; CHECK-LABEL: sin_v6f32:
192; CHECK:       // %bb.0:
193; CHECK-NEXT:    stp d13, d12, [sp, #-64]! // 16-byte Folded Spill
194; CHECK-NEXT:    stp d11, d10, [sp, #16] // 16-byte Folded Spill
195; CHECK-NEXT:    fmov s10, s3
196; CHECK-NEXT:    fmov s11, s2
197; CHECK-NEXT:    stp d9, d8, [sp, #32] // 16-byte Folded Spill
198; CHECK-NEXT:    fmov s8, s5
199; CHECK-NEXT:    fmov s9, s4
200; CHECK-NEXT:    str x30, [sp, #48] // 8-byte Folded Spill
201; CHECK-NEXT:    fmov s12, s1
202; CHECK-NEXT:    bl sinf
203; CHECK-NEXT:    fmov s13, s0
204; CHECK-NEXT:    fmov s0, s12
205; CHECK-NEXT:    bl sinf
206; CHECK-NEXT:    fmov s12, s0
207; CHECK-NEXT:    fmov s0, s11
208; CHECK-NEXT:    bl sinf
209; CHECK-NEXT:    fmov s11, s0
210; CHECK-NEXT:    fmov s0, s10
211; CHECK-NEXT:    bl sinf
212; CHECK-NEXT:    fmov s10, s0
213; CHECK-NEXT:    fmov s0, s9
214; CHECK-NEXT:    bl sinf
215; CHECK-NEXT:    fmov s9, s0
216; CHECK-NEXT:    fmov s0, s8
217; CHECK-NEXT:    bl sinf
218; CHECK-NEXT:    fmov s2, s11
219; CHECK-NEXT:    fmov s3, s10
220; CHECK-NEXT:    ldr x30, [sp, #48] // 8-byte Folded Reload
221; CHECK-NEXT:    fmov s4, s9
222; CHECK-NEXT:    ldp d9, d8, [sp, #32] // 16-byte Folded Reload
223; CHECK-NEXT:    ldp d11, d10, [sp, #16] // 16-byte Folded Reload
224; CHECK-NEXT:    fmov s5, s0
225; CHECK-NEXT:    fmov s0, s13
226; CHECK-NEXT:    fmov s1, s12
227; CHECK-NEXT:    ldp d13, d12, [sp], #64 // 16-byte Folded Reload
228; CHECK-NEXT:    ret
229  %r = call <6 x float> @llvm.sin.v6f32(<6 x float> %x)
230  ret <6 x float> %r
231}
232
233define <3 x double> @sin_v3f64(<3 x double> %x) nounwind {
234; CHECK-LABEL: sin_v3f64:
235; CHECK:       // %bb.0:
236; CHECK-NEXT:    str d10, [sp, #-32]! // 8-byte Folded Spill
237; CHECK-NEXT:    stp d9, d8, [sp, #8] // 16-byte Folded Spill
238; CHECK-NEXT:    fmov d8, d2
239; CHECK-NEXT:    fmov d9, d1
240; CHECK-NEXT:    str x30, [sp, #24] // 8-byte Folded Spill
241; CHECK-NEXT:    bl sin
242; CHECK-NEXT:    fmov d10, d0
243; CHECK-NEXT:    fmov d0, d9
244; CHECK-NEXT:    bl sin
245; CHECK-NEXT:    fmov d9, d0
246; CHECK-NEXT:    fmov d0, d8
247; CHECK-NEXT:    bl sin
248; CHECK-NEXT:    fmov d1, d9
249; CHECK-NEXT:    ldp d9, d8, [sp, #8] // 16-byte Folded Reload
250; CHECK-NEXT:    ldr x30, [sp, #24] // 8-byte Folded Reload
251; CHECK-NEXT:    fmov d2, d0
252; CHECK-NEXT:    fmov d0, d10
253; CHECK-NEXT:    ldr d10, [sp], #32 // 8-byte Folded Reload
254; CHECK-NEXT:    ret
255  %r = call <3 x double> @llvm.sin.v3f64(<3 x double> %x)
256  ret <3 x double> %r
257}
258
259define <3 x float> @fabs_v3f32(<3 x float> %x) nounwind {
260; CHECK-LABEL: fabs_v3f32:
261; CHECK:       // %bb.0:
262; CHECK-NEXT:    fabs v0.4s, v0.4s
263; CHECK-NEXT:    ret
264  %r = call <3 x float> @llvm.fabs.v3f32(<3 x float> %x)
265  ret <3 x float> %r
266}
267
268define <3 x float> @ceil_v3f32(<3 x float> %x) nounwind {
269; CHECK-LABEL: ceil_v3f32:
270; CHECK:       // %bb.0:
271; CHECK-NEXT:    frintp v0.4s, v0.4s
272; CHECK-NEXT:    ret
273  %r = call <3 x float> @llvm.ceil.v3f32(<3 x float> %x)
274  ret <3 x float> %r
275}
276
277define <3 x float> @cos_v3f32(<3 x float> %x) nounwind {
278; CHECK-LABEL: cos_v3f32:
279; CHECK:       // %bb.0:
280; CHECK-NEXT:    sub sp, sp, #48
281; CHECK-NEXT:    str q0, [sp, #16] // 16-byte Folded Spill
282; CHECK-NEXT:    mov s0, v0.s[1]
283; CHECK-NEXT:    str x30, [sp, #32] // 8-byte Folded Spill
284; CHECK-NEXT:    bl cosf
285; CHECK-NEXT:    // kill: def $s0 killed $s0 def $q0
286; CHECK-NEXT:    str q0, [sp] // 16-byte Folded Spill
287; CHECK-NEXT:    ldr q0, [sp, #16] // 16-byte Folded Reload
288; CHECK-NEXT:    // kill: def $s0 killed $s0 killed $q0
289; CHECK-NEXT:    bl cosf
290; CHECK-NEXT:    ldr q1, [sp] // 16-byte Folded Reload
291; CHECK-NEXT:    // kill: def $s0 killed $s0 def $q0
292; CHECK-NEXT:    mov v0.s[1], v1.s[0]
293; CHECK-NEXT:    str q0, [sp] // 16-byte Folded Spill
294; CHECK-NEXT:    ldr q0, [sp, #16] // 16-byte Folded Reload
295; CHECK-NEXT:    mov s0, v0.s[2]
296; CHECK-NEXT:    bl cosf
297; CHECK-NEXT:    ldr q1, [sp] // 16-byte Folded Reload
298; CHECK-NEXT:    // kill: def $s0 killed $s0 def $q0
299; CHECK-NEXT:    ldr x30, [sp, #32] // 8-byte Folded Reload
300; CHECK-NEXT:    mov v1.s[2], v0.s[0]
301; CHECK-NEXT:    mov v0.16b, v1.16b
302; CHECK-NEXT:    add sp, sp, #48
303; CHECK-NEXT:    ret
304  %r = call <3 x float> @llvm.cos.v3f32(<3 x float> %x)
305  ret <3 x float> %r
306}
307
308define <3 x float> @tan_v3f32(<3 x float> %x) nounwind {
309; CHECK-LABEL: tan_v3f32:
310; CHECK:       // %bb.0:
311; CHECK-NEXT:    sub sp, sp, #48
312; CHECK-NEXT:    str q0, [sp, #16] // 16-byte Folded Spill
313; CHECK-NEXT:    mov s0, v0.s[1]
314; CHECK-NEXT:    str x30, [sp, #32] // 8-byte Folded Spill
315; CHECK-NEXT:    bl tanf
316; CHECK-NEXT:    // kill: def $s0 killed $s0 def $q0
317; CHECK-NEXT:    str q0, [sp] // 16-byte Folded Spill
318; CHECK-NEXT:    ldr q0, [sp, #16] // 16-byte Folded Reload
319; CHECK-NEXT:    // kill: def $s0 killed $s0 killed $q0
320; CHECK-NEXT:    bl tanf
321; CHECK-NEXT:    ldr q1, [sp] // 16-byte Folded Reload
322; CHECK-NEXT:    // kill: def $s0 killed $s0 def $q0
323; CHECK-NEXT:    mov v0.s[1], v1.s[0]
324; CHECK-NEXT:    str q0, [sp] // 16-byte Folded Spill
325; CHECK-NEXT:    ldr q0, [sp, #16] // 16-byte Folded Reload
326; CHECK-NEXT:    mov s0, v0.s[2]
327; CHECK-NEXT:    bl tanf
328; CHECK-NEXT:    ldr q1, [sp] // 16-byte Folded Reload
329; CHECK-NEXT:    // kill: def $s0 killed $s0 def $q0
330; CHECK-NEXT:    ldr x30, [sp, #32] // 8-byte Folded Reload
331; CHECK-NEXT:    mov v1.s[2], v0.s[0]
332; CHECK-NEXT:    mov v0.16b, v1.16b
333; CHECK-NEXT:    add sp, sp, #48
334; CHECK-NEXT:    ret
335  %r = call <3 x float> @llvm.tan.v3f32(<3 x float> %x)
336  ret <3 x float> %r
337}
338
339define <3 x float> @asin_v3f32(<3 x float> %x) nounwind {
340; CHECK-LABEL: asin_v3f32:
341; CHECK:       // %bb.0:
342; CHECK-NEXT:    sub sp, sp, #48
343; CHECK-NEXT:    str q0, [sp, #16] // 16-byte Folded Spill
344; CHECK-NEXT:    mov s0, v0.s[1]
345; CHECK-NEXT:    str x30, [sp, #32] // 8-byte Folded Spill
346; CHECK-NEXT:    bl asinf
347; CHECK-NEXT:    // kill: def $s0 killed $s0 def $q0
348; CHECK-NEXT:    str q0, [sp] // 16-byte Folded Spill
349; CHECK-NEXT:    ldr q0, [sp, #16] // 16-byte Folded Reload
350; CHECK-NEXT:    // kill: def $s0 killed $s0 killed $q0
351; CHECK-NEXT:    bl asinf
352; CHECK-NEXT:    ldr q1, [sp] // 16-byte Folded Reload
353; CHECK-NEXT:    // kill: def $s0 killed $s0 def $q0
354; CHECK-NEXT:    mov v0.s[1], v1.s[0]
355; CHECK-NEXT:    str q0, [sp] // 16-byte Folded Spill
356; CHECK-NEXT:    ldr q0, [sp, #16] // 16-byte Folded Reload
357; CHECK-NEXT:    mov s0, v0.s[2]
358; CHECK-NEXT:    bl asinf
359; CHECK-NEXT:    ldr q1, [sp] // 16-byte Folded Reload
360; CHECK-NEXT:    // kill: def $s0 killed $s0 def $q0
361; CHECK-NEXT:    ldr x30, [sp, #32] // 8-byte Folded Reload
362; CHECK-NEXT:    mov v1.s[2], v0.s[0]
363; CHECK-NEXT:    mov v0.16b, v1.16b
364; CHECK-NEXT:    add sp, sp, #48
365; CHECK-NEXT:    ret
366  %r = call <3 x float> @llvm.asin.v3f32(<3 x float> %x)
367  ret <3 x float> %r
368}
369
370define <3 x float> @acos_v3f32(<3 x float> %x) nounwind {
371; CHECK-LABEL: acos_v3f32:
372; CHECK:       // %bb.0:
373; CHECK-NEXT:    sub sp, sp, #48
374; CHECK-NEXT:    str q0, [sp, #16] // 16-byte Folded Spill
375; CHECK-NEXT:    mov s0, v0.s[1]
376; CHECK-NEXT:    str x30, [sp, #32] // 8-byte Folded Spill
377; CHECK-NEXT:    bl acosf
378; CHECK-NEXT:    // kill: def $s0 killed $s0 def $q0
379; CHECK-NEXT:    str q0, [sp] // 16-byte Folded Spill
380; CHECK-NEXT:    ldr q0, [sp, #16] // 16-byte Folded Reload
381; CHECK-NEXT:    // kill: def $s0 killed $s0 killed $q0
382; CHECK-NEXT:    bl acosf
383; CHECK-NEXT:    ldr q1, [sp] // 16-byte Folded Reload
384; CHECK-NEXT:    // kill: def $s0 killed $s0 def $q0
385; CHECK-NEXT:    mov v0.s[1], v1.s[0]
386; CHECK-NEXT:    str q0, [sp] // 16-byte Folded Spill
387; CHECK-NEXT:    ldr q0, [sp, #16] // 16-byte Folded Reload
388; CHECK-NEXT:    mov s0, v0.s[2]
389; CHECK-NEXT:    bl acosf
390; CHECK-NEXT:    ldr q1, [sp] // 16-byte Folded Reload
391; CHECK-NEXT:    // kill: def $s0 killed $s0 def $q0
392; CHECK-NEXT:    ldr x30, [sp, #32] // 8-byte Folded Reload
393; CHECK-NEXT:    mov v1.s[2], v0.s[0]
394; CHECK-NEXT:    mov v0.16b, v1.16b
395; CHECK-NEXT:    add sp, sp, #48
396; CHECK-NEXT:    ret
397  %r = call <3 x float> @llvm.acos.v3f32(<3 x float> %x)
398  ret <3 x float> %r
399}
400
401define <3 x float> @atan_v3f32(<3 x float> %x) nounwind {
402; CHECK-LABEL: atan_v3f32:
403; CHECK:       // %bb.0:
404; CHECK-NEXT:    sub sp, sp, #48
405; CHECK-NEXT:    str q0, [sp, #16] // 16-byte Folded Spill
406; CHECK-NEXT:    mov s0, v0.s[1]
407; CHECK-NEXT:    str x30, [sp, #32] // 8-byte Folded Spill
408; CHECK-NEXT:    bl atanf
409; CHECK-NEXT:    // kill: def $s0 killed $s0 def $q0
410; CHECK-NEXT:    str q0, [sp] // 16-byte Folded Spill
411; CHECK-NEXT:    ldr q0, [sp, #16] // 16-byte Folded Reload
412; CHECK-NEXT:    // kill: def $s0 killed $s0 killed $q0
413; CHECK-NEXT:    bl atanf
414; CHECK-NEXT:    ldr q1, [sp] // 16-byte Folded Reload
415; CHECK-NEXT:    // kill: def $s0 killed $s0 def $q0
416; CHECK-NEXT:    mov v0.s[1], v1.s[0]
417; CHECK-NEXT:    str q0, [sp] // 16-byte Folded Spill
418; CHECK-NEXT:    ldr q0, [sp, #16] // 16-byte Folded Reload
419; CHECK-NEXT:    mov s0, v0.s[2]
420; CHECK-NEXT:    bl atanf
421; CHECK-NEXT:    ldr q1, [sp] // 16-byte Folded Reload
422; CHECK-NEXT:    // kill: def $s0 killed $s0 def $q0
423; CHECK-NEXT:    ldr x30, [sp, #32] // 8-byte Folded Reload
424; CHECK-NEXT:    mov v1.s[2], v0.s[0]
425; CHECK-NEXT:    mov v0.16b, v1.16b
426; CHECK-NEXT:    add sp, sp, #48
427; CHECK-NEXT:    ret
428  %r = call <3 x float> @llvm.atan.v3f32(<3 x float> %x)
429  ret <3 x float> %r
430}
431
432define <3 x float> @atan2_v3f32(<3 x float> %x, <3 x float> %y) nounwind {
433; CHECK-LABEL: atan2_v3f32:
434; CHECK:       // %bb.0:
435; CHECK-NEXT:    sub sp, sp, #64
436; CHECK-NEXT:    stp q0, q1, [sp, #16] // 32-byte Folded Spill
437; CHECK-NEXT:    mov s0, v0.s[1]
438; CHECK-NEXT:    mov s1, v1.s[1]
439; CHECK-NEXT:    str x30, [sp, #48] // 8-byte Folded Spill
440; CHECK-NEXT:    bl atan2f
441; CHECK-NEXT:    // kill: def $s0 killed $s0 def $q0
442; CHECK-NEXT:    str q0, [sp] // 16-byte Folded Spill
443; CHECK-NEXT:    ldp q0, q1, [sp, #16] // 32-byte Folded Reload
444; CHECK-NEXT:    // kill: def $s0 killed $s0 killed $q0
445; CHECK-NEXT:    // kill: def $s1 killed $s1 killed $q1
446; CHECK-NEXT:    bl atan2f
447; CHECK-NEXT:    ldr q1, [sp] // 16-byte Folded Reload
448; CHECK-NEXT:    // kill: def $s0 killed $s0 def $q0
449; CHECK-NEXT:    mov v0.s[1], v1.s[0]
450; CHECK-NEXT:    str q0, [sp] // 16-byte Folded Spill
451; CHECK-NEXT:    ldp q0, q1, [sp, #16] // 32-byte Folded Reload
452; CHECK-NEXT:    mov s0, v0.s[2]
453; CHECK-NEXT:    mov s1, v1.s[2]
454; CHECK-NEXT:    bl atan2f
455; CHECK-NEXT:    ldr q1, [sp] // 16-byte Folded Reload
456; CHECK-NEXT:    // kill: def $s0 killed $s0 def $q0
457; CHECK-NEXT:    ldr x30, [sp, #48] // 8-byte Folded Reload
458; CHECK-NEXT:    mov v1.s[2], v0.s[0]
459; CHECK-NEXT:    mov v0.16b, v1.16b
460; CHECK-NEXT:    add sp, sp, #64
461; CHECK-NEXT:    ret
462  %r = call <3 x float> @llvm.atan2.v3f32(<3 x float> %x, <3 x float> %y)
463  ret <3 x float> %r
464}
465
466define <3 x float> @sinh_v3f32(<3 x float> %x) nounwind {
467; CHECK-LABEL: sinh_v3f32:
468; CHECK:       // %bb.0:
469; CHECK-NEXT:    sub sp, sp, #48
470; CHECK-NEXT:    str q0, [sp, #16] // 16-byte Folded Spill
471; CHECK-NEXT:    mov s0, v0.s[1]
472; CHECK-NEXT:    str x30, [sp, #32] // 8-byte Folded Spill
473; CHECK-NEXT:    bl sinhf
474; CHECK-NEXT:    // kill: def $s0 killed $s0 def $q0
475; CHECK-NEXT:    str q0, [sp] // 16-byte Folded Spill
476; CHECK-NEXT:    ldr q0, [sp, #16] // 16-byte Folded Reload
477; CHECK-NEXT:    // kill: def $s0 killed $s0 killed $q0
478; CHECK-NEXT:    bl sinhf
479; CHECK-NEXT:    ldr q1, [sp] // 16-byte Folded Reload
480; CHECK-NEXT:    // kill: def $s0 killed $s0 def $q0
481; CHECK-NEXT:    mov v0.s[1], v1.s[0]
482; CHECK-NEXT:    str q0, [sp] // 16-byte Folded Spill
483; CHECK-NEXT:    ldr q0, [sp, #16] // 16-byte Folded Reload
484; CHECK-NEXT:    mov s0, v0.s[2]
485; CHECK-NEXT:    bl sinhf
486; CHECK-NEXT:    ldr q1, [sp] // 16-byte Folded Reload
487; CHECK-NEXT:    // kill: def $s0 killed $s0 def $q0
488; CHECK-NEXT:    ldr x30, [sp, #32] // 8-byte Folded Reload
489; CHECK-NEXT:    mov v1.s[2], v0.s[0]
490; CHECK-NEXT:    mov v0.16b, v1.16b
491; CHECK-NEXT:    add sp, sp, #48
492; CHECK-NEXT:    ret
493  %r = call <3 x float> @llvm.sinh.v3f32(<3 x float> %x)
494  ret <3 x float> %r
495}
496define <3 x float> @cosh_v3f32(<3 x float> %x) nounwind {
497; CHECK-LABEL: cosh_v3f32:
498; CHECK:       // %bb.0:
499; CHECK-NEXT:    sub sp, sp, #48
500; CHECK-NEXT:    str q0, [sp, #16] // 16-byte Folded Spill
501; CHECK-NEXT:    mov s0, v0.s[1]
502; CHECK-NEXT:    str x30, [sp, #32] // 8-byte Folded Spill
503; CHECK-NEXT:    bl coshf
504; CHECK-NEXT:    // kill: def $s0 killed $s0 def $q0
505; CHECK-NEXT:    str q0, [sp] // 16-byte Folded Spill
506; CHECK-NEXT:    ldr q0, [sp, #16] // 16-byte Folded Reload
507; CHECK-NEXT:    // kill: def $s0 killed $s0 killed $q0
508; CHECK-NEXT:    bl coshf
509; CHECK-NEXT:    ldr q1, [sp] // 16-byte Folded Reload
510; CHECK-NEXT:    // kill: def $s0 killed $s0 def $q0
511; CHECK-NEXT:    mov v0.s[1], v1.s[0]
512; CHECK-NEXT:    str q0, [sp] // 16-byte Folded Spill
513; CHECK-NEXT:    ldr q0, [sp, #16] // 16-byte Folded Reload
514; CHECK-NEXT:    mov s0, v0.s[2]
515; CHECK-NEXT:    bl coshf
516; CHECK-NEXT:    ldr q1, [sp] // 16-byte Folded Reload
517; CHECK-NEXT:    // kill: def $s0 killed $s0 def $q0
518; CHECK-NEXT:    ldr x30, [sp, #32] // 8-byte Folded Reload
519; CHECK-NEXT:    mov v1.s[2], v0.s[0]
520; CHECK-NEXT:    mov v0.16b, v1.16b
521; CHECK-NEXT:    add sp, sp, #48
522; CHECK-NEXT:    ret
523  %r = call <3 x float> @llvm.cosh.v3f32(<3 x float> %x)
524  ret <3 x float> %r
525}
526
527define <3 x float> @tanh_v3f32(<3 x float> %x) nounwind {
528; CHECK-LABEL: tanh_v3f32:
529; CHECK:       // %bb.0:
530; CHECK-NEXT:    sub sp, sp, #48
531; CHECK-NEXT:    str q0, [sp, #16] // 16-byte Folded Spill
532; CHECK-NEXT:    mov s0, v0.s[1]
533; CHECK-NEXT:    str x30, [sp, #32] // 8-byte Folded Spill
534; CHECK-NEXT:    bl tanhf
535; CHECK-NEXT:    // kill: def $s0 killed $s0 def $q0
536; CHECK-NEXT:    str q0, [sp] // 16-byte Folded Spill
537; CHECK-NEXT:    ldr q0, [sp, #16] // 16-byte Folded Reload
538; CHECK-NEXT:    // kill: def $s0 killed $s0 killed $q0
539; CHECK-NEXT:    bl tanhf
540; CHECK-NEXT:    ldr q1, [sp] // 16-byte Folded Reload
541; CHECK-NEXT:    // kill: def $s0 killed $s0 def $q0
542; CHECK-NEXT:    mov v0.s[1], v1.s[0]
543; CHECK-NEXT:    str q0, [sp] // 16-byte Folded Spill
544; CHECK-NEXT:    ldr q0, [sp, #16] // 16-byte Folded Reload
545; CHECK-NEXT:    mov s0, v0.s[2]
546; CHECK-NEXT:    bl tanhf
547; CHECK-NEXT:    ldr q1, [sp] // 16-byte Folded Reload
548; CHECK-NEXT:    // kill: def $s0 killed $s0 def $q0
549; CHECK-NEXT:    ldr x30, [sp, #32] // 8-byte Folded Reload
550; CHECK-NEXT:    mov v1.s[2], v0.s[0]
551; CHECK-NEXT:    mov v0.16b, v1.16b
552; CHECK-NEXT:    add sp, sp, #48
553; CHECK-NEXT:    ret
554  %r = call <3 x float> @llvm.tanh.v3f32(<3 x float> %x)
555  ret <3 x float> %r
556}
557
558define <3 x float> @exp_v3f32(<3 x float> %x) nounwind {
559; CHECK-LABEL: exp_v3f32:
560; CHECK:       // %bb.0:
561; CHECK-NEXT:    sub sp, sp, #48
562; CHECK-NEXT:    str q0, [sp, #16] // 16-byte Folded Spill
563; CHECK-NEXT:    mov s0, v0.s[1]
564; CHECK-NEXT:    str x30, [sp, #32] // 8-byte Folded Spill
565; CHECK-NEXT:    bl expf
566; CHECK-NEXT:    // kill: def $s0 killed $s0 def $q0
567; CHECK-NEXT:    str q0, [sp] // 16-byte Folded Spill
568; CHECK-NEXT:    ldr q0, [sp, #16] // 16-byte Folded Reload
569; CHECK-NEXT:    // kill: def $s0 killed $s0 killed $q0
570; CHECK-NEXT:    bl expf
571; CHECK-NEXT:    ldr q1, [sp] // 16-byte Folded Reload
572; CHECK-NEXT:    // kill: def $s0 killed $s0 def $q0
573; CHECK-NEXT:    mov v0.s[1], v1.s[0]
574; CHECK-NEXT:    str q0, [sp] // 16-byte Folded Spill
575; CHECK-NEXT:    ldr q0, [sp, #16] // 16-byte Folded Reload
576; CHECK-NEXT:    mov s0, v0.s[2]
577; CHECK-NEXT:    bl expf
578; CHECK-NEXT:    ldr q1, [sp] // 16-byte Folded Reload
579; CHECK-NEXT:    // kill: def $s0 killed $s0 def $q0
580; CHECK-NEXT:    ldr x30, [sp, #32] // 8-byte Folded Reload
581; CHECK-NEXT:    mov v1.s[2], v0.s[0]
582; CHECK-NEXT:    mov v0.16b, v1.16b
583; CHECK-NEXT:    add sp, sp, #48
584; CHECK-NEXT:    ret
585  %r = call <3 x float> @llvm.exp.v3f32(<3 x float> %x)
586  ret <3 x float> %r
587}
588
589define <3 x float> @exp2_v3f32(<3 x float> %x) nounwind {
590; CHECK-LABEL: exp2_v3f32:
591; CHECK:       // %bb.0:
592; CHECK-NEXT:    sub sp, sp, #48
593; CHECK-NEXT:    str q0, [sp, #16] // 16-byte Folded Spill
594; CHECK-NEXT:    mov s0, v0.s[1]
595; CHECK-NEXT:    str x30, [sp, #32] // 8-byte Folded Spill
596; CHECK-NEXT:    bl exp2f
597; CHECK-NEXT:    // kill: def $s0 killed $s0 def $q0
598; CHECK-NEXT:    str q0, [sp] // 16-byte Folded Spill
599; CHECK-NEXT:    ldr q0, [sp, #16] // 16-byte Folded Reload
600; CHECK-NEXT:    // kill: def $s0 killed $s0 killed $q0
601; CHECK-NEXT:    bl exp2f
602; CHECK-NEXT:    ldr q1, [sp] // 16-byte Folded Reload
603; CHECK-NEXT:    // kill: def $s0 killed $s0 def $q0
604; CHECK-NEXT:    mov v0.s[1], v1.s[0]
605; CHECK-NEXT:    str q0, [sp] // 16-byte Folded Spill
606; CHECK-NEXT:    ldr q0, [sp, #16] // 16-byte Folded Reload
607; CHECK-NEXT:    mov s0, v0.s[2]
608; CHECK-NEXT:    bl exp2f
609; CHECK-NEXT:    ldr q1, [sp] // 16-byte Folded Reload
610; CHECK-NEXT:    // kill: def $s0 killed $s0 def $q0
611; CHECK-NEXT:    ldr x30, [sp, #32] // 8-byte Folded Reload
612; CHECK-NEXT:    mov v1.s[2], v0.s[0]
613; CHECK-NEXT:    mov v0.16b, v1.16b
614; CHECK-NEXT:    add sp, sp, #48
615; CHECK-NEXT:    ret
616  %r = call <3 x float> @llvm.exp2.v3f32(<3 x float> %x)
617  ret <3 x float> %r
618}
619
620define <3 x float> @floor_v3f32(<3 x float> %x) nounwind {
621; CHECK-LABEL: floor_v3f32:
622; CHECK:       // %bb.0:
623; CHECK-NEXT:    frintm v0.4s, v0.4s
624; CHECK-NEXT:    ret
625  %r = call <3 x float> @llvm.floor.v3f32(<3 x float> %x)
626  ret <3 x float> %r
627}
628
629define <3 x float> @log_v3f32(<3 x float> %x) nounwind {
630; CHECK-LABEL: log_v3f32:
631; CHECK:       // %bb.0:
632; CHECK-NEXT:    sub sp, sp, #48
633; CHECK-NEXT:    str q0, [sp, #16] // 16-byte Folded Spill
634; CHECK-NEXT:    mov s0, v0.s[1]
635; CHECK-NEXT:    str x30, [sp, #32] // 8-byte Folded Spill
636; CHECK-NEXT:    bl logf
637; CHECK-NEXT:    // kill: def $s0 killed $s0 def $q0
638; CHECK-NEXT:    str q0, [sp] // 16-byte Folded Spill
639; CHECK-NEXT:    ldr q0, [sp, #16] // 16-byte Folded Reload
640; CHECK-NEXT:    // kill: def $s0 killed $s0 killed $q0
641; CHECK-NEXT:    bl logf
642; CHECK-NEXT:    ldr q1, [sp] // 16-byte Folded Reload
643; CHECK-NEXT:    // kill: def $s0 killed $s0 def $q0
644; CHECK-NEXT:    mov v0.s[1], v1.s[0]
645; CHECK-NEXT:    str q0, [sp] // 16-byte Folded Spill
646; CHECK-NEXT:    ldr q0, [sp, #16] // 16-byte Folded Reload
647; CHECK-NEXT:    mov s0, v0.s[2]
648; CHECK-NEXT:    bl logf
649; CHECK-NEXT:    ldr q1, [sp] // 16-byte Folded Reload
650; CHECK-NEXT:    // kill: def $s0 killed $s0 def $q0
651; CHECK-NEXT:    ldr x30, [sp, #32] // 8-byte Folded Reload
652; CHECK-NEXT:    mov v1.s[2], v0.s[0]
653; CHECK-NEXT:    mov v0.16b, v1.16b
654; CHECK-NEXT:    add sp, sp, #48
655; CHECK-NEXT:    ret
656  %r = call <3 x float> @llvm.log.v3f32(<3 x float> %x)
657  ret <3 x float> %r
658}
659
660define <3 x float> @log10_v3f32(<3 x float> %x) nounwind {
661; CHECK-LABEL: log10_v3f32:
662; CHECK:       // %bb.0:
663; CHECK-NEXT:    sub sp, sp, #48
664; CHECK-NEXT:    str q0, [sp, #16] // 16-byte Folded Spill
665; CHECK-NEXT:    mov s0, v0.s[1]
666; CHECK-NEXT:    str x30, [sp, #32] // 8-byte Folded Spill
667; CHECK-NEXT:    bl log10f
668; CHECK-NEXT:    // kill: def $s0 killed $s0 def $q0
669; CHECK-NEXT:    str q0, [sp] // 16-byte Folded Spill
670; CHECK-NEXT:    ldr q0, [sp, #16] // 16-byte Folded Reload
671; CHECK-NEXT:    // kill: def $s0 killed $s0 killed $q0
672; CHECK-NEXT:    bl log10f
673; CHECK-NEXT:    ldr q1, [sp] // 16-byte Folded Reload
674; CHECK-NEXT:    // kill: def $s0 killed $s0 def $q0
675; CHECK-NEXT:    mov v0.s[1], v1.s[0]
676; CHECK-NEXT:    str q0, [sp] // 16-byte Folded Spill
677; CHECK-NEXT:    ldr q0, [sp, #16] // 16-byte Folded Reload
678; CHECK-NEXT:    mov s0, v0.s[2]
679; CHECK-NEXT:    bl log10f
680; CHECK-NEXT:    ldr q1, [sp] // 16-byte Folded Reload
681; CHECK-NEXT:    // kill: def $s0 killed $s0 def $q0
682; CHECK-NEXT:    ldr x30, [sp, #32] // 8-byte Folded Reload
683; CHECK-NEXT:    mov v1.s[2], v0.s[0]
684; CHECK-NEXT:    mov v0.16b, v1.16b
685; CHECK-NEXT:    add sp, sp, #48
686; CHECK-NEXT:    ret
687  %r = call <3 x float> @llvm.log10.v3f32(<3 x float> %x)
688  ret <3 x float> %r
689}
690
691define <3 x float> @log2_v3f32(<3 x float> %x) nounwind {
692; CHECK-LABEL: log2_v3f32:
693; CHECK:       // %bb.0:
694; CHECK-NEXT:    sub sp, sp, #48
695; CHECK-NEXT:    str q0, [sp, #16] // 16-byte Folded Spill
696; CHECK-NEXT:    mov s0, v0.s[1]
697; CHECK-NEXT:    str x30, [sp, #32] // 8-byte Folded Spill
698; CHECK-NEXT:    bl log2f
699; CHECK-NEXT:    // kill: def $s0 killed $s0 def $q0
700; CHECK-NEXT:    str q0, [sp] // 16-byte Folded Spill
701; CHECK-NEXT:    ldr q0, [sp, #16] // 16-byte Folded Reload
702; CHECK-NEXT:    // kill: def $s0 killed $s0 killed $q0
703; CHECK-NEXT:    bl log2f
704; CHECK-NEXT:    ldr q1, [sp] // 16-byte Folded Reload
705; CHECK-NEXT:    // kill: def $s0 killed $s0 def $q0
706; CHECK-NEXT:    mov v0.s[1], v1.s[0]
707; CHECK-NEXT:    str q0, [sp] // 16-byte Folded Spill
708; CHECK-NEXT:    ldr q0, [sp, #16] // 16-byte Folded Reload
709; CHECK-NEXT:    mov s0, v0.s[2]
710; CHECK-NEXT:    bl log2f
711; CHECK-NEXT:    ldr q1, [sp] // 16-byte Folded Reload
712; CHECK-NEXT:    // kill: def $s0 killed $s0 def $q0
713; CHECK-NEXT:    ldr x30, [sp, #32] // 8-byte Folded Reload
714; CHECK-NEXT:    mov v1.s[2], v0.s[0]
715; CHECK-NEXT:    mov v0.16b, v1.16b
716; CHECK-NEXT:    add sp, sp, #48
717; CHECK-NEXT:    ret
718  %r = call <3 x float> @llvm.log2.v3f32(<3 x float> %x)
719  ret <3 x float> %r
720}
721
722define <3 x float> @nearbyint__v3f32(<3 x float> %x) nounwind {
723; CHECK-LABEL: nearbyint__v3f32:
724; CHECK:       // %bb.0:
725; CHECK-NEXT:    frinti v0.4s, v0.4s
726; CHECK-NEXT:    ret
727  %r = call <3 x float> @llvm.nearbyint.v3f32(<3 x float> %x)
728  ret <3 x float> %r
729}
730
731define <3 x float> @rint_v3f32(<3 x float> %x) nounwind {
732; CHECK-LABEL: rint_v3f32:
733; CHECK:       // %bb.0:
734; CHECK-NEXT:    frintx v0.4s, v0.4s
735; CHECK-NEXT:    ret
736  %r = call <3 x float> @llvm.rint.v3f32(<3 x float> %x)
737  ret <3 x float> %r
738}
739
740define <3 x float> @round_v3f32(<3 x float> %x) nounwind {
741; CHECK-LABEL: round_v3f32:
742; CHECK:       // %bb.0:
743; CHECK-NEXT:    frinta v0.4s, v0.4s
744; CHECK-NEXT:    ret
745  %r = call <3 x float> @llvm.round.v3f32(<3 x float> %x)
746  ret <3 x float> %r
747}
748
749define <3 x float> @roundeven_v3f32(<3 x float> %x) nounwind {
750; CHECK-LABEL: roundeven_v3f32:
751; CHECK:       // %bb.0:
752; CHECK-NEXT:    frintn v0.4s, v0.4s
753; CHECK-NEXT:    ret
754  %r = call <3 x float> @llvm.roundeven.v3f32(<3 x float> %x)
755  ret <3 x float> %r
756}
757
758define <3 x float> @sqrt_v3f32(<3 x float> %x) nounwind {
759; CHECK-LABEL: sqrt_v3f32:
760; CHECK:       // %bb.0:
761; CHECK-NEXT:    fsqrt v0.4s, v0.4s
762; CHECK-NEXT:    ret
763  %r = call <3 x float> @llvm.sqrt.v3f32(<3 x float> %x)
764  ret <3 x float> %r
765}
766
767define <3 x float> @trunc_v3f32(<3 x float> %x) nounwind {
768; CHECK-LABEL: trunc_v3f32:
769; CHECK:       // %bb.0:
770; CHECK-NEXT:    frintz v0.4s, v0.4s
771; CHECK-NEXT:    ret
772  %r = call <3 x float> @llvm.trunc.v3f32(<3 x float> %x)
773  ret <3 x float> %r
774}
775
776