• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1; RUN: llc -mtriple=aarch64-linux-gnu -mattr=+sve < %s 2>%t | FileCheck %s
2; RUN: FileCheck --check-prefix=WARN --allow-empty %s <%t
3
4; If this check fails please read test/CodeGen/AArch64/README for instructions on how to resolve it.
5; WARN-NOT: warning
6
7;
8; FABD
9;
10
11define <vscale x 8 x half> @fabd_h(<vscale x 8 x i1> %pg, <vscale x 8 x half> %a, <vscale x 8 x half> %b) {
12; CHECK-LABEL: fabd_h:
13; CHECK: fabd z0.h, p0/m, z0.h, z1.h
14; CHECK-NEXT: ret
15  %out = call <vscale x 8 x half> @llvm.aarch64.sve.fabd.nxv8f16(<vscale x 8 x i1> %pg,
16                                                                 <vscale x 8 x half> %a,
17                                                                 <vscale x 8 x half> %b)
18  ret <vscale x 8 x half> %out
19}
20
21define <vscale x 4 x float> @fabd_s(<vscale x 4 x i1> %pg, <vscale x 4 x float> %a, <vscale x 4 x float> %b) {
22; CHECK-LABEL: fabd_s:
23; CHECK: fabd z0.s, p0/m, z0.s, z1.s
24; CHECK-NEXT: ret
25  %out = call <vscale x 4 x float> @llvm.aarch64.sve.fabd.nxv4f32(<vscale x 4 x i1> %pg,
26                                                                  <vscale x 4 x float> %a,
27                                                                  <vscale x 4 x float> %b)
28  ret <vscale x 4 x float> %out
29}
30
31define <vscale x 2 x double> @fabd_d(<vscale x 2 x i1> %pg, <vscale x 2 x double> %a, <vscale x 2 x double> %b) {
32; CHECK-LABEL: fabd_d:
33; CHECK: fabd z0.d, p0/m, z0.d, z1.d
34; CHECK-NEXT: ret
35  %out = call <vscale x 2 x double> @llvm.aarch64.sve.fabd.nxv2f64(<vscale x 2 x i1> %pg,
36                                                                   <vscale x 2 x double> %a,
37                                                                   <vscale x 2 x double> %b)
38  ret <vscale x 2 x double> %out
39}
40
41;
42; FABS
43;
44
45define <vscale x 8 x half> @fabs_h(<vscale x 8 x half> %a, <vscale x 8 x i1> %pg, <vscale x 8 x half> %b) {
46; CHECK-LABEL: fabs_h:
47; CHECK: fabs z0.h, p0/m, z1.h
48; CHECK-NEXT: ret
49  %out = call <vscale x 8 x half> @llvm.aarch64.sve.fabs.nxv8f16(<vscale x 8 x half> %a,
50                                                                 <vscale x 8 x i1> %pg,
51                                                                 <vscale x 8 x half> %b)
52  ret <vscale x 8 x half> %out
53}
54
55define <vscale x 4 x float> @fabs_s(<vscale x 4 x float> %a, <vscale x 4 x i1> %pg, <vscale x 4 x float> %b) {
56; CHECK-LABEL: fabs_s:
57; CHECK: fabs z0.s, p0/m, z1.s
58; CHECK-NEXT: ret
59  %out = call <vscale x 4 x float> @llvm.aarch64.sve.fabs.nxv4f32(<vscale x 4 x float> %a,
60                                                                  <vscale x 4 x i1> %pg,
61                                                                  <vscale x 4 x float> %b)
62  ret <vscale x 4 x float> %out
63}
64
65define <vscale x 2 x double> @fabs_d(<vscale x 2 x double> %a, <vscale x 2 x i1> %pg, <vscale x 2 x double> %b) {
66; CHECK-LABEL: fabs_d:
67; CHECK: fabs z0.d, p0/m, z1.d
68; CHECK-NEXT: ret
69  %out = call <vscale x 2 x double> @llvm.aarch64.sve.fabs.nxv2f64(<vscale x 2 x double> %a,
70                                                                   <vscale x 2 x i1> %pg,
71                                                                   <vscale x 2 x double> %b)
72  ret <vscale x 2 x double> %out
73}
74
75;
76; FADD
77;
78
79define <vscale x 8 x half> @fadd_h(<vscale x 8 x i1> %pg, <vscale x 8 x half> %a, <vscale x 8 x half> %b) {
80; CHECK-LABEL: fadd_h:
81; CHECK: fadd z0.h, p0/m, z0.h, z1.h
82; CHECK-NEXT: ret
83  %out = call <vscale x 8 x half> @llvm.aarch64.sve.fadd.nxv8f16(<vscale x 8 x i1> %pg,
84                                                                 <vscale x 8 x half> %a,
85                                                                 <vscale x 8 x half> %b)
86  ret <vscale x 8 x half> %out
87}
88
89define <vscale x 4 x float> @fadd_s(<vscale x 4 x i1> %pg, <vscale x 4 x float> %a, <vscale x 4 x float> %b) {
90; CHECK-LABEL: fadd_s:
91; CHECK: fadd z0.s, p0/m, z0.s, z1.s
92; CHECK-NEXT: ret
93  %out = call <vscale x 4 x float> @llvm.aarch64.sve.fadd.nxv4f32(<vscale x 4 x i1> %pg,
94                                                                  <vscale x 4 x float> %a,
95                                                                  <vscale x 4 x float> %b)
96  ret <vscale x 4 x float> %out
97}
98
99define <vscale x 2 x double> @fadd_d(<vscale x 2 x i1> %pg, <vscale x 2 x double> %a, <vscale x 2 x double> %b) {
100; CHECK-LABEL: fadd_d:
101; CHECK: fadd z0.d, p0/m, z0.d, z1.d
102; CHECK-NEXT: ret
103  %out = call <vscale x 2 x double> @llvm.aarch64.sve.fadd.nxv2f64(<vscale x 2 x i1> %pg,
104                                                                   <vscale x 2 x double> %a,
105                                                                   <vscale x 2 x double> %b)
106  ret <vscale x 2 x double> %out
107}
108
109;
110; FCADD
111;
112
113define <vscale x 8 x half> @fcadd_h(<vscale x 8 x i1> %pg, <vscale x 8 x half> %a, <vscale x 8 x half> %b) {
114; CHECK-LABEL: fcadd_h:
115; CHECK: fcadd z0.h, p0/m, z0.h, z1.h, #90
116; CHECK-NEXT: ret
117  %out = call <vscale x 8 x half> @llvm.aarch64.sve.fcadd.nxv8f16(<vscale x 8 x i1> %pg,
118                                                                  <vscale x 8 x half> %a,
119                                                                  <vscale x 8 x half> %b,
120                                                                  i32 90)
121  ret <vscale x 8 x half> %out
122}
123
124define <vscale x 4 x float> @fcadd_s(<vscale x 4 x i1> %pg, <vscale x 4 x float> %a, <vscale x 4 x float> %b) {
125; CHECK-LABEL: fcadd_s:
126; CHECK: fcadd z0.s, p0/m, z0.s, z1.s, #270
127; CHECK-NEXT: ret
128  %out = call <vscale x 4 x float> @llvm.aarch64.sve.fcadd.nxv4f32(<vscale x 4 x i1> %pg,
129                                                                   <vscale x 4 x float> %a,
130                                                                   <vscale x 4 x float> %b,
131                                                                   i32 270)
132  ret <vscale x 4 x float> %out
133}
134
135define <vscale x 2 x double> @fcadd_d(<vscale x 2 x i1> %pg, <vscale x 2 x double> %a, <vscale x 2 x double> %b) {
136; CHECK-LABEL: fcadd_d:
137; CHECK: fcadd z0.d, p0/m, z0.d, z1.d, #90
138; CHECK-NEXT: ret
139  %out = call <vscale x 2 x double> @llvm.aarch64.sve.fcadd.nxv2f64(<vscale x 2 x i1> %pg,
140                                                                    <vscale x 2 x double> %a,
141                                                                    <vscale x 2 x double> %b,
142                                                                    i32 90)
143  ret <vscale x 2 x double> %out
144}
145
146;
147; FCMLA
148;
149
150define <vscale x 8 x half> @fcmla_h(<vscale x 8 x i1> %pg, <vscale x 8 x half> %a, <vscale x 8 x half> %b, <vscale x 8 x half> %c) {
151; CHECK-LABEL: fcmla_h:
152; CHECK: fcmla z0.h, p0/m, z1.h, z2.h, #90
153; CHECK-NEXT: ret
154  %out = call <vscale x 8 x half> @llvm.aarch64.sve.fcmla.nxv8f16(<vscale x 8 x i1> %pg,
155                                                                  <vscale x 8 x half> %a,
156                                                                  <vscale x 8 x half> %b,
157                                                                  <vscale x 8 x half> %c,
158                                                                  i32 90)
159  ret <vscale x 8 x half> %out
160}
161
162define <vscale x 4 x float> @fcmla_s(<vscale x 4 x i1> %pg, <vscale x 4 x float> %a, <vscale x 4 x float> %b, <vscale x 4 x float> %c) {
163; CHECK-LABEL: fcmla_s:
164; CHECK: fcmla z0.s, p0/m, z1.s, z2.s, #180
165; CHECK-NEXT: ret
166  %out = call <vscale x 4 x float> @llvm.aarch64.sve.fcmla.nxv4f32(<vscale x 4 x i1> %pg,
167                                                                   <vscale x 4 x float> %a,
168                                                                   <vscale x 4 x float> %b,
169                                                                   <vscale x 4 x float> %c,
170                                                                   i32 180)
171  ret <vscale x 4 x float> %out
172}
173
174define <vscale x 2 x double> @fcmla_d(<vscale x 2 x i1> %pg, <vscale x 2 x double> %a, <vscale x 2 x double> %b, <vscale x 2 x double> %c) {
175; CHECK-LABEL: fcmla_d:
176; CHECK: fcmla z0.d, p0/m, z1.d, z2.d, #270
177; CHECK-NEXT: ret
178  %out = call <vscale x 2 x double> @llvm.aarch64.sve.fcmla.nxv2f64(<vscale x 2 x i1> %pg,
179                                                                    <vscale x 2 x double> %a,
180                                                                    <vscale x 2 x double> %b,
181                                                                    <vscale x 2 x double> %c,
182                                                                    i32 270)
183  ret <vscale x 2 x double> %out
184}
185
186;
187; FCMLA (Indexed)
188;
189
190define <vscale x 8 x half> @fcmla_lane_h(<vscale x 8 x half> %a, <vscale x 8 x half> %b, <vscale x 8 x half> %c) {
191; CHECK-LABEL: fcmla_lane_h:
192; CHECK: fcmla z0.h, z1.h, z2.h[3], #0
193; CHECK-NEXT: ret
194  %out = call <vscale x 8 x half> @llvm.aarch64.sve.fcmla.lane.nxv8f16(<vscale x 8 x half> %a,
195                                                                       <vscale x 8 x half> %b,
196                                                                       <vscale x 8 x half> %c,
197                                                                       i32 3,
198                                                                       i32 0)
199  ret <vscale x 8 x half> %out
200}
201
202define <vscale x 4 x float> @fcmla_lane_s(<vscale x 4 x float> %a, <vscale x 4 x float> %b, <vscale x 4 x float> %c) {
203; CHECK-LABEL: fcmla_lane_s:
204; CHECK: fcmla z0.s, z1.s, z2.s[1], #90
205; CHECK-NEXT: ret
206  %out = call <vscale x 4 x float> @llvm.aarch64.sve.fcmla.lane.nxv4f32(<vscale x 4 x float> %a,
207                                                                        <vscale x 4 x float> %b,
208                                                                        <vscale x 4 x float> %c,
209                                                                        i32 1,
210                                                                        i32 90)
211  ret <vscale x 4 x float> %out
212}
213
214;
215; FDIV
216;
217
218define <vscale x 8 x half> @fdiv_h(<vscale x 8 x i1> %pg, <vscale x 8 x half> %a, <vscale x 8 x half> %b) {
219; CHECK-LABEL: fdiv_h:
220; CHECK: fdiv z0.h, p0/m, z0.h, z1.h
221; CHECK-NEXT: ret
222  %out = call <vscale x 8 x half> @llvm.aarch64.sve.fdiv.nxv8f16(<vscale x 8 x i1> %pg,
223                                                                 <vscale x 8 x half> %a,
224                                                                 <vscale x 8 x half> %b)
225  ret <vscale x 8 x half> %out
226}
227
228define <vscale x 4 x float> @fdiv_s(<vscale x 4 x i1> %pg, <vscale x 4 x float> %a, <vscale x 4 x float> %b) {
229; CHECK-LABEL: fdiv_s:
230; CHECK: fdiv z0.s, p0/m, z0.s, z1.s
231; CHECK-NEXT: ret
232  %out = call <vscale x 4 x float> @llvm.aarch64.sve.fdiv.nxv4f32(<vscale x 4 x i1> %pg,
233                                                                  <vscale x 4 x float> %a,
234                                                                  <vscale x 4 x float> %b)
235  ret <vscale x 4 x float> %out
236}
237
238define <vscale x 2 x double> @fdiv_d(<vscale x 2 x i1> %pg, <vscale x 2 x double> %a, <vscale x 2 x double> %b) {
239; CHECK-LABEL: fdiv_d:
240; CHECK: fdiv z0.d, p0/m, z0.d, z1.d
241; CHECK-NEXT: ret
242  %out = call <vscale x 2 x double> @llvm.aarch64.sve.fdiv.nxv2f64(<vscale x 2 x i1> %pg,
243                                                                   <vscale x 2 x double> %a,
244                                                                   <vscale x 2 x double> %b)
245  ret <vscale x 2 x double> %out
246}
247
248;
249; FDIVR
250;
251
252define <vscale x 8 x half> @fdivr_h(<vscale x 8 x i1> %pg, <vscale x 8 x half> %a, <vscale x 8 x half> %b) {
253; CHECK-LABEL: fdivr_h:
254; CHECK: fdivr z0.h, p0/m, z0.h, z1.h
255; CHECK-NEXT: ret
256  %out = call <vscale x 8 x half> @llvm.aarch64.sve.fdivr.nxv8f16(<vscale x 8 x i1> %pg,
257                                                                  <vscale x 8 x half> %a,
258                                                                  <vscale x 8 x half> %b)
259  ret <vscale x 8 x half> %out
260}
261
262define <vscale x 4 x float> @fdivr_s(<vscale x 4 x i1> %pg, <vscale x 4 x float> %a, <vscale x 4 x float> %b) {
263; CHECK-LABEL: fdivr_s:
264; CHECK: fdivr z0.s, p0/m, z0.s, z1.s
265; CHECK-NEXT: ret
266  %out = call <vscale x 4 x float> @llvm.aarch64.sve.fdivr.nxv4f32(<vscale x 4 x i1> %pg,
267                                                                   <vscale x 4 x float> %a,
268                                                                   <vscale x 4 x float> %b)
269  ret <vscale x 4 x float> %out
270}
271
272define <vscale x 2 x double> @fdivr_d(<vscale x 2 x i1> %pg, <vscale x 2 x double> %a, <vscale x 2 x double> %b) {
273; CHECK-LABEL: fdivr_d:
274; CHECK: fdivr z0.d, p0/m, z0.d, z1.d
275; CHECK-NEXT: ret
276  %out = call <vscale x 2 x double> @llvm.aarch64.sve.fdivr.nxv2f64(<vscale x 2 x i1> %pg,
277                                                                    <vscale x 2 x double> %a,
278                                                                    <vscale x 2 x double> %b)
279  ret <vscale x 2 x double> %out
280}
281
282;
283; FEXPA
284;
285
286define <vscale x 8 x half> @fexpa_h(<vscale x 8 x i16> %a) {
287; CHECK-LABEL: fexpa_h:
288; CHECK: fexpa z0.h, z0.h
289; CHECK-NEXT: ret
290  %out = call <vscale x 8 x half> @llvm.aarch64.sve.fexpa.x.nxv8f16(<vscale x 8 x i16> %a)
291  ret <vscale x 8 x half> %out
292}
293
294define <vscale x 4 x float> @fexpa_s(<vscale x 4 x i32> %a) {
295; CHECK-LABEL: fexpa_s:
296; CHECK: fexpa z0.s, z0.s
297; CHECK-NEXT: ret
298  %out = call <vscale x 4 x float> @llvm.aarch64.sve.fexpa.x.nxv4f32(<vscale x 4 x i32> %a)
299  ret <vscale x 4 x float> %out
300}
301
302define <vscale x 2 x double> @fexpa_d(<vscale x 2 x i1> %pg, <vscale x 2 x i64> %a) {
303; CHECK-LABEL: fexpa_d:
304; CHECK: fexpa z0.d, z0.d
305; CHECK-NEXT: ret
306  %out = call <vscale x 2 x double> @llvm.aarch64.sve.fexpa.x.nxv2f64(<vscale x 2 x i64> %a)
307  ret <vscale x 2 x double> %out
308}
309
310;
311; FMAD
312;
313
314define <vscale x 8 x half> @fmad_h(<vscale x 8 x i1> %pg, <vscale x 8 x half> %a, <vscale x 8 x half> %b, <vscale x 8 x half> %c) {
315; CHECK-LABEL: fmad_h:
316; CHECK: fmad z0.h, p0/m, z1.h, z2.h
317; CHECK-NEXT: ret
318  %out = call <vscale x 8 x half> @llvm.aarch64.sve.fmad.nxv8f16(<vscale x 8 x i1> %pg,
319                                                                 <vscale x 8 x half> %a,
320                                                                 <vscale x 8 x half> %b,
321                                                                 <vscale x 8 x half> %c)
322  ret <vscale x 8 x half> %out
323}
324
325define <vscale x 4 x float> @fmad_s(<vscale x 4 x i1> %pg, <vscale x 4 x float> %a, <vscale x 4 x float> %b, <vscale x 4 x float> %c) {
326; CHECK-LABEL: fmad_s:
327; CHECK: fmad z0.s, p0/m, z1.s, z2.s
328; CHECK-NEXT: ret
329  %out = call <vscale x 4 x float> @llvm.aarch64.sve.fmad.nxv4f32(<vscale x 4 x i1> %pg,
330                                                                  <vscale x 4 x float> %a,
331                                                                  <vscale x 4 x float> %b,
332                                                                  <vscale x 4 x float> %c)
333  ret <vscale x 4 x float> %out
334}
335
336define <vscale x 2 x double> @fmad_d(<vscale x 2 x i1> %pg, <vscale x 2 x double> %a, <vscale x 2 x double> %b, <vscale x 2 x double> %c) {
337; CHECK-LABEL: fmad_d:
338; CHECK: fmad z0.d, p0/m, z1.d, z2.d
339; CHECK-NEXT: ret
340  %out = call <vscale x 2 x double> @llvm.aarch64.sve.fmad.nxv2f64(<vscale x 2 x i1> %pg,
341                                                                   <vscale x 2 x double> %a,
342                                                                   <vscale x 2 x double> %b,
343                                                                   <vscale x 2 x double> %c)
344  ret <vscale x 2 x double> %out
345}
346
347;
348; FMAX
349;
350
351define <vscale x 8 x half> @fmax_h(<vscale x 8 x i1> %pg, <vscale x 8 x half> %a, <vscale x 8 x half> %b) {
352; CHECK-LABEL: fmax_h:
353; CHECK: fmax z0.h, p0/m, z0.h, z1.h
354; CHECK-NEXT: ret
355  %out = call <vscale x 8 x half> @llvm.aarch64.sve.fmax.nxv8f16(<vscale x 8 x i1> %pg,
356                                                                 <vscale x 8 x half> %a,
357                                                                 <vscale x 8 x half> %b)
358  ret <vscale x 8 x half> %out
359}
360
361define <vscale x 4 x float> @fmax_s(<vscale x 4 x i1> %pg, <vscale x 4 x float> %a, <vscale x 4 x float> %b) {
362; CHECK-LABEL: fmax_s:
363; CHECK: fmax z0.s, p0/m, z0.s, z1.s
364; CHECK-NEXT: ret
365  %out = call <vscale x 4 x float> @llvm.aarch64.sve.fmax.nxv4f32(<vscale x 4 x i1> %pg,
366                                                                  <vscale x 4 x float> %a,
367                                                                  <vscale x 4 x float> %b)
368  ret <vscale x 4 x float> %out
369}
370
371define <vscale x 2 x double> @fmax_d(<vscale x 2 x i1> %pg, <vscale x 2 x double> %a, <vscale x 2 x double> %b) {
372; CHECK-LABEL: fmax_d:
373; CHECK: fmax z0.d, p0/m, z0.d, z1.d
374; CHECK-NEXT: ret
375  %out = call <vscale x 2 x double> @llvm.aarch64.sve.fmax.nxv2f64(<vscale x 2 x i1> %pg,
376                                                                   <vscale x 2 x double> %a,
377                                                                   <vscale x 2 x double> %b)
378  ret <vscale x 2 x double> %out
379}
380
381;
382; FMAXNM
383;
384
385define <vscale x 8 x half> @fmaxnm_h(<vscale x 8 x i1> %pg, <vscale x 8 x half> %a, <vscale x 8 x half> %b) {
386; CHECK-LABEL: fmaxnm_h:
387; CHECK: fmaxnm z0.h, p0/m, z0.h, z1.h
388; CHECK-NEXT: ret
389  %out = call <vscale x 8 x half> @llvm.aarch64.sve.fmaxnm.nxv8f16(<vscale x 8 x i1> %pg,
390                                                                   <vscale x 8 x half> %a,
391                                                                   <vscale x 8 x half> %b)
392  ret <vscale x 8 x half> %out
393}
394
395define <vscale x 4 x float> @fmaxnm_s(<vscale x 4 x i1> %pg, <vscale x 4 x float> %a, <vscale x 4 x float> %b) {
396; CHECK-LABEL: fmaxnm_s:
397; CHECK: fmaxnm z0.s, p0/m, z0.s, z1.s
398; CHECK-NEXT: ret
399  %out = call <vscale x 4 x float> @llvm.aarch64.sve.fmaxnm.nxv4f32(<vscale x 4 x i1> %pg,
400                                                                    <vscale x 4 x float> %a,
401                                                                    <vscale x 4 x float> %b)
402  ret <vscale x 4 x float> %out
403}
404
405define <vscale x 2 x double> @fmaxnm_d(<vscale x 2 x i1> %pg, <vscale x 2 x double> %a, <vscale x 2 x double> %b) {
406; CHECK-LABEL: fmaxnm_d:
407; CHECK: fmaxnm z0.d, p0/m, z0.d, z1.d
408; CHECK-NEXT: ret
409  %out = call <vscale x 2 x double> @llvm.aarch64.sve.fmaxnm.nxv2f64(<vscale x 2 x i1> %pg,
410                                                                     <vscale x 2 x double> %a,
411                                                                     <vscale x 2 x double> %b)
412  ret <vscale x 2 x double> %out
413}
414
415;
416; FMIN
417;
418
419define <vscale x 8 x half> @fmin_h(<vscale x 8 x i1> %pg, <vscale x 8 x half> %a, <vscale x 8 x half> %b) {
420; CHECK-LABEL: fmin_h:
421; CHECK: fmin z0.h, p0/m, z0.h, z1.h
422; CHECK-NEXT: ret
423  %out = call <vscale x 8 x half> @llvm.aarch64.sve.fmin.nxv8f16(<vscale x 8 x i1> %pg,
424                                                                 <vscale x 8 x half> %a,
425                                                                 <vscale x 8 x half> %b)
426  ret <vscale x 8 x half> %out
427}
428
429define <vscale x 4 x float> @fmin_s(<vscale x 4 x i1> %pg, <vscale x 4 x float> %a, <vscale x 4 x float> %b) {
430; CHECK-LABEL: fmin_s:
431; CHECK: fmin z0.s, p0/m, z0.s, z1.s
432; CHECK-NEXT: ret
433  %out = call <vscale x 4 x float> @llvm.aarch64.sve.fmin.nxv4f32(<vscale x 4 x i1> %pg,
434                                                                  <vscale x 4 x float> %a,
435                                                                  <vscale x 4 x float> %b)
436  ret <vscale x 4 x float> %out
437}
438
439define <vscale x 2 x double> @fmin_d(<vscale x 2 x i1> %pg, <vscale x 2 x double> %a, <vscale x 2 x double> %b) {
440; CHECK-LABEL: fmin_d:
441; CHECK: fmin z0.d, p0/m, z0.d, z1.d
442; CHECK-NEXT: ret
443  %out = call <vscale x 2 x double> @llvm.aarch64.sve.fmin.nxv2f64(<vscale x 2 x i1> %pg,
444                                                                   <vscale x 2 x double> %a,
445                                                                   <vscale x 2 x double> %b)
446  ret <vscale x 2 x double> %out
447}
448
449;
450; FMINNM
451;
452
453define <vscale x 8 x half> @fminnm_h(<vscale x 8 x i1> %pg, <vscale x 8 x half> %a, <vscale x 8 x half> %b) {
454; CHECK-LABEL: fminnm_h:
455; CHECK: fminnm z0.h, p0/m, z0.h, z1.h
456; CHECK-NEXT: ret
457  %out = call <vscale x 8 x half> @llvm.aarch64.sve.fminnm.nxv8f16(<vscale x 8 x i1> %pg,
458                                                                   <vscale x 8 x half> %a,
459                                                                   <vscale x 8 x half> %b)
460  ret <vscale x 8 x half> %out
461}
462
463define <vscale x 4 x float> @fminnm_s(<vscale x 4 x i1> %pg, <vscale x 4 x float> %a, <vscale x 4 x float> %b) {
464; CHECK-LABEL: fminnm_s:
465; CHECK: fminnm z0.s, p0/m, z0.s, z1.s
466; CHECK-NEXT: ret
467  %out = call <vscale x 4 x float> @llvm.aarch64.sve.fminnm.nxv4f32(<vscale x 4 x i1> %pg,
468                                                                    <vscale x 4 x float> %a,
469                                                                    <vscale x 4 x float> %b)
470  ret <vscale x 4 x float> %out
471}
472
473define <vscale x 2 x double> @fminnm_d(<vscale x 2 x i1> %pg, <vscale x 2 x double> %a, <vscale x 2 x double> %b) {
474; CHECK-LABEL: fminnm_d:
475; CHECK: fminnm z0.d, p0/m, z0.d, z1.d
476; CHECK-NEXT: ret
477  %out = call <vscale x 2 x double> @llvm.aarch64.sve.fminnm.nxv2f64(<vscale x 2 x i1> %pg,
478                                                                     <vscale x 2 x double> %a,
479                                                                     <vscale x 2 x double> %b)
480  ret <vscale x 2 x double> %out
481}
482
483;
484; FMLA
485;
486
487define <vscale x 8 x half> @fmla_h(<vscale x 8 x i1> %pg, <vscale x 8 x half> %a, <vscale x 8 x half> %b, <vscale x 8 x half> %c) {
488; CHECK-LABEL: fmla_h:
489; CHECK: fmla z0.h, p0/m, z1.h, z2.h
490; CHECK-NEXT: ret
491  %out = call <vscale x 8 x half> @llvm.aarch64.sve.fmla.nxv8f16(<vscale x 8 x i1> %pg,
492                                                                 <vscale x 8 x half> %a,
493                                                                 <vscale x 8 x half> %b,
494                                                                 <vscale x 8 x half> %c)
495  ret <vscale x 8 x half> %out
496}
497
498define <vscale x 4 x float> @fmla_s(<vscale x 4 x i1> %pg, <vscale x 4 x float> %a, <vscale x 4 x float> %b, <vscale x 4 x float> %c) {
499; CHECK-LABEL: fmla_s:
500; CHECK: fmla z0.s, p0/m, z1.s, z2.s
501; CHECK-NEXT: ret
502  %out = call <vscale x 4 x float> @llvm.aarch64.sve.fmla.nxv4f32(<vscale x 4 x i1> %pg,
503                                                                  <vscale x 4 x float> %a,
504                                                                  <vscale x 4 x float> %b,
505                                                                  <vscale x 4 x float> %c)
506  ret <vscale x 4 x float> %out
507}
508
509define <vscale x 2 x double> @fmla_d(<vscale x 2 x i1> %pg, <vscale x 2 x double> %a, <vscale x 2 x double> %b, <vscale x 2 x double> %c) {
510; CHECK-LABEL: fmla_d:
511; CHECK: fmla z0.d, p0/m, z1.d, z2.d
512; CHECK-NEXT: ret
513  %out = call <vscale x 2 x double> @llvm.aarch64.sve.fmla.nxv2f64(<vscale x 2 x i1> %pg,
514                                                                   <vscale x 2 x double> %a,
515                                                                   <vscale x 2 x double> %b,
516                                                                   <vscale x 2 x double> %c)
517  ret <vscale x 2 x double> %out
518}
519
520;
521; FMLA (Indexed)
522;
523
524define <vscale x 8 x half> @fmla_lane_h(<vscale x 8 x half> %a, <vscale x 8 x half> %b, <vscale x 8 x half> %c) {
525; CHECK-LABEL: fmla_lane_h:
526; CHECK: fmla z0.h, z1.h, z2.h[3]
527; CHECK-NEXT: ret
528  %out = call <vscale x 8 x half> @llvm.aarch64.sve.fmla.lane.nxv8f16(<vscale x 8 x half> %a,
529                                                                      <vscale x 8 x half> %b,
530                                                                      <vscale x 8 x half> %c,
531                                                                      i32 3)
532  ret <vscale x 8 x half> %out
533}
534
535define <vscale x 4 x float> @fmla_lane_s(<vscale x 4 x float> %a, <vscale x 4 x float> %b, <vscale x 4 x float> %c) {
536; CHECK-LABEL: fmla_lane_s:
537; CHECK: fmla z0.s, z1.s, z2.s[2]
538; CHECK-NEXT: ret
539  %out = call <vscale x 4 x float> @llvm.aarch64.sve.fmla.lane.nxv4f32(<vscale x 4 x float> %a,
540                                                                       <vscale x 4 x float> %b,
541                                                                       <vscale x 4 x float> %c,
542                                                                       i32 2)
543  ret <vscale x 4 x float> %out
544}
545
546define <vscale x 2 x double> @fmla_lane_d(<vscale x 2 x double> %a, <vscale x 2 x double> %b, <vscale x 2 x double> %c) {
547; CHECK-LABEL: fmla_lane_d:
548; CHECK: fmla z0.d, z1.d, z2.d[1]
549; CHECK-NEXT: ret
550  %out = call <vscale x 2 x double> @llvm.aarch64.sve.fmla.lane.nxv2f64(<vscale x 2 x double> %a,
551                                                                        <vscale x 2 x double> %b,
552                                                                        <vscale x 2 x double> %c,
553                                                                        i32 1)
554  ret <vscale x 2 x double> %out
555}
556
557;
558; FMLS
559;
560
561define <vscale x 8 x half> @fmls_h(<vscale x 8 x i1> %pg, <vscale x 8 x half> %a, <vscale x 8 x half> %b, <vscale x 8 x half> %c) {
562; CHECK-LABEL: fmls_h:
563; CHECK: fmls z0.h, p0/m, z1.h, z2.h
564; CHECK-NEXT: ret
565  %out = call <vscale x 8 x half> @llvm.aarch64.sve.fmls.nxv8f16(<vscale x 8 x i1> %pg,
566                                                                 <vscale x 8 x half> %a,
567                                                                 <vscale x 8 x half> %b,
568                                                                 <vscale x 8 x half> %c)
569  ret <vscale x 8 x half> %out
570}
571
572define <vscale x 4 x float> @fmls_s(<vscale x 4 x i1> %pg, <vscale x 4 x float> %a, <vscale x 4 x float> %b, <vscale x 4 x float> %c) {
573; CHECK-LABEL: fmls_s:
574; CHECK: fmls z0.s, p0/m, z1.s, z2.s
575; CHECK-NEXT: ret
576  %out = call <vscale x 4 x float> @llvm.aarch64.sve.fmls.nxv4f32(<vscale x 4 x i1> %pg,
577                                                                  <vscale x 4 x float> %a,
578                                                                  <vscale x 4 x float> %b,
579                                                                  <vscale x 4 x float> %c)
580  ret <vscale x 4 x float> %out
581}
582
583define <vscale x 2 x double> @fmls_d(<vscale x 2 x i1> %pg, <vscale x 2 x double> %a, <vscale x 2 x double> %b, <vscale x 2 x double> %c) {
584; CHECK-LABEL: fmls_d:
585; CHECK: fmls z0.d, p0/m, z1.d, z2.d
586; CHECK-NEXT: ret
587  %out = call <vscale x 2 x double> @llvm.aarch64.sve.fmls.nxv2f64(<vscale x 2 x i1> %pg,
588                                                                   <vscale x 2 x double> %a,
589                                                                   <vscale x 2 x double> %b,
590                                                                   <vscale x 2 x double> %c)
591  ret <vscale x 2 x double> %out
592}
593
594;
595; FMLS (Indexed)
596;
597
598define <vscale x 8 x half> @fmls_lane_h(<vscale x 8 x half> %a, <vscale x 8 x half> %b, <vscale x 8 x half> %c) {
599; CHECK-LABEL: fmls_lane_h:
600; CHECK: fmls z0.h, z1.h, z2.h[3]
601; CHECK-NEXT: ret
602  %out = call <vscale x 8 x half> @llvm.aarch64.sve.fmls.lane.nxv8f16(<vscale x 8 x half> %a,
603                                                                 <vscale x 8 x half> %b,
604                                                                 <vscale x 8 x half> %c,
605                                                                 i32 3)
606  ret <vscale x 8 x half> %out
607}
608
609define <vscale x 4 x float> @fmls_lane_s(<vscale x 4 x float> %a, <vscale x 4 x float> %b, <vscale x 4 x float> %c) {
610; CHECK-LABEL: fmls_lane_s:
611; CHECK: fmls z0.s, z1.s, z2.s[2]
612; CHECK-NEXT: ret
613  %out = call <vscale x 4 x float> @llvm.aarch64.sve.fmls.lane.nxv4f32(<vscale x 4 x float> %a,
614                                                                  <vscale x 4 x float> %b,
615                                                                  <vscale x 4 x float> %c,
616                                                                  i32 2)
617  ret <vscale x 4 x float> %out
618}
619
620define <vscale x 2 x double> @fmls_lane_d(<vscale x 2 x double> %a, <vscale x 2 x double> %b, <vscale x 2 x double> %c) {
621; CHECK-LABEL: fmls_lane_d:
622; CHECK: fmls z0.d, z1.d, z2.d[1]
623; CHECK-NEXT: ret
624  %out = call <vscale x 2 x double> @llvm.aarch64.sve.fmls.lane.nxv2f64(<vscale x 2 x double> %a,
625                                                                   <vscale x 2 x double> %b,
626                                                                   <vscale x 2 x double> %c,
627                                                                   i32 1)
628  ret <vscale x 2 x double> %out
629}
630
631;
632; FMSB
633;
634
635define <vscale x 8 x half> @fmsb_h(<vscale x 8 x i1> %pg, <vscale x 8 x half> %a, <vscale x 8 x half> %b, <vscale x 8 x half> %c) {
636; CHECK-LABEL: fmsb_h:
637; CHECK: fmsb z0.h, p0/m, z1.h, z2.h
638; CHECK-NEXT: ret
639  %out = call <vscale x 8 x half> @llvm.aarch64.sve.fmsb.nxv8f16(<vscale x 8 x i1> %pg,
640                                                                 <vscale x 8 x half> %a,
641                                                                 <vscale x 8 x half> %b,
642                                                                 <vscale x 8 x half> %c)
643  ret <vscale x 8 x half> %out
644}
645
646define <vscale x 4 x float> @fmsb_s(<vscale x 4 x i1> %pg, <vscale x 4 x float> %a, <vscale x 4 x float> %b, <vscale x 4 x float> %c) {
647; CHECK-LABEL: fmsb_s:
648; CHECK: fmsb z0.s, p0/m, z1.s, z2.s
649; CHECK-NEXT: ret
650  %out = call <vscale x 4 x float> @llvm.aarch64.sve.fmsb.nxv4f32(<vscale x 4 x i1> %pg,
651                                                                  <vscale x 4 x float> %a,
652                                                                  <vscale x 4 x float> %b,
653                                                                  <vscale x 4 x float> %c)
654  ret <vscale x 4 x float> %out
655}
656
657define <vscale x 2 x double> @fmsb_d(<vscale x 2 x i1> %pg, <vscale x 2 x double> %a, <vscale x 2 x double> %b, <vscale x 2 x double> %c) {
658; CHECK-LABEL: fmsb_d:
659; CHECK: fmsb z0.d, p0/m, z1.d, z2.d
660; CHECK-NEXT: ret
661  %out = call <vscale x 2 x double> @llvm.aarch64.sve.fmsb.nxv2f64(<vscale x 2 x i1> %pg,
662                                                                   <vscale x 2 x double> %a,
663                                                                   <vscale x 2 x double> %b,
664                                                                   <vscale x 2 x double> %c)
665  ret <vscale x 2 x double> %out
666}
667
668;
669; FMUL
670;
671
672define <vscale x 8 x half> @fmul_h(<vscale x 8 x i1> %pg, <vscale x 8 x half> %a, <vscale x 8 x half> %b) {
673; CHECK-LABEL: fmul_h:
674; CHECK: fmul z0.h, p0/m, z0.h, z1.h
675; CHECK-NEXT: ret
676  %out = call <vscale x 8 x half> @llvm.aarch64.sve.fmul.nxv8f16(<vscale x 8 x i1> %pg,
677                                                                 <vscale x 8 x half> %a,
678                                                                 <vscale x 8 x half> %b)
679  ret <vscale x 8 x half> %out
680}
681
682define <vscale x 4 x float> @fmul_s(<vscale x 4 x i1> %pg, <vscale x 4 x float> %a, <vscale x 4 x float> %b) {
683; CHECK-LABEL: fmul_s:
684; CHECK: fmul z0.s, p0/m, z0.s, z1.s
685; CHECK-NEXT: ret
686  %out = call <vscale x 4 x float> @llvm.aarch64.sve.fmul.nxv4f32(<vscale x 4 x i1> %pg,
687                                                                  <vscale x 4 x float> %a,
688                                                                  <vscale x 4 x float> %b)
689  ret <vscale x 4 x float> %out
690}
691
692define <vscale x 2 x double> @fmul_d(<vscale x 2 x i1> %pg, <vscale x 2 x double> %a, <vscale x 2 x double> %b) {
693; CHECK-LABEL: fmul_d:
694; CHECK: fmul z0.d, p0/m, z0.d, z1.d
695; CHECK-NEXT: ret
696  %out = call <vscale x 2 x double> @llvm.aarch64.sve.fmul.nxv2f64(<vscale x 2 x i1> %pg,
697                                                                   <vscale x 2 x double> %a,
698                                                                   <vscale x 2 x double> %b)
699  ret <vscale x 2 x double> %out
700}
701
702;
703; FMUL (Indexed)
704;
705
706define <vscale x 8 x half> @fmul_lane_h(<vscale x 8 x half> %a, <vscale x 8 x half> %b) {
707; CHECK-LABEL: fmul_lane_h:
708; CHECK: fmul z0.h, z0.h, z1.h[3]
709; CHECK-NEXT: ret
710  %out = call <vscale x 8 x half> @llvm.aarch64.sve.fmul.lane.nxv8f16(<vscale x 8 x half> %a,
711                                                                      <vscale x 8 x half> %b,
712                                                                      i32 3)
713  ret <vscale x 8 x half> %out
714}
715
716define <vscale x 4 x float> @fmul_lane_s(<vscale x 4 x float> %a, <vscale x 4 x float> %b) {
717; CHECK-LABEL: fmul_lane_s:
718; CHECK: fmul z0.s, z0.s, z1.s[2]
719; CHECK-NEXT: ret
720  %out = call <vscale x 4 x float> @llvm.aarch64.sve.fmul.lane.nxv4f32(<vscale x 4 x float> %a,
721                                                                       <vscale x 4 x float> %b,
722                                                                       i32 2)
723  ret <vscale x 4 x float> %out
724}
725
726define <vscale x 2 x double> @fmul_lane_d(<vscale x 2 x double> %a, <vscale x 2 x double> %b) {
727; CHECK-LABEL: fmul_lane_d:
728; CHECK: fmul z0.d, z0.d, z1.d[1]
729; CHECK-NEXT: ret
730  %out = call <vscale x 2 x double> @llvm.aarch64.sve.fmul.lane.nxv2f64(<vscale x 2 x double> %a,
731                                                                        <vscale x 2 x double> %b,
732                                                                        i32 1)
733  ret <vscale x 2 x double> %out
734}
735
736;
737; FMULX
738;
739
740define <vscale x 8 x half> @fmulx_h(<vscale x 8 x i1> %pg, <vscale x 8 x half> %a, <vscale x 8 x half> %b) {
741; CHECK-LABEL: fmulx_h:
742; CHECK: fmulx z0.h, p0/m, z0.h, z1.h
743; CHECK-NEXT: ret
744  %out = call <vscale x 8 x half> @llvm.aarch64.sve.fmulx.nxv8f16(<vscale x 8 x i1> %pg,
745                                                                  <vscale x 8 x half> %a,
746                                                                  <vscale x 8 x half> %b)
747  ret <vscale x 8 x half> %out
748}
749
750define <vscale x 4 x float> @fmulx_s(<vscale x 4 x i1> %pg, <vscale x 4 x float> %a, <vscale x 4 x float> %b) {
751; CHECK-LABEL: fmulx_s:
752; CHECK: fmulx z0.s, p0/m, z0.s, z1.s
753; CHECK-NEXT: ret
754  %out = call <vscale x 4 x float> @llvm.aarch64.sve.fmulx.nxv4f32(<vscale x 4 x i1> %pg,
755                                                                   <vscale x 4 x float> %a,
756                                                                   <vscale x 4 x float> %b)
757  ret <vscale x 4 x float> %out
758}
759
760define <vscale x 2 x double> @fmulx_d(<vscale x 2 x i1> %pg, <vscale x 2 x double> %a, <vscale x 2 x double> %b) {
761; CHECK-LABEL: fmulx_d:
762; CHECK: fmulx z0.d, p0/m, z0.d, z1.d
763; CHECK-NEXT: ret
764  %out = call <vscale x 2 x double> @llvm.aarch64.sve.fmulx.nxv2f64(<vscale x 2 x i1> %pg,
765                                                                    <vscale x 2 x double> %a,
766                                                                    <vscale x 2 x double> %b)
767  ret <vscale x 2 x double> %out
768}
769
770;
771; FNEG
772;
773
774define <vscale x 8 x half> @fneg_h(<vscale x 8 x half> %a, <vscale x 8 x i1> %pg, <vscale x 8 x half> %b) {
775; CHECK-LABEL: fneg_h:
776; CHECK: fneg z0.h, p0/m, z1.h
777; CHECK-NEXT: ret
778  %out = call <vscale x 8 x half> @llvm.aarch64.sve.fneg.nxv8f16(<vscale x 8 x half> %a,
779                                                                 <vscale x 8 x i1> %pg,
780                                                                 <vscale x 8 x half> %b)
781  ret <vscale x 8 x half> %out
782}
783
784define <vscale x 4 x float> @fneg_s(<vscale x 4 x float> %a, <vscale x 4 x i1> %pg, <vscale x 4 x float> %b) {
785; CHECK-LABEL: fneg_s:
786; CHECK: fneg z0.s, p0/m, z1.s
787; CHECK-NEXT: ret
788  %out = call <vscale x 4 x float> @llvm.aarch64.sve.fneg.nxv4f32(<vscale x 4 x float> %a,
789                                                                  <vscale x 4 x i1> %pg,
790                                                                  <vscale x 4 x float> %b)
791  ret <vscale x 4 x float> %out
792}
793
794define <vscale x 2 x double> @fneg_d(<vscale x 2 x double> %a, <vscale x 2 x i1> %pg, <vscale x 2 x double> %b) {
795; CHECK-LABEL: fneg_d:
796; CHECK: fneg z0.d, p0/m, z1.d
797; CHECK-NEXT: ret
798  %out = call <vscale x 2 x double> @llvm.aarch64.sve.fneg.nxv2f64(<vscale x 2 x double> %a,
799                                                                   <vscale x 2 x i1> %pg,
800                                                                   <vscale x 2 x double> %b)
801  ret <vscale x 2 x double> %out
802}
803
804;
805; FNMAD
806;
807
808define <vscale x 8 x half> @fnmad_h(<vscale x 8 x i1> %pg, <vscale x 8 x half> %a, <vscale x 8 x half> %b, <vscale x 8 x half> %c) {
809; CHECK-LABEL: fnmad_h:
810; CHECK: fnmad z0.h, p0/m, z1.h, z2.h
811; CHECK-NEXT: ret
812  %out = call <vscale x 8 x half> @llvm.aarch64.sve.fnmad.nxv8f16(<vscale x 8 x i1> %pg,
813                                                                  <vscale x 8 x half> %a,
814                                                                  <vscale x 8 x half> %b,
815                                                                  <vscale x 8 x half> %c)
816  ret <vscale x 8 x half> %out
817}
818
819define <vscale x 4 x float> @fnmad_s(<vscale x 4 x i1> %pg, <vscale x 4 x float> %a, <vscale x 4 x float> %b, <vscale x 4 x float> %c) {
820; CHECK-LABEL: fnmad_s:
821; CHECK: fnmad z0.s, p0/m, z1.s, z2.s
822; CHECK-NEXT: ret
823  %out = call <vscale x 4 x float> @llvm.aarch64.sve.fnmad.nxv4f32(<vscale x 4 x i1> %pg,
824                                                                   <vscale x 4 x float> %a,
825                                                                   <vscale x 4 x float> %b,
826                                                                   <vscale x 4 x float> %c)
827  ret <vscale x 4 x float> %out
828}
829
830define <vscale x 2 x double> @fnmad_d(<vscale x 2 x i1> %pg, <vscale x 2 x double> %a, <vscale x 2 x double> %b, <vscale x 2 x double> %c) {
831; CHECK-LABEL: fnmad_d:
832; CHECK: fnmad z0.d, p0/m, z1.d, z2.d
833; CHECK-NEXT: ret
834  %out = call <vscale x 2 x double> @llvm.aarch64.sve.fnmad.nxv2f64(<vscale x 2 x i1> %pg,
835                                                                    <vscale x 2 x double> %a,
836                                                                    <vscale x 2 x double> %b,
837                                                                    <vscale x 2 x double> %c)
838  ret <vscale x 2 x double> %out
839}
840
841;
842; FNMLA
843;
844
845define <vscale x 8 x half> @fnmla_h(<vscale x 8 x i1> %pg, <vscale x 8 x half> %a, <vscale x 8 x half> %b, <vscale x 8 x half> %c) {
846; CHECK-LABEL: fnmla_h:
847; CHECK: fnmla z0.h, p0/m, z1.h, z2.h
848; CHECK-NEXT: ret
849  %out = call <vscale x 8 x half> @llvm.aarch64.sve.fnmla.nxv8f16(<vscale x 8 x i1> %pg,
850                                                                  <vscale x 8 x half> %a,
851                                                                  <vscale x 8 x half> %b,
852                                                                  <vscale x 8 x half> %c)
853  ret <vscale x 8 x half> %out
854}
855
856define <vscale x 4 x float> @fnmla_s(<vscale x 4 x i1> %pg, <vscale x 4 x float> %a, <vscale x 4 x float> %b, <vscale x 4 x float> %c) {
857; CHECK-LABEL: fnmla_s:
858; CHECK: fnmla z0.s, p0/m, z1.s, z2.s
859; CHECK-NEXT: ret
860  %out = call <vscale x 4 x float> @llvm.aarch64.sve.fnmla.nxv4f32(<vscale x 4 x i1> %pg,
861                                                                   <vscale x 4 x float> %a,
862                                                                   <vscale x 4 x float> %b,
863                                                                   <vscale x 4 x float> %c)
864  ret <vscale x 4 x float> %out
865}
866
867define <vscale x 2 x double> @fnmla_d(<vscale x 2 x i1> %pg, <vscale x 2 x double> %a, <vscale x 2 x double> %b, <vscale x 2 x double> %c) {
868; CHECK-LABEL: fnmla_d:
869; CHECK: fnmla z0.d, p0/m, z1.d, z2.d
870; CHECK-NEXT: ret
871  %out = call <vscale x 2 x double> @llvm.aarch64.sve.fnmla.nxv2f64(<vscale x 2 x i1> %pg,
872                                                                    <vscale x 2 x double> %a,
873                                                                    <vscale x 2 x double> %b,
874                                                                    <vscale x 2 x double> %c)
875  ret <vscale x 2 x double> %out
876}
877
878;
879; FNMLS
880;
881
882define <vscale x 8 x half> @fnmls_h(<vscale x 8 x i1> %pg, <vscale x 8 x half> %a, <vscale x 8 x half> %b, <vscale x 8 x half> %c) {
883; CHECK-LABEL: fnmls_h:
884; CHECK: fnmls z0.h, p0/m, z1.h, z2.h
885; CHECK-NEXT: ret
886  %out = call <vscale x 8 x half> @llvm.aarch64.sve.fnmls.nxv8f16(<vscale x 8 x i1> %pg,
887                                                                  <vscale x 8 x half> %a,
888                                                                  <vscale x 8 x half> %b,
889                                                                  <vscale x 8 x half> %c)
890  ret <vscale x 8 x half> %out
891}
892
893define <vscale x 4 x float> @fnmls_s(<vscale x 4 x i1> %pg, <vscale x 4 x float> %a, <vscale x 4 x float> %b, <vscale x 4 x float> %c) {
894; CHECK-LABEL: fnmls_s:
895; CHECK: fnmls z0.s, p0/m, z1.s, z2.s
896; CHECK-NEXT: ret
897  %out = call <vscale x 4 x float> @llvm.aarch64.sve.fnmls.nxv4f32(<vscale x 4 x i1> %pg,
898                                                                   <vscale x 4 x float> %a,
899                                                                   <vscale x 4 x float> %b,
900                                                                   <vscale x 4 x float> %c)
901  ret <vscale x 4 x float> %out
902}
903
904define <vscale x 2 x double> @fnmls_d(<vscale x 2 x i1> %pg, <vscale x 2 x double> %a, <vscale x 2 x double> %b, <vscale x 2 x double> %c) {
905; CHECK-LABEL: fnmls_d:
906; CHECK: fnmls z0.d, p0/m, z1.d, z2.d
907; CHECK-NEXT: ret
908  %out = call <vscale x 2 x double> @llvm.aarch64.sve.fnmls.nxv2f64(<vscale x 2 x i1> %pg,
909                                                                    <vscale x 2 x double> %a,
910                                                                    <vscale x 2 x double> %b,
911                                                                    <vscale x 2 x double> %c)
912  ret <vscale x 2 x double> %out
913}
914
915;
916; FNMSB
917;
918
919define <vscale x 8 x half> @fnmsb_h(<vscale x 8 x i1> %pg, <vscale x 8 x half> %a, <vscale x 8 x half> %b, <vscale x 8 x half> %c) {
920; CHECK-LABEL: fnmsb_h:
921; CHECK: fnmsb z0.h, p0/m, z1.h, z2.h
922; CHECK-NEXT: ret
923  %out = call <vscale x 8 x half> @llvm.aarch64.sve.fnmsb.nxv8f16(<vscale x 8 x i1> %pg,
924                                                                  <vscale x 8 x half> %a,
925                                                                  <vscale x 8 x half> %b,
926                                                                  <vscale x 8 x half> %c)
927  ret <vscale x 8 x half> %out
928}
929
930define <vscale x 4 x float> @fnmsb_s(<vscale x 4 x i1> %pg, <vscale x 4 x float> %a, <vscale x 4 x float> %b, <vscale x 4 x float> %c) {
931; CHECK-LABEL: fnmsb_s:
932; CHECK: fnmsb z0.s, p0/m, z1.s, z2.s
933; CHECK-NEXT: ret
934  %out = call <vscale x 4 x float> @llvm.aarch64.sve.fnmsb.nxv4f32(<vscale x 4 x i1> %pg,
935                                                                   <vscale x 4 x float> %a,
936                                                                   <vscale x 4 x float> %b,
937                                                                   <vscale x 4 x float> %c)
938  ret <vscale x 4 x float> %out
939}
940
941define <vscale x 2 x double> @fnmsb_d(<vscale x 2 x i1> %pg, <vscale x 2 x double> %a, <vscale x 2 x double> %b, <vscale x 2 x double> %c) {
942; CHECK-LABEL: fnmsb_d:
943; CHECK: fnmsb z0.d, p0/m, z1.d, z2.d
944; CHECK-NEXT: ret
945  %out = call <vscale x 2 x double> @llvm.aarch64.sve.fnmsb.nxv2f64(<vscale x 2 x i1> %pg,
946                                                                    <vscale x 2 x double> %a,
947                                                                    <vscale x 2 x double> %b,
948                                                                    <vscale x 2 x double> %c)
949  ret <vscale x 2 x double> %out
950}
951
952;
953; FRECPE
954;
955
956define <vscale x 8 x half> @frecpe_h(<vscale x 8 x half> %a) {
957; CHECK-LABEL: frecpe_h:
958; CHECK: frecpe z0.h, z0.h
959; CHECK-NEXT: ret
960  %out = call <vscale x 8 x half> @llvm.aarch64.sve.frecpe.x.nxv8f16(<vscale x 8 x half> %a)
961  ret <vscale x 8 x half> %out
962}
963
964define <vscale x 4 x float> @frecpe_s(<vscale x 4 x float> %a) {
965; CHECK-LABEL: frecpe_s:
966; CHECK: frecpe z0.s, z0.s
967; CHECK-NEXT: ret
968  %out = call <vscale x 4 x float> @llvm.aarch64.sve.frecpe.x.nxv4f32(<vscale x 4 x float> %a)
969  ret <vscale x 4 x float> %out
970}
971
972define <vscale x 2 x double> @frecpe_d(<vscale x 2 x i1> %pg, <vscale x 2 x double> %a) {
973; CHECK-LABEL: frecpe_d:
974; CHECK: frecpe z0.d, z0.d
975; CHECK-NEXT: ret
976  %out = call <vscale x 2 x double> @llvm.aarch64.sve.frecpe.x.nxv2f64(<vscale x 2 x double> %a)
977  ret <vscale x 2 x double> %out
978}
979
980;
981; FRECPX
982;
983
984define <vscale x 8 x half> @frecpx_h(<vscale x 8 x half> %a, <vscale x 8 x i1> %pg, <vscale x 8 x half> %b) {
985; CHECK-LABEL: frecpx_h:
986; CHECK: frecpx z0.h, p0/m, z1.h
987; CHECK-NEXT: ret
988  %out = call <vscale x 8 x half> @llvm.aarch64.sve.frecpx.nxv8f16(<vscale x 8 x half> %a,
989                                                                  <vscale x 8 x i1> %pg,
990                                                                  <vscale x 8 x half> %b)
991  ret <vscale x 8 x half> %out
992}
993
994define <vscale x 4 x float> @frecpx_s(<vscale x 4 x float> %a, <vscale x 4 x i1> %pg, <vscale x 4 x float> %b) {
995; CHECK-LABEL: frecpx_s:
996; CHECK: frecpx z0.s, p0/m, z1.s
997; CHECK-NEXT: ret
998  %out = call <vscale x 4 x float> @llvm.aarch64.sve.frecpx.nxv4f32(<vscale x 4 x float> %a,
999                                                                   <vscale x 4 x i1> %pg,
1000                                                                   <vscale x 4 x float> %b)
1001  ret <vscale x 4 x float> %out
1002}
1003
1004define <vscale x 2 x double> @frecpx_d(<vscale x 2 x double> %a, <vscale x 2 x i1> %pg, <vscale x 2 x double> %b) {
1005; CHECK-LABEL: frecpx_d:
1006; CHECK: frecpx z0.d, p0/m, z1.d
1007; CHECK-NEXT: ret
1008  %out = call <vscale x 2 x double> @llvm.aarch64.sve.frecpx.nxv2f64(<vscale x 2 x double> %a,
1009                                                                    <vscale x 2 x i1> %pg,
1010                                                                    <vscale x 2 x double> %b)
1011  ret <vscale x 2 x double> %out
1012}
1013
1014;
1015; FRINTA
1016;
1017
1018define <vscale x 8 x half> @frinta_h(<vscale x 8 x half> %a, <vscale x 8 x i1> %pg, <vscale x 8 x half> %b) {
1019; CHECK-LABEL: frinta_h:
1020; CHECK: frinta z0.h, p0/m, z1.h
1021; CHECK-NEXT: ret
1022  %out = call <vscale x 8 x half> @llvm.aarch64.sve.frinta.nxv8f16(<vscale x 8 x half> %a,
1023                                                                   <vscale x 8 x i1> %pg,
1024                                                                   <vscale x 8 x half> %b)
1025  ret <vscale x 8 x half> %out
1026}
1027
1028define <vscale x 4 x float> @frinta_s(<vscale x 4 x float> %a, <vscale x 4 x i1> %pg, <vscale x 4 x float> %b) {
1029; CHECK-LABEL: frinta_s:
1030; CHECK: frinta z0.s, p0/m, z1.s
1031; CHECK-NEXT: ret
1032  %out = call <vscale x 4 x float> @llvm.aarch64.sve.frinta.nxv4f32(<vscale x 4 x float> %a,
1033                                                                    <vscale x 4 x i1> %pg,
1034                                                                    <vscale x 4 x float> %b)
1035  ret <vscale x 4 x float> %out
1036}
1037
1038define <vscale x 2 x double> @frinta_d(<vscale x 2 x double> %a, <vscale x 2 x i1> %pg, <vscale x 2 x double> %b) {
1039; CHECK-LABEL: frinta_d:
1040; CHECK: frinta z0.d, p0/m, z1.d
1041; CHECK-NEXT: ret
1042  %out = call <vscale x 2 x double> @llvm.aarch64.sve.frinta.nxv2f64(<vscale x 2 x double> %a,
1043                                                                     <vscale x 2 x i1> %pg,
1044                                                                     <vscale x 2 x double> %b)
1045  ret <vscale x 2 x double> %out
1046}
1047
1048;
1049; FRINTI
1050;
1051
1052define <vscale x 8 x half> @frinti_h(<vscale x 8 x half> %a, <vscale x 8 x i1> %pg, <vscale x 8 x half> %b) {
1053; CHECK-LABEL: frinti_h:
1054; CHECK: frinti z0.h, p0/m, z1.h
1055; CHECK-NEXT: ret
1056  %out = call <vscale x 8 x half> @llvm.aarch64.sve.frinti.nxv8f16(<vscale x 8 x half> %a,
1057                                                                   <vscale x 8 x i1> %pg,
1058                                                                   <vscale x 8 x half> %b)
1059  ret <vscale x 8 x half> %out
1060}
1061
1062define <vscale x 4 x float> @frinti_s(<vscale x 4 x float> %a, <vscale x 4 x i1> %pg, <vscale x 4 x float> %b) {
1063; CHECK-LABEL: frinti_s:
1064; CHECK: frinti z0.s, p0/m, z1.s
1065; CHECK-NEXT: ret
1066  %out = call <vscale x 4 x float> @llvm.aarch64.sve.frinti.nxv4f32(<vscale x 4 x float> %a,
1067                                                                    <vscale x 4 x i1> %pg,
1068                                                                    <vscale x 4 x float> %b)
1069  ret <vscale x 4 x float> %out
1070}
1071
1072define <vscale x 2 x double> @frinti_d(<vscale x 2 x double> %a, <vscale x 2 x i1> %pg, <vscale x 2 x double> %b) {
1073; CHECK-LABEL: frinti_d:
1074; CHECK: frinti z0.d, p0/m, z1.d
1075; CHECK-NEXT: ret
1076  %out = call <vscale x 2 x double> @llvm.aarch64.sve.frinti.nxv2f64(<vscale x 2 x double> %a,
1077                                                                     <vscale x 2 x i1> %pg,
1078                                                                     <vscale x 2 x double> %b)
1079  ret <vscale x 2 x double> %out
1080}
1081
1082;
1083; FRINTM
1084;
1085
1086define <vscale x 8 x half> @frintm_h(<vscale x 8 x half> %a, <vscale x 8 x i1> %pg, <vscale x 8 x half> %b) {
1087; CHECK-LABEL: frintm_h:
1088; CHECK: frintm z0.h, p0/m, z1.h
1089; CHECK-NEXT: ret
1090  %out = call <vscale x 8 x half> @llvm.aarch64.sve.frintm.nxv8f16(<vscale x 8 x half> %a,
1091                                                                   <vscale x 8 x i1> %pg,
1092                                                                   <vscale x 8 x half> %b)
1093  ret <vscale x 8 x half> %out
1094}
1095
1096define <vscale x 4 x float> @frintm_s(<vscale x 4 x float> %a, <vscale x 4 x i1> %pg, <vscale x 4 x float> %b) {
1097; CHECK-LABEL: frintm_s:
1098; CHECK: frintm z0.s, p0/m, z1.s
1099; CHECK-NEXT: ret
1100  %out = call <vscale x 4 x float> @llvm.aarch64.sve.frintm.nxv4f32(<vscale x 4 x float> %a,
1101                                                                    <vscale x 4 x i1> %pg,
1102                                                                    <vscale x 4 x float> %b)
1103  ret <vscale x 4 x float> %out
1104}
1105
1106define <vscale x 2 x double> @frintm_d(<vscale x 2 x double> %a, <vscale x 2 x i1> %pg, <vscale x 2 x double> %b) {
1107; CHECK-LABEL: frintm_d:
1108; CHECK: frintm z0.d, p0/m, z1.d
1109; CHECK-NEXT: ret
1110  %out = call <vscale x 2 x double> @llvm.aarch64.sve.frintm.nxv2f64(<vscale x 2 x double> %a,
1111                                                                     <vscale x 2 x i1> %pg,
1112                                                                     <vscale x 2 x double> %b)
1113  ret <vscale x 2 x double> %out
1114}
1115
1116;
1117; FRINTN
1118;
1119
1120define <vscale x 8 x half> @frintn_h(<vscale x 8 x half> %a, <vscale x 8 x i1> %pg, <vscale x 8 x half> %b) {
1121; CHECK-LABEL: frintn_h:
1122; CHECK: frintn z0.h, p0/m, z1.h
1123; CHECK-NEXT: ret
1124  %out = call <vscale x 8 x half> @llvm.aarch64.sve.frintn.nxv8f16(<vscale x 8 x half> %a,
1125                                                                   <vscale x 8 x i1> %pg,
1126                                                                   <vscale x 8 x half> %b)
1127  ret <vscale x 8 x half> %out
1128}
1129
1130define <vscale x 4 x float> @frintn_s(<vscale x 4 x float> %a, <vscale x 4 x i1> %pg, <vscale x 4 x float> %b) {
1131; CHECK-LABEL: frintn_s:
1132; CHECK: frintn z0.s, p0/m, z1.s
1133; CHECK-NEXT: ret
1134  %out = call <vscale x 4 x float> @llvm.aarch64.sve.frintn.nxv4f32(<vscale x 4 x float> %a,
1135                                                                    <vscale x 4 x i1> %pg,
1136                                                                    <vscale x 4 x float> %b)
1137  ret <vscale x 4 x float> %out
1138}
1139
1140define <vscale x 2 x double> @frintn_d(<vscale x 2 x double> %a, <vscale x 2 x i1> %pg, <vscale x 2 x double> %b) {
1141; CHECK-LABEL: frintn_d:
1142; CHECK: frintn z0.d, p0/m, z1.d
1143; CHECK-NEXT: ret
1144  %out = call <vscale x 2 x double> @llvm.aarch64.sve.frintn.nxv2f64(<vscale x 2 x double> %a,
1145                                                                     <vscale x 2 x i1> %pg,
1146                                                                     <vscale x 2 x double> %b)
1147  ret <vscale x 2 x double> %out
1148}
1149
1150;
1151; FRINTP
1152;
1153
1154define <vscale x 8 x half> @frintp_h(<vscale x 8 x half> %a, <vscale x 8 x i1> %pg, <vscale x 8 x half> %b) {
1155; CHECK-LABEL: frintp_h:
1156; CHECK: frintp z0.h, p0/m, z1.h
1157; CHECK-NEXT: ret
1158  %out = call <vscale x 8 x half> @llvm.aarch64.sve.frintp.nxv8f16(<vscale x 8 x half> %a,
1159                                                                   <vscale x 8 x i1> %pg,
1160                                                                   <vscale x 8 x half> %b)
1161  ret <vscale x 8 x half> %out
1162}
1163
1164define <vscale x 4 x float> @frintp_s(<vscale x 4 x float> %a, <vscale x 4 x i1> %pg, <vscale x 4 x float> %b) {
1165; CHECK-LABEL: frintp_s:
1166; CHECK: frintp z0.s, p0/m, z1.s
1167; CHECK-NEXT: ret
1168  %out = call <vscale x 4 x float> @llvm.aarch64.sve.frintp.nxv4f32(<vscale x 4 x float> %a,
1169                                                                    <vscale x 4 x i1> %pg,
1170                                                                    <vscale x 4 x float> %b)
1171  ret <vscale x 4 x float> %out
1172}
1173
1174define <vscale x 2 x double> @frintp_d(<vscale x 2 x double> %a, <vscale x 2 x i1> %pg, <vscale x 2 x double> %b) {
1175; CHECK-LABEL: frintp_d:
1176; CHECK: frintp z0.d, p0/m, z1.d
1177; CHECK-NEXT: ret
1178  %out = call <vscale x 2 x double> @llvm.aarch64.sve.frintp.nxv2f64(<vscale x 2 x double> %a,
1179                                                                     <vscale x 2 x i1> %pg,
1180                                                                     <vscale x 2 x double> %b)
1181  ret <vscale x 2 x double> %out
1182}
1183
1184;
1185; FRINTX
1186;
1187
1188define <vscale x 8 x half> @frintx_h(<vscale x 8 x half> %a, <vscale x 8 x i1> %pg, <vscale x 8 x half> %b) {
1189; CHECK-LABEL: frintx_h:
1190; CHECK: frintx z0.h, p0/m, z1.h
1191; CHECK-NEXT: ret
1192  %out = call <vscale x 8 x half> @llvm.aarch64.sve.frintx.nxv8f16(<vscale x 8 x half> %a,
1193                                                                   <vscale x 8 x i1> %pg,
1194                                                                   <vscale x 8 x half> %b)
1195  ret <vscale x 8 x half> %out
1196}
1197
1198define <vscale x 4 x float> @frintx_s(<vscale x 4 x float> %a, <vscale x 4 x i1> %pg, <vscale x 4 x float> %b) {
1199; CHECK-LABEL: frintx_s:
1200; CHECK: frintx z0.s, p0/m, z1.s
1201; CHECK-NEXT: ret
1202  %out = call <vscale x 4 x float> @llvm.aarch64.sve.frintx.nxv4f32(<vscale x 4 x float> %a,
1203                                                                    <vscale x 4 x i1> %pg,
1204                                                                    <vscale x 4 x float> %b)
1205  ret <vscale x 4 x float> %out
1206}
1207
1208define <vscale x 2 x double> @frintx_d(<vscale x 2 x double> %a, <vscale x 2 x i1> %pg, <vscale x 2 x double> %b) {
1209; CHECK-LABEL: frintx_d:
1210; CHECK: frintx z0.d, p0/m, z1.d
1211; CHECK-NEXT: ret
1212  %out = call <vscale x 2 x double> @llvm.aarch64.sve.frintx.nxv2f64(<vscale x 2 x double> %a,
1213                                                                     <vscale x 2 x i1> %pg,
1214                                                                     <vscale x 2 x double> %b)
1215  ret <vscale x 2 x double> %out
1216}
1217
1218;
1219; FRINTZ
1220;
1221
1222define <vscale x 8 x half> @frintz_h(<vscale x 8 x half> %a, <vscale x 8 x i1> %pg, <vscale x 8 x half> %b) {
1223; CHECK-LABEL: frintz_h:
1224; CHECK: frintz z0.h, p0/m, z1.h
1225; CHECK-NEXT: ret
1226  %out = call <vscale x 8 x half> @llvm.aarch64.sve.frintz.nxv8f16(<vscale x 8 x half> %a,
1227                                                                   <vscale x 8 x i1> %pg,
1228                                                                   <vscale x 8 x half> %b)
1229  ret <vscale x 8 x half> %out
1230}
1231
1232define <vscale x 4 x float> @frintz_s(<vscale x 4 x float> %a, <vscale x 4 x i1> %pg, <vscale x 4 x float> %b) {
1233; CHECK-LABEL: frintz_s:
1234; CHECK: frintz z0.s, p0/m, z1.s
1235; CHECK-NEXT: ret
1236  %out = call <vscale x 4 x float> @llvm.aarch64.sve.frintz.nxv4f32(<vscale x 4 x float> %a,
1237                                                                    <vscale x 4 x i1> %pg,
1238                                                                    <vscale x 4 x float> %b)
1239  ret <vscale x 4 x float> %out
1240}
1241
1242define <vscale x 2 x double> @frintz_d(<vscale x 2 x double> %a, <vscale x 2 x i1> %pg, <vscale x 2 x double> %b) {
1243; CHECK-LABEL: frintz_d:
1244; CHECK: frintz z0.d, p0/m, z1.d
1245; CHECK-NEXT: ret
1246  %out = call <vscale x 2 x double> @llvm.aarch64.sve.frintz.nxv2f64(<vscale x 2 x double> %a,
1247                                                                     <vscale x 2 x i1> %pg,
1248                                                                     <vscale x 2 x double> %b)
1249  ret <vscale x 2 x double> %out
1250}
1251
1252;
1253; FRSQRTE
1254;
1255
1256define <vscale x 8 x half> @frsqrte_h(<vscale x 8 x half> %a) {
1257; CHECK-LABEL: frsqrte_h:
1258; CHECK: frsqrte z0.h, z0.h
1259; CHECK-NEXT: ret
1260  %out = call <vscale x 8 x half> @llvm.aarch64.sve.frsqrte.x.nxv8f16(<vscale x 8 x half> %a)
1261  ret <vscale x 8 x half> %out
1262}
1263
1264define <vscale x 4 x float> @frsqrte_s(<vscale x 4 x float> %a) {
1265; CHECK-LABEL: frsqrte_s:
1266; CHECK: frsqrte z0.s, z0.s
1267; CHECK-NEXT: ret
1268  %out = call <vscale x 4 x float> @llvm.aarch64.sve.frsqrte.x.nxv4f32(<vscale x 4 x float> %a)
1269  ret <vscale x 4 x float> %out
1270}
1271
1272define <vscale x 2 x double> @frsqrte_d(<vscale x 2 x i1> %pg, <vscale x 2 x double> %a) {
1273; CHECK-LABEL: frsqrte_d:
1274; CHECK: frsqrte z0.d, z0.d
1275; CHECK-NEXT: ret
1276  %out = call <vscale x 2 x double> @llvm.aarch64.sve.frsqrte.x.nxv2f64(<vscale x 2 x double> %a)
1277  ret <vscale x 2 x double> %out
1278}
1279
1280;
1281; FSCALE
1282;
1283
1284define <vscale x 8 x half> @fscale_h(<vscale x 8 x i1> %pg, <vscale x 8 x half> %a, <vscale x 8 x i16> %b) {
1285; CHECK-LABEL: fscale_h:
1286; CHECK: fscale z0.h, p0/m, z0.h, z1.h
1287; CHECK-NEXT: ret
1288  %out = call <vscale x 8 x half> @llvm.aarch64.sve.fscale.nxv8f16(<vscale x 8 x i1> %pg,
1289                                                                   <vscale x 8 x half> %a,
1290                                                                   <vscale x 8 x i16> %b)
1291  ret <vscale x 8 x half> %out
1292}
1293
1294define <vscale x 4 x float> @fscale_s(<vscale x 4 x i1> %pg, <vscale x 4 x float> %a, <vscale x 4 x i32> %b) {
1295; CHECK-LABEL: fscale_s:
1296; CHECK: fscale z0.s, p0/m, z0.s, z1.s
1297; CHECK-NEXT: ret
1298  %out = call <vscale x 4 x float> @llvm.aarch64.sve.fscale.nxv4f32(<vscale x 4 x i1> %pg,
1299                                                                    <vscale x 4 x float> %a,
1300                                                                    <vscale x 4 x i32> %b)
1301  ret <vscale x 4 x float> %out
1302}
1303
1304define <vscale x 2 x double> @fscale_d(<vscale x 2 x i1> %pg, <vscale x 2 x double> %a, <vscale x 2 x i64> %b) {
1305; CHECK-LABEL: fscale_d:
1306; CHECK: fscale z0.d, p0/m, z0.d, z1.d
1307; CHECK-NEXT: ret
1308  %out = call <vscale x 2 x double> @llvm.aarch64.sve.fscale.nxv2f64(<vscale x 2 x i1> %pg,
1309                                                                     <vscale x 2 x double> %a,
1310                                                                     <vscale x 2 x i64> %b)
1311  ret <vscale x 2 x double> %out
1312}
1313
1314;
1315; FSQRT
1316;
1317
1318define <vscale x 8 x half> @fsqrt_h(<vscale x 8 x half> %a, <vscale x 8 x i1> %pg, <vscale x 8 x half> %b) {
1319; CHECK-LABEL: fsqrt_h:
1320; CHECK: fsqrt z0.h, p0/m, z1.h
1321; CHECK-NEXT: ret
1322  %out = call <vscale x 8 x half> @llvm.aarch64.sve.fsqrt.nxv8f16(<vscale x 8 x half> %a,
1323                                                                  <vscale x 8 x i1> %pg,
1324                                                                  <vscale x 8 x half> %b)
1325  ret <vscale x 8 x half> %out
1326}
1327
1328define <vscale x 4 x float> @fsqrt_s(<vscale x 4 x float> %a, <vscale x 4 x i1> %pg, <vscale x 4 x float> %b) {
1329; CHECK-LABEL: fsqrt_s:
1330; CHECK: fsqrt z0.s, p0/m, z1.s
1331; CHECK-NEXT: ret
1332  %out = call <vscale x 4 x float> @llvm.aarch64.sve.fsqrt.nxv4f32(<vscale x 4 x float> %a,
1333                                                                   <vscale x 4 x i1> %pg,
1334                                                                   <vscale x 4 x float> %b)
1335  ret <vscale x 4 x float> %out
1336}
1337
1338define <vscale x 2 x double> @fsqrt_d(<vscale x 2 x double> %a, <vscale x 2 x i1> %pg, <vscale x 2 x double> %b) {
1339; CHECK-LABEL: fsqrt_d:
1340; CHECK: fsqrt z0.d, p0/m, z1.d
1341; CHECK-NEXT: ret
1342  %out = call <vscale x 2 x double> @llvm.aarch64.sve.fsqrt.nxv2f64(<vscale x 2 x double> %a,
1343                                                                    <vscale x 2 x i1> %pg,
1344                                                                    <vscale x 2 x double> %b)
1345  ret <vscale x 2 x double> %out
1346}
1347
1348;
1349; FSUB
1350;
1351
1352define <vscale x 8 x half> @fsub_h(<vscale x 8 x i1> %pg, <vscale x 8 x half> %a, <vscale x 8 x half> %b) {
1353; CHECK-LABEL: fsub_h:
1354; CHECK: fsub z0.h, p0/m, z0.h, z1.h
1355; CHECK-NEXT: ret
1356  %out = call <vscale x 8 x half> @llvm.aarch64.sve.fsub.nxv8f16(<vscale x 8 x i1> %pg,
1357                                                                 <vscale x 8 x half> %a,
1358                                                                 <vscale x 8 x half> %b)
1359  ret <vscale x 8 x half> %out
1360}
1361
1362define <vscale x 4 x float> @fsub_s(<vscale x 4 x i1> %pg, <vscale x 4 x float> %a, <vscale x 4 x float> %b) {
1363; CHECK-LABEL: fsub_s:
1364; CHECK: fsub z0.s, p0/m, z0.s, z1.s
1365; CHECK-NEXT: ret
1366  %out = call <vscale x 4 x float> @llvm.aarch64.sve.fsub.nxv4f32(<vscale x 4 x i1> %pg,
1367                                                                  <vscale x 4 x float> %a,
1368                                                                  <vscale x 4 x float> %b)
1369  ret <vscale x 4 x float> %out
1370}
1371
1372define <vscale x 2 x double> @fsub_d(<vscale x 2 x i1> %pg, <vscale x 2 x double> %a, <vscale x 2 x double> %b) {
1373; CHECK-LABEL: fsub_d:
1374; CHECK: fsub z0.d, p0/m, z0.d, z1.d
1375; CHECK-NEXT: ret
1376  %out = call <vscale x 2 x double> @llvm.aarch64.sve.fsub.nxv2f64(<vscale x 2 x i1> %pg,
1377                                                                   <vscale x 2 x double> %a,
1378                                                                   <vscale x 2 x double> %b)
1379  ret <vscale x 2 x double> %out
1380}
1381
1382;
1383; FSUBR
1384;
1385
1386define <vscale x 8 x half> @fsubr_h(<vscale x 8 x i1> %pg, <vscale x 8 x half> %a, <vscale x 8 x half> %b) {
1387; CHECK-LABEL: fsubr_h:
1388; CHECK: fsubr z0.h, p0/m, z0.h, z1.h
1389; CHECK-NEXT: ret
1390  %out = call <vscale x 8 x half> @llvm.aarch64.sve.fsubr.nxv8f16(<vscale x 8 x i1> %pg,
1391                                                                  <vscale x 8 x half> %a,
1392                                                                  <vscale x 8 x half> %b)
1393  ret <vscale x 8 x half> %out
1394}
1395
1396define <vscale x 4 x float> @fsubr_s(<vscale x 4 x i1> %pg, <vscale x 4 x float> %a, <vscale x 4 x float> %b) {
1397; CHECK-LABEL: fsubr_s:
1398; CHECK: fsubr z0.s, p0/m, z0.s, z1.s
1399; CHECK-NEXT: ret
1400  %out = call <vscale x 4 x float> @llvm.aarch64.sve.fsubr.nxv4f32(<vscale x 4 x i1> %pg,
1401                                                                   <vscale x 4 x float> %a,
1402                                                                   <vscale x 4 x float> %b)
1403  ret <vscale x 4 x float> %out
1404}
1405
1406define <vscale x 2 x double> @fsubr_d(<vscale x 2 x i1> %pg, <vscale x 2 x double> %a, <vscale x 2 x double> %b) {
1407; CHECK-LABEL: fsubr_d:
1408; CHECK: fsubr z0.d, p0/m, z0.d, z1.d
1409; CHECK-NEXT: ret
1410  %out = call <vscale x 2 x double> @llvm.aarch64.sve.fsubr.nxv2f64(<vscale x 2 x i1> %pg,
1411                                                                    <vscale x 2 x double> %a,
1412                                                                    <vscale x 2 x double> %b)
1413  ret <vscale x 2 x double> %out
1414}
1415
1416;
1417; FTMAD
1418;
1419
1420define <vscale x 8 x half> @ftmad_h(<vscale x 8 x half> %a, <vscale x 8 x half> %b) {
1421; CHECK-LABEL: ftmad_h:
1422; CHECK: ftmad z0.h, z0.h, z1.h, #0
1423; CHECK-NEXT: ret
1424  %out = call <vscale x 8 x half> @llvm.aarch64.sve.ftmad.x.nxv8f16(<vscale x 8 x half> %a,
1425                                                                    <vscale x 8 x half> %b,
1426                                                                    i32 0)
1427  ret <vscale x 8 x half> %out
1428}
1429
1430define <vscale x 4 x float> @ftmad_s(<vscale x 4 x float> %a, <vscale x 4 x float> %b) {
1431; CHECK-LABEL: ftmad_s:
1432; CHECK: ftmad z0.s, z0.s, z1.s, #0
1433; CHECK-NEXT: ret
1434  %out = call <vscale x 4 x float> @llvm.aarch64.sve.ftmad.x.nxv4f32(<vscale x 4 x float> %a,
1435                                                                     <vscale x 4 x float> %b,
1436                                                                     i32 0)
1437  ret <vscale x 4 x float> %out
1438}
1439
1440define <vscale x 2 x double> @ftmad_d(<vscale x 2 x double> %a, <vscale x 2 x double> %b) {
1441; CHECK-LABEL: ftmad_d:
1442; CHECK: ftmad z0.d, z0.d, z1.d, #7
1443; CHECK-NEXT: ret
1444  %out = call <vscale x 2 x double> @llvm.aarch64.sve.ftmad.x.nxv2f64(<vscale x 2 x double> %a,
1445                                                                      <vscale x 2 x double> %b,
1446                                                                      i32 7)
1447  ret <vscale x 2 x double> %out
1448}
1449
1450;
1451; FTSMUL
1452;
1453
1454define <vscale x 8 x half> @ftsmul_h(<vscale x 8 x half> %a, <vscale x 8 x i16> %b) {
1455; CHECK-LABEL: ftsmul_h:
1456; CHECK: ftsmul z0.h, z0.h, z1.h
1457; CHECK-NEXT: ret
1458  %out = call <vscale x 8 x half> @llvm.aarch64.sve.ftsmul.x.nxv8f16(<vscale x 8 x half> %a,
1459                                                                     <vscale x 8 x i16> %b)
1460  ret <vscale x 8 x half> %out
1461}
1462
1463define <vscale x 4 x float> @ftsmul_s(<vscale x 4 x float> %a, <vscale x 4 x i32> %b) {
1464; CHECK-LABEL: ftsmul_s:
1465; CHECK: ftsmul z0.s, z0.s, z1.s
1466; CHECK-NEXT: ret
1467  %out = call <vscale x 4 x float> @llvm.aarch64.sve.ftsmul.x.nxv4f32(<vscale x 4 x float> %a,
1468                                                                      <vscale x 4 x i32> %b)
1469  ret <vscale x 4 x float> %out
1470}
1471
1472define <vscale x 2 x double> @ftsmul_d(<vscale x 2 x double> %a, <vscale x 2 x i64> %b) {
1473; CHECK-LABEL: ftsmul_d:
1474; CHECK: ftsmul z0.d, z0.d, z1.d
1475; CHECK-NEXT: ret
1476  %out = call <vscale x 2 x double> @llvm.aarch64.sve.ftsmul.x.nxv2f64(<vscale x 2 x double> %a,
1477                                                                       <vscale x 2 x i64> %b)
1478  ret <vscale x 2 x double> %out
1479}
1480
1481;
1482; FTSSEL
1483;
1484
1485define <vscale x 8 x half> @ftssel_h(<vscale x 8 x half> %a, <vscale x 8 x i16> %b) {
1486; CHECK-LABEL: ftssel_h:
1487; CHECK: ftssel z0.h, z0.h, z1.h
1488; CHECK-NEXT: ret
1489  %out = call <vscale x 8 x half> @llvm.aarch64.sve.ftssel.x.nxv8f16(<vscale x 8 x half> %a,
1490                                                                     <vscale x 8 x i16> %b)
1491  ret <vscale x 8 x half> %out
1492}
1493
1494define <vscale x 4 x float> @ftssel_s(<vscale x 4 x float> %a, <vscale x 4 x i32> %b) {
1495; CHECK-LABEL: ftssel_s:
1496; CHECK: ftssel z0.s, z0.s, z1.s
1497; CHECK-NEXT: ret
1498  %out = call <vscale x 4 x float> @llvm.aarch64.sve.ftssel.x.nxv4f32(<vscale x 4 x float> %a,
1499                                                                      <vscale x 4 x i32> %b)
1500  ret <vscale x 4 x float> %out
1501}
1502
1503define <vscale x 2 x double> @ftssel_d(<vscale x 2 x double> %a, <vscale x 2 x i64> %b) {
1504; CHECK-LABEL: ftssel_d:
1505; CHECK: ftssel z0.d, z0.d, z1.d
1506; CHECK-NEXT: ret
1507  %out = call <vscale x 2 x double> @llvm.aarch64.sve.ftssel.x.nxv2f64(<vscale x 2 x double> %a,
1508                                                                       <vscale x 2 x i64> %b)
1509  ret <vscale x 2 x double> %out
1510}
1511
1512declare <vscale x 8 x half> @llvm.aarch64.sve.fabd.nxv8f16(<vscale x 8 x i1>, <vscale x 8 x half>, <vscale x 8 x half>)
1513declare <vscale x 4 x float> @llvm.aarch64.sve.fabd.nxv4f32(<vscale x 4 x i1>, <vscale x 4 x float>, <vscale x 4 x float>)
1514declare <vscale x 2 x double> @llvm.aarch64.sve.fabd.nxv2f64(<vscale x 2 x i1>, <vscale x 2 x double>, <vscale x 2 x double>)
1515
1516declare <vscale x 8 x half> @llvm.aarch64.sve.fabs.nxv8f16(<vscale x 8 x half>, <vscale x 8 x i1>, <vscale x 8 x half>)
1517declare <vscale x 4 x float> @llvm.aarch64.sve.fabs.nxv4f32(<vscale x 4 x float>, <vscale x 4 x i1>, <vscale x 4 x float>)
1518declare <vscale x 2 x double> @llvm.aarch64.sve.fabs.nxv2f64(<vscale x 2 x double>, <vscale x 2 x i1>, <vscale x 2 x double>)
1519
1520declare <vscale x 8 x half> @llvm.aarch64.sve.fadd.nxv8f16(<vscale x 8 x i1>, <vscale x 8 x half>, <vscale x 8 x half>)
1521declare <vscale x 4 x float> @llvm.aarch64.sve.fadd.nxv4f32(<vscale x 4 x i1>, <vscale x 4 x float>, <vscale x 4 x float>)
1522declare <vscale x 2 x double> @llvm.aarch64.sve.fadd.nxv2f64(<vscale x 2 x i1>, <vscale x 2 x double>, <vscale x 2 x double>)
1523
1524declare <vscale x 8 x half> @llvm.aarch64.sve.fcadd.nxv8f16(<vscale x 8 x i1>, <vscale x 8 x half>, <vscale x 8 x half>, i32)
1525declare <vscale x 4 x float> @llvm.aarch64.sve.fcadd.nxv4f32(<vscale x 4 x i1>, <vscale x 4 x float>, <vscale x 4 x float>, i32)
1526declare <vscale x 2 x double> @llvm.aarch64.sve.fcadd.nxv2f64(<vscale x 2 x i1>, <vscale x 2 x double>, <vscale x 2 x double>, i32)
1527
1528declare <vscale x 8 x half> @llvm.aarch64.sve.fcmla.nxv8f16(<vscale x 8 x i1>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, i32)
1529declare <vscale x 4 x float> @llvm.aarch64.sve.fcmla.nxv4f32(<vscale x 4 x i1>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, i32)
1530declare <vscale x 2 x double> @llvm.aarch64.sve.fcmla.nxv2f64(<vscale x 2 x i1>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, i32)
1531
1532declare <vscale x 8 x half> @llvm.aarch64.sve.fcmla.lane.nxv8f16(<vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, i32, i32)
1533declare <vscale x 4 x float> @llvm.aarch64.sve.fcmla.lane.nxv4f32(<vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, i32, i32)
1534
1535declare <vscale x 8 x half> @llvm.aarch64.sve.fdiv.nxv8f16(<vscale x 8 x i1>, <vscale x 8 x half>, <vscale x 8 x half>)
1536declare <vscale x 4 x float> @llvm.aarch64.sve.fdiv.nxv4f32(<vscale x 4 x i1>, <vscale x 4 x float>, <vscale x 4 x float>)
1537declare <vscale x 2 x double> @llvm.aarch64.sve.fdiv.nxv2f64(<vscale x 2 x i1>, <vscale x 2 x double>, <vscale x 2 x double>)
1538
1539declare <vscale x 8 x half> @llvm.aarch64.sve.fdivr.nxv8f16(<vscale x 8 x i1>, <vscale x 8 x half>, <vscale x 8 x half>)
1540declare <vscale x 4 x float> @llvm.aarch64.sve.fdivr.nxv4f32(<vscale x 4 x i1>, <vscale x 4 x float>, <vscale x 4 x float>)
1541declare <vscale x 2 x double> @llvm.aarch64.sve.fdivr.nxv2f64(<vscale x 2 x i1>, <vscale x 2 x double>, <vscale x 2 x double>)
1542
1543declare <vscale x 8 x half> @llvm.aarch64.sve.fexpa.x.nxv8f16(<vscale x 8 x i16>)
1544declare <vscale x 4 x float> @llvm.aarch64.sve.fexpa.x.nxv4f32(<vscale x 4 x i32>)
1545declare <vscale x 2 x double> @llvm.aarch64.sve.fexpa.x.nxv2f64(<vscale x 2 x i64>)
1546
1547declare <vscale x 8 x half> @llvm.aarch64.sve.fmad.nxv8f16(<vscale x 8 x i1>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>)
1548declare <vscale x 4 x float> @llvm.aarch64.sve.fmad.nxv4f32(<vscale x 4 x i1>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>)
1549declare <vscale x 2 x double> @llvm.aarch64.sve.fmad.nxv2f64(<vscale x 2 x i1>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>)
1550
1551declare <vscale x 8 x half> @llvm.aarch64.sve.fmax.nxv8f16(<vscale x 8 x i1>, <vscale x 8 x half>, <vscale x 8 x half>)
1552declare <vscale x 4 x float> @llvm.aarch64.sve.fmax.nxv4f32(<vscale x 4 x i1>, <vscale x 4 x float>, <vscale x 4 x float>)
1553declare <vscale x 2 x double> @llvm.aarch64.sve.fmax.nxv2f64(<vscale x 2 x i1>, <vscale x 2 x double>, <vscale x 2 x double>)
1554
1555declare <vscale x 8 x half> @llvm.aarch64.sve.fmaxnm.nxv8f16(<vscale x 8 x i1>, <vscale x 8 x half>, <vscale x 8 x half>)
1556declare <vscale x 4 x float> @llvm.aarch64.sve.fmaxnm.nxv4f32(<vscale x 4 x i1>, <vscale x 4 x float>, <vscale x 4 x float>)
1557declare <vscale x 2 x double> @llvm.aarch64.sve.fmaxnm.nxv2f64(<vscale x 2 x i1>, <vscale x 2 x double>, <vscale x 2 x double>)
1558
1559declare <vscale x 8 x half> @llvm.aarch64.sve.fmin.nxv8f16(<vscale x 8 x i1>, <vscale x 8 x half>, <vscale x 8 x half>)
1560declare <vscale x 4 x float> @llvm.aarch64.sve.fmin.nxv4f32(<vscale x 4 x i1>, <vscale x 4 x float>, <vscale x 4 x float>)
1561declare <vscale x 2 x double> @llvm.aarch64.sve.fmin.nxv2f64(<vscale x 2 x i1>, <vscale x 2 x double>, <vscale x 2 x double>)
1562
1563declare <vscale x 8 x half> @llvm.aarch64.sve.fminnm.nxv8f16(<vscale x 8 x i1>, <vscale x 8 x half>, <vscale x 8 x half>)
1564declare <vscale x 4 x float> @llvm.aarch64.sve.fminnm.nxv4f32(<vscale x 4 x i1>, <vscale x 4 x float>, <vscale x 4 x float>)
1565declare <vscale x 2 x double> @llvm.aarch64.sve.fminnm.nxv2f64(<vscale x 2 x i1>, <vscale x 2 x double>, <vscale x 2 x double>)
1566
1567declare <vscale x 8 x half> @llvm.aarch64.sve.fmla.nxv8f16(<vscale x 8 x i1>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>)
1568declare <vscale x 4 x float> @llvm.aarch64.sve.fmla.nxv4f32(<vscale x 4 x i1>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>)
1569declare <vscale x 2 x double> @llvm.aarch64.sve.fmla.nxv2f64(<vscale x 2 x i1>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>)
1570
1571declare <vscale x 8 x half> @llvm.aarch64.sve.fmla.lane.nxv8f16(<vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, i32)
1572declare <vscale x 4 x float> @llvm.aarch64.sve.fmla.lane.nxv4f32(<vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, i32)
1573declare <vscale x 2 x double> @llvm.aarch64.sve.fmla.lane.nxv2f64(<vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, i32)
1574
1575declare <vscale x 8 x half> @llvm.aarch64.sve.fmls.nxv8f16(<vscale x 8 x i1>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>)
1576declare <vscale x 4 x float> @llvm.aarch64.sve.fmls.nxv4f32(<vscale x 4 x i1>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>)
1577declare <vscale x 2 x double> @llvm.aarch64.sve.fmls.nxv2f64(<vscale x 2 x i1>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>)
1578
1579declare <vscale x 8 x half> @llvm.aarch64.sve.fmls.lane.nxv8f16(<vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, i32)
1580declare <vscale x 4 x float> @llvm.aarch64.sve.fmls.lane.nxv4f32(<vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, i32)
1581declare <vscale x 2 x double> @llvm.aarch64.sve.fmls.lane.nxv2f64(<vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, i32)
1582
1583declare <vscale x 8 x half> @llvm.aarch64.sve.fmsb.nxv8f16(<vscale x 8 x i1>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>)
1584declare <vscale x 4 x float> @llvm.aarch64.sve.fmsb.nxv4f32(<vscale x 4 x i1>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>)
1585declare <vscale x 2 x double> @llvm.aarch64.sve.fmsb.nxv2f64(<vscale x 2 x i1>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>)
1586
1587declare <vscale x 8 x half> @llvm.aarch64.sve.fmul.nxv8f16(<vscale x 8 x i1>, <vscale x 8 x half>, <vscale x 8 x half>)
1588declare <vscale x 4 x float> @llvm.aarch64.sve.fmul.nxv4f32(<vscale x 4 x i1>, <vscale x 4 x float>, <vscale x 4 x float>)
1589declare <vscale x 2 x double> @llvm.aarch64.sve.fmul.nxv2f64(<vscale x 2 x i1>, <vscale x 2 x double>, <vscale x 2 x double>)
1590
1591declare <vscale x 8 x half> @llvm.aarch64.sve.fmul.lane.nxv8f16(<vscale x 8 x half>, <vscale x 8 x half>, i32)
1592declare <vscale x 4 x float> @llvm.aarch64.sve.fmul.lane.nxv4f32(<vscale x 4 x float>, <vscale x 4 x float>, i32)
1593declare <vscale x 2 x double> @llvm.aarch64.sve.fmul.lane.nxv2f64(<vscale x 2 x double>, <vscale x 2 x double>, i32)
1594
1595declare <vscale x 8 x half> @llvm.aarch64.sve.fmulx.nxv8f16(<vscale x 8 x i1>, <vscale x 8 x half>, <vscale x 8 x half>)
1596declare <vscale x 4 x float> @llvm.aarch64.sve.fmulx.nxv4f32(<vscale x 4 x i1>, <vscale x 4 x float>, <vscale x 4 x float>)
1597declare <vscale x 2 x double> @llvm.aarch64.sve.fmulx.nxv2f64(<vscale x 2 x i1>, <vscale x 2 x double>, <vscale x 2 x double>)
1598
1599declare <vscale x 8 x half> @llvm.aarch64.sve.fneg.nxv8f16(<vscale x 8 x half>, <vscale x 8 x i1>, <vscale x 8 x half>)
1600declare <vscale x 4 x float> @llvm.aarch64.sve.fneg.nxv4f32(<vscale x 4 x float>, <vscale x 4 x i1>, <vscale x 4 x float>)
1601declare <vscale x 2 x double> @llvm.aarch64.sve.fneg.nxv2f64(<vscale x 2 x double>, <vscale x 2 x i1>, <vscale x 2 x double>)
1602
1603declare <vscale x 8 x half> @llvm.aarch64.sve.fnmad.nxv8f16(<vscale x 8 x i1>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>)
1604declare <vscale x 4 x float> @llvm.aarch64.sve.fnmad.nxv4f32(<vscale x 4 x i1>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>)
1605declare <vscale x 2 x double> @llvm.aarch64.sve.fnmad.nxv2f64(<vscale x 2 x i1>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>)
1606
1607declare <vscale x 8 x half> @llvm.aarch64.sve.fnmla.nxv8f16(<vscale x 8 x i1>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>)
1608declare <vscale x 4 x float> @llvm.aarch64.sve.fnmla.nxv4f32(<vscale x 4 x i1>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>)
1609declare <vscale x 2 x double> @llvm.aarch64.sve.fnmla.nxv2f64(<vscale x 2 x i1>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>)
1610
1611declare <vscale x 8 x half> @llvm.aarch64.sve.fnmls.nxv8f16(<vscale x 8 x i1>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>)
1612declare <vscale x 4 x float> @llvm.aarch64.sve.fnmls.nxv4f32(<vscale x 4 x i1>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>)
1613declare <vscale x 2 x double> @llvm.aarch64.sve.fnmls.nxv2f64(<vscale x 2 x i1>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>)
1614
1615declare <vscale x 8 x half> @llvm.aarch64.sve.fnmsb.nxv8f16(<vscale x 8 x i1>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>)
1616declare <vscale x 4 x float> @llvm.aarch64.sve.fnmsb.nxv4f32(<vscale x 4 x i1>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>)
1617declare <vscale x 2 x double> @llvm.aarch64.sve.fnmsb.nxv2f64(<vscale x 2 x i1>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>)
1618
1619declare <vscale x 8 x half> @llvm.aarch64.sve.frecpe.x.nxv8f16(<vscale x 8 x half>)
1620declare <vscale x 4 x float> @llvm.aarch64.sve.frecpe.x.nxv4f32(<vscale x 4 x float>)
1621declare <vscale x 2 x double> @llvm.aarch64.sve.frecpe.x.nxv2f64(<vscale x 2 x double>)
1622
1623declare <vscale x 8 x half> @llvm.aarch64.sve.frecpx.nxv8f16(<vscale x 8 x half>, <vscale x 8 x i1>, <vscale x 8 x half>)
1624declare <vscale x 4 x float> @llvm.aarch64.sve.frecpx.nxv4f32(<vscale x 4 x float>, <vscale x 4 x i1>, <vscale x 4 x float>)
1625declare <vscale x 2 x double> @llvm.aarch64.sve.frecpx.nxv2f64(<vscale x 2 x double>, <vscale x 2 x i1>, <vscale x 2 x double>)
1626
1627declare <vscale x 8 x half> @llvm.aarch64.sve.frinta.nxv8f16(<vscale x 8 x half>, <vscale x 8 x i1>, <vscale x 8 x half>)
1628declare <vscale x 4 x float> @llvm.aarch64.sve.frinta.nxv4f32(<vscale x 4 x float>, <vscale x 4 x i1>, <vscale x 4 x float>)
1629declare <vscale x 2 x double> @llvm.aarch64.sve.frinta.nxv2f64(<vscale x 2 x double>, <vscale x 2 x i1>, <vscale x 2 x double>)
1630
1631declare <vscale x 8 x half> @llvm.aarch64.sve.frinti.nxv8f16(<vscale x 8 x half>, <vscale x 8 x i1>, <vscale x 8 x half>)
1632declare <vscale x 4 x float> @llvm.aarch64.sve.frinti.nxv4f32(<vscale x 4 x float>, <vscale x 4 x i1>, <vscale x 4 x float>)
1633declare <vscale x 2 x double> @llvm.aarch64.sve.frinti.nxv2f64(<vscale x 2 x double>, <vscale x 2 x i1>, <vscale x 2 x double>)
1634
1635declare <vscale x 8 x half> @llvm.aarch64.sve.frintm.nxv8f16(<vscale x 8 x half>, <vscale x 8 x i1>, <vscale x 8 x half>)
1636declare <vscale x 4 x float> @llvm.aarch64.sve.frintm.nxv4f32(<vscale x 4 x float>, <vscale x 4 x i1>, <vscale x 4 x float>)
1637declare <vscale x 2 x double> @llvm.aarch64.sve.frintm.nxv2f64(<vscale x 2 x double>, <vscale x 2 x i1>, <vscale x 2 x double>)
1638
1639declare <vscale x 8 x half> @llvm.aarch64.sve.frintn.nxv8f16(<vscale x 8 x half>, <vscale x 8 x i1>, <vscale x 8 x half>)
1640declare <vscale x 4 x float> @llvm.aarch64.sve.frintn.nxv4f32(<vscale x 4 x float>, <vscale x 4 x i1>, <vscale x 4 x float>)
1641declare <vscale x 2 x double> @llvm.aarch64.sve.frintn.nxv2f64(<vscale x 2 x double>, <vscale x 2 x i1>, <vscale x 2 x double>)
1642
1643declare <vscale x 8 x half> @llvm.aarch64.sve.frintp.nxv8f16(<vscale x 8 x half>, <vscale x 8 x i1>, <vscale x 8 x half>)
1644declare <vscale x 4 x float> @llvm.aarch64.sve.frintp.nxv4f32(<vscale x 4 x float>, <vscale x 4 x i1>, <vscale x 4 x float>)
1645declare <vscale x 2 x double> @llvm.aarch64.sve.frintp.nxv2f64(<vscale x 2 x double>, <vscale x 2 x i1>, <vscale x 2 x double>)
1646
1647declare <vscale x 8 x half> @llvm.aarch64.sve.frintx.nxv8f16(<vscale x 8 x half>, <vscale x 8 x i1>, <vscale x 8 x half>)
1648declare <vscale x 4 x float> @llvm.aarch64.sve.frintx.nxv4f32(<vscale x 4 x float>, <vscale x 4 x i1>, <vscale x 4 x float>)
1649declare <vscale x 2 x double> @llvm.aarch64.sve.frintx.nxv2f64(<vscale x 2 x double>, <vscale x 2 x i1>, <vscale x 2 x double>)
1650
1651declare <vscale x 8 x half> @llvm.aarch64.sve.frintz.nxv8f16(<vscale x 8 x half>, <vscale x 8 x i1>, <vscale x 8 x half>)
1652declare <vscale x 4 x float> @llvm.aarch64.sve.frintz.nxv4f32(<vscale x 4 x float>, <vscale x 4 x i1>, <vscale x 4 x float>)
1653declare <vscale x 2 x double> @llvm.aarch64.sve.frintz.nxv2f64(<vscale x 2 x double>, <vscale x 2 x i1>, <vscale x 2 x double>)
1654
1655declare <vscale x 8 x half> @llvm.aarch64.sve.frsqrte.x.nxv8f16(<vscale x 8 x half>)
1656declare <vscale x 4 x float> @llvm.aarch64.sve.frsqrte.x.nxv4f32(<vscale x 4 x float>)
1657declare <vscale x 2 x double> @llvm.aarch64.sve.frsqrte.x.nxv2f64(<vscale x 2 x double>)
1658
1659declare <vscale x 8 x half> @llvm.aarch64.sve.fscale.nxv8f16(<vscale x 8 x i1>, <vscale x 8 x half>, <vscale x 8 x i16>)
1660declare <vscale x 4 x float> @llvm.aarch64.sve.fscale.nxv4f32(<vscale x 4 x i1>, <vscale x 4 x float>, <vscale x 4 x i32>)
1661declare <vscale x 2 x double> @llvm.aarch64.sve.fscale.nxv2f64(<vscale x 2 x i1>, <vscale x 2 x double>, <vscale x 2 x i64>)
1662
1663declare <vscale x 8 x half> @llvm.aarch64.sve.fsqrt.nxv8f16(<vscale x 8 x half>, <vscale x 8 x i1>, <vscale x 8 x half>)
1664declare <vscale x 4 x float> @llvm.aarch64.sve.fsqrt.nxv4f32(<vscale x 4 x float>, <vscale x 4 x i1>, <vscale x 4 x float>)
1665declare <vscale x 2 x double> @llvm.aarch64.sve.fsqrt.nxv2f64(<vscale x 2 x double>, <vscale x 2 x i1>, <vscale x 2 x double>)
1666
1667declare <vscale x 8 x half> @llvm.aarch64.sve.fsub.nxv8f16(<vscale x 8 x i1>, <vscale x 8 x half>, <vscale x 8 x half>)
1668declare <vscale x 4 x float> @llvm.aarch64.sve.fsub.nxv4f32(<vscale x 4 x i1>, <vscale x 4 x float>, <vscale x 4 x float>)
1669declare <vscale x 2 x double> @llvm.aarch64.sve.fsub.nxv2f64(<vscale x 2 x i1>, <vscale x 2 x double>, <vscale x 2 x double>)
1670
1671declare <vscale x 8 x half> @llvm.aarch64.sve.fsubr.nxv8f16(<vscale x 8 x i1>, <vscale x 8 x half>, <vscale x 8 x half>)
1672declare <vscale x 4 x float> @llvm.aarch64.sve.fsubr.nxv4f32(<vscale x 4 x i1>, <vscale x 4 x float>, <vscale x 4 x float>)
1673declare <vscale x 2 x double> @llvm.aarch64.sve.fsubr.nxv2f64(<vscale x 2 x i1>, <vscale x 2 x double>, <vscale x 2 x double>)
1674
1675declare <vscale x 8 x half> @llvm.aarch64.sve.ftmad.x.nxv8f16(<vscale x 8 x half>, <vscale x 8 x half>, i32)
1676declare <vscale x 4 x float> @llvm.aarch64.sve.ftmad.x.nxv4f32(<vscale x 4 x float>, <vscale x 4 x float>, i32)
1677declare <vscale x 2 x double> @llvm.aarch64.sve.ftmad.x.nxv2f64(<vscale x 2 x double>, <vscale x 2 x double>, i32)
1678
1679declare <vscale x 8 x half> @llvm.aarch64.sve.ftsmul.x.nxv8f16(<vscale x 8 x half>, <vscale x 8 x i16>)
1680declare <vscale x 4 x float> @llvm.aarch64.sve.ftsmul.x.nxv4f32(<vscale x 4 x float>, <vscale x 4 x i32>)
1681declare <vscale x 2 x double> @llvm.aarch64.sve.ftsmul.x.nxv2f64(<vscale x 2 x double>, <vscale x 2 x i64>)
1682
1683declare <vscale x 8 x half> @llvm.aarch64.sve.ftssel.x.nxv8f16(<vscale x 8 x half>, <vscale x 8 x i16>)
1684declare <vscale x 4 x float> @llvm.aarch64.sve.ftssel.x.nxv4f32(<vscale x 4 x float>, <vscale x 4 x i32>)
1685declare <vscale x 2 x double> @llvm.aarch64.sve.ftssel.x.nxv2f64(<vscale x 2 x double>, <vscale x 2 x i64>)
1686