xref: /llvm-project/llvm/test/CodeGen/AArch64/sve-intrinsics-gather-loads-vector-base-imm-offset.ll (revision fadea4413ecbfffa4d28ad8298e0628165b543f1)
1; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
2; RUN: llc -mtriple=aarch64-linux-gnu -mattr=+sve < %s | FileCheck %s
3
4;
5; LD1B, LD1W, LD1H, LD1D: vector base + immediate offset (index)
6;   e.g. ld1h { z0.s }, p0/z, [z0.s, #16]
7;
8
9; LD1B
10define <vscale x 4 x i32> @gld1b_s_imm_offset(<vscale x 4 x i1> %pg, <vscale x 4 x i32> %base) {
11; CHECK-LABEL: gld1b_s_imm_offset:
12; CHECK:       // %bb.0:
13; CHECK-NEXT:    ld1b { z0.s }, p0/z, [z0.s, #16]
14; CHECK-NEXT:    ret
15  %load = call <vscale x 4 x i8> @llvm.aarch64.sve.ld1.gather.scalar.offset.nxv4i8.nxv4i32(<vscale x 4 x i1> %pg,
16                                                                                           <vscale x 4 x i32> %base,
17                                                                                           i64 16)
18  %res = zext <vscale x 4 x i8> %load to <vscale x 4 x i32>
19  ret <vscale x 4 x i32> %res
20}
21
22define <vscale x 2 x i64> @gld1b_d_imm_offset(<vscale x 2 x i1> %pg, <vscale x 2 x i64> %base) {
23; CHECK-LABEL: gld1b_d_imm_offset:
24; CHECK:       // %bb.0:
25; CHECK-NEXT:    ld1b { z0.d }, p0/z, [z0.d, #16]
26; CHECK-NEXT:    ret
27  %load = call <vscale x 2 x i8> @llvm.aarch64.sve.ld1.gather.scalar.offset.nxv2i8.nxv2i64(<vscale x 2 x i1> %pg,
28                                                                                           <vscale x 2 x i64> %base,
29                                                                                           i64 16)
30  %res = zext <vscale x 2 x i8> %load to <vscale x 2 x i64>
31  ret <vscale x 2 x i64> %res
32}
33
34; LD1H
35define <vscale x 4 x i32> @gld1h_s_imm_offset(<vscale x 4 x i1> %pg, <vscale x 4 x i32> %base) {
36; CHECK-LABEL: gld1h_s_imm_offset:
37; CHECK:       // %bb.0:
38; CHECK-NEXT:    ld1h { z0.s }, p0/z, [z0.s, #16]
39; CHECK-NEXT:    ret
40  %load = call <vscale x 4 x i16> @llvm.aarch64.sve.ld1.gather.scalar.offset.nxv4i16.nxv4i32(<vscale x 4 x i1> %pg,
41                                                                                             <vscale x 4 x i32> %base,
42                                                                                             i64 16)
43  %res = zext <vscale x 4 x i16> %load to <vscale x 4 x i32>
44  ret <vscale x 4 x i32> %res
45}
46
47define <vscale x 2 x i64> @gld1h_d_imm_offset(<vscale x 2 x i1> %pg, <vscale x 2 x i64> %base) {
48; CHECK-LABEL: gld1h_d_imm_offset:
49; CHECK:       // %bb.0:
50; CHECK-NEXT:    ld1h { z0.d }, p0/z, [z0.d, #16]
51; CHECK-NEXT:    ret
52  %load = call <vscale x 2 x i16> @llvm.aarch64.sve.ld1.gather.scalar.offset.nxv2i16.nxv2i64(<vscale x 2 x i1> %pg,
53                                                                                             <vscale x 2 x i64> %base,
54                                                                                             i64 16)
55  %res = zext <vscale x 2 x i16> %load to <vscale x 2 x i64>
56  ret <vscale x 2 x i64> %res
57}
58
59; LD1W
60define <vscale x 4 x i32> @gld1w_s_imm_offset(<vscale x 4 x i1> %pg, <vscale x 4 x i32> %base) {
61; CHECK-LABEL: gld1w_s_imm_offset:
62; CHECK:       // %bb.0:
63; CHECK-NEXT:    ld1w { z0.s }, p0/z, [z0.s, #16]
64; CHECK-NEXT:    ret
65  %load = call <vscale x 4 x i32> @llvm.aarch64.sve.ld1.gather.scalar.offset.nxv4i32.nxv4i32(<vscale x 4 x i1> %pg,
66                                                                                             <vscale x 4 x i32> %base,
67                                                                                             i64 16)
68  ret <vscale x 4 x i32> %load
69}
70
71define <vscale x 2 x i64> @gld1w_d_imm_offset(<vscale x 2 x i1> %pg, <vscale x 2 x i64> %base) {
72; CHECK-LABEL: gld1w_d_imm_offset:
73; CHECK:       // %bb.0:
74; CHECK-NEXT:    ld1w { z0.d }, p0/z, [z0.d, #16]
75; CHECK-NEXT:    ret
76  %load = call <vscale x 2 x i32> @llvm.aarch64.sve.ld1.gather.scalar.offset.nxv2i32.nxv2i64(<vscale x 2 x i1> %pg,
77                                                                                             <vscale x 2 x i64> %base,
78                                                                                             i64 16)
79  %res = zext <vscale x 2 x i32> %load to <vscale x 2 x i64>
80  ret <vscale x 2 x i64> %res
81}
82
83define <vscale x 4 x float> @gld1w_s_imm_offset_float(<vscale x 4 x i1> %pg, <vscale x 4 x i32> %base) {
84; CHECK-LABEL: gld1w_s_imm_offset_float:
85; CHECK:       // %bb.0:
86; CHECK-NEXT:    ld1w { z0.s }, p0/z, [z0.s, #16]
87; CHECK-NEXT:    ret
88  %load = call <vscale x 4 x float> @llvm.aarch64.sve.ld1.gather.scalar.offset.nxv4f32.nxv4i32(<vscale x 4 x i1> %pg,
89                                                                                               <vscale x 4 x i32> %base,
90                                                                                               i64 16)
91  ret <vscale x 4 x float> %load
92}
93
94; LD1D
95define <vscale x 2 x i64> @gld1d_d_imm_offset(<vscale x 2 x i1> %pg, <vscale x 2 x i64> %base) {
96; CHECK-LABEL: gld1d_d_imm_offset:
97; CHECK:       // %bb.0:
98; CHECK-NEXT:    ld1d { z0.d }, p0/z, [z0.d, #16]
99; CHECK-NEXT:    ret
100  %load = call <vscale x 2 x i64> @llvm.aarch64.sve.ld1.gather.scalar.offset.nxv2i64.nxv2i64(<vscale x 2 x i1> %pg,
101                                                                                             <vscale x 2 x i64> %base,
102                                                                                             i64 16)
103  ret <vscale x 2 x i64> %load
104}
105
106define <vscale x 2 x double> @gld1d_d_imm_offset_double(<vscale x 2 x i1> %pg, <vscale x 2 x i64> %base) {
107; CHECK-LABEL: gld1d_d_imm_offset_double:
108; CHECK:       // %bb.0:
109; CHECK-NEXT:    ld1d { z0.d }, p0/z, [z0.d, #16]
110; CHECK-NEXT:    ret
111  %load = call <vscale x 2 x double> @llvm.aarch64.sve.ld1.gather.scalar.offset.nxv2f64.nxv2i64(<vscale x 2 x i1> %pg,
112                                                                                                <vscale x 2 x i64> %base,
113                                                                                                i64 16)
114  ret <vscale x 2 x double> %load
115}
116
117;
118; LD1SB, LD1SW, LD1SH: vector base + immediate offset (index)
119;   e.g. ld1sh { z0.s }, p0/z, [z0.s, #16]
120;
121
122; LD1SB
123define <vscale x 4 x i32> @gld1sb_s_imm_offset(<vscale x 4 x i1> %pg, <vscale x 4 x i32> %base) {
124; CHECK-LABEL: gld1sb_s_imm_offset:
125; CHECK:       // %bb.0:
126; CHECK-NEXT:    ld1sb { z0.s }, p0/z, [z0.s, #16]
127; CHECK-NEXT:    ret
128  %load = call <vscale x 4 x i8> @llvm.aarch64.sve.ld1.gather.scalar.offset.nxv4i8.nxv4i32(<vscale x 4 x i1> %pg,
129                                                                                           <vscale x 4 x i32> %base,
130                                                                                           i64 16)
131  %res = sext <vscale x 4 x i8> %load to <vscale x 4 x i32>
132  ret <vscale x 4 x i32> %res
133}
134
135define <vscale x 2 x i64> @gld1sb_d_imm_offset(<vscale x 2 x i1> %pg, <vscale x 2 x i64> %base) {
136; CHECK-LABEL: gld1sb_d_imm_offset:
137; CHECK:       // %bb.0:
138; CHECK-NEXT:    ld1sb { z0.d }, p0/z, [z0.d, #16]
139; CHECK-NEXT:    ret
140  %load = call <vscale x 2 x i8> @llvm.aarch64.sve.ld1.gather.scalar.offset.nxv2i8.nxv2i64(<vscale x 2 x i1> %pg,
141                                                                                           <vscale x 2 x i64> %base,
142                                                                                           i64 16)
143  %res = sext <vscale x 2 x i8> %load to <vscale x 2 x i64>
144  ret <vscale x 2 x i64> %res
145}
146
147; LD1SH
148define <vscale x 4 x i32> @gld1sh_s_imm_offset(<vscale x 4 x i1> %pg, <vscale x 4 x i32> %base) {
149; CHECK-LABEL: gld1sh_s_imm_offset:
150; CHECK:       // %bb.0:
151; CHECK-NEXT:    ld1sh { z0.s }, p0/z, [z0.s, #16]
152; CHECK-NEXT:    ret
153  %load = call <vscale x 4 x i16> @llvm.aarch64.sve.ld1.gather.scalar.offset.nxv4i16.nxv4i32(<vscale x 4 x i1> %pg,
154                                                                                             <vscale x 4 x i32> %base,
155                                                                                             i64 16)
156  %res = sext <vscale x 4 x i16> %load to <vscale x 4 x i32>
157  ret <vscale x 4 x i32> %res
158}
159
160define <vscale x 2 x i64> @gld1sh_d_imm_offset(<vscale x 2 x i1> %pg, <vscale x 2 x i64> %base) {
161; CHECK-LABEL: gld1sh_d_imm_offset:
162; CHECK:       // %bb.0:
163; CHECK-NEXT:    ld1sh { z0.d }, p0/z, [z0.d, #16]
164; CHECK-NEXT:    ret
165  %load = call <vscale x 2 x i16> @llvm.aarch64.sve.ld1.gather.scalar.offset.nxv2i16.nxv2i64(<vscale x 2 x i1> %pg,
166                                                                                             <vscale x 2 x i64> %base,
167                                                                                             i64 16)
168  %res = sext <vscale x 2 x i16> %load to <vscale x 2 x i64>
169  ret <vscale x 2 x i64> %res
170}
171
172; LD1SW
173define <vscale x 2 x i64> @gld1sw_d_imm_offset(<vscale x 2 x i1> %pg, <vscale x 2 x i64> %base) {
174; CHECK-LABEL: gld1sw_d_imm_offset:
175; CHECK:       // %bb.0:
176; CHECK-NEXT:    ld1sw { z0.d }, p0/z, [z0.d, #16]
177; CHECK-NEXT:    ret
178  %load = call <vscale x 2 x i32> @llvm.aarch64.sve.ld1.gather.scalar.offset.nxv2i32.nxv2i64(<vscale x 2 x i1> %pg,
179                                                                                             <vscale x 2 x i64> %base,
180                                                                                             i64 16)
181  %res = sext <vscale x 2 x i32> %load to <vscale x 2 x i64>
182  ret <vscale x 2 x i64> %res
183}
184
185;
186; LD1B, LD1W, LD1H, LD1D: vector base + out of range immediate offset
187;   e.g. ld1b { z0.d }, p0/z, [x0, z0.d]
188;
189
190; LD1B
191define <vscale x 4 x i32> @gld1b_s_imm_offset_out_of_range(<vscale x 4 x i1> %pg, <vscale x 4 x i32> %base) {
192; CHECK-LABEL: gld1b_s_imm_offset_out_of_range:
193; CHECK:       // %bb.0:
194; CHECK-NEXT:    mov w8, #32
195; CHECK-NEXT:    ld1b { z0.s }, p0/z, [x8, z0.s, uxtw]
196; CHECK-NEXT:    ret
197  %load = call <vscale x 4 x i8> @llvm.aarch64.sve.ld1.gather.scalar.offset.nxv4i8.nxv4i32(<vscale x 4 x i1> %pg,
198                                                                                           <vscale x 4 x i32> %base,
199                                                                                           i64 32)
200  %res = zext <vscale x 4 x i8> %load to <vscale x 4 x i32>
201  ret <vscale x 4 x i32> %res
202}
203
204define <vscale x 2 x i64> @gld1b_d_imm_offset_out_of_range(<vscale x 2 x i1> %pg, <vscale x 2 x i64> %base) {
205; CHECK-LABEL: gld1b_d_imm_offset_out_of_range:
206; CHECK:       // %bb.0:
207; CHECK-NEXT:    mov w8, #32
208; CHECK-NEXT:    ld1b { z0.d }, p0/z, [x8, z0.d]
209; CHECK-NEXT:    ret
210  %load = call <vscale x 2 x i8> @llvm.aarch64.sve.ld1.gather.scalar.offset.nxv2i8.nxv2i64(<vscale x 2 x i1> %pg,
211                                                                                           <vscale x 2 x i64> %base,
212                                                                                           i64 32)
213  %res = zext <vscale x 2 x i8> %load to <vscale x 2 x i64>
214  ret <vscale x 2 x i64> %res
215}
216
217; LD1H
218define <vscale x 4 x i32> @gld1h_s_imm_offset_out_of_range(<vscale x 4 x i1> %pg, <vscale x 4 x i32> %base) {
219; CHECK-LABEL: gld1h_s_imm_offset_out_of_range:
220; CHECK:       // %bb.0:
221; CHECK-NEXT:    mov w8, #63
222; CHECK-NEXT:    ld1h { z0.s }, p0/z, [x8, z0.s, uxtw]
223; CHECK-NEXT:    ret
224  %load = call <vscale x 4 x i16> @llvm.aarch64.sve.ld1.gather.scalar.offset.nxv4i16.nxv4i32(<vscale x 4 x i1> %pg,
225                                                                                             <vscale x 4 x i32> %base,
226                                                                                             i64 63)
227  %res = zext <vscale x 4 x i16> %load to <vscale x 4 x i32>
228  ret <vscale x 4 x i32> %res
229}
230
231define <vscale x 2 x i64> @gld1h_d_imm_offset_out_of_range(<vscale x 2 x i1> %pg, <vscale x 2 x i64> %base) {
232; CHECK-LABEL: gld1h_d_imm_offset_out_of_range:
233; CHECK:       // %bb.0:
234; CHECK-NEXT:    mov w8, #63
235; CHECK-NEXT:    ld1h { z0.d }, p0/z, [x8, z0.d]
236; CHECK-NEXT:    ret
237  %load = call <vscale x 2 x i16> @llvm.aarch64.sve.ld1.gather.scalar.offset.nxv2i16.nxv2i64(<vscale x 2 x i1> %pg,
238                                                                                             <vscale x 2 x i64> %base,
239                                                                                             i64 63)
240  %res = zext <vscale x 2 x i16> %load to <vscale x 2 x i64>
241  ret <vscale x 2 x i64> %res
242}
243
244; LD1W
245define <vscale x 4 x i32> @gld1w_s_imm_offset_out_of_range(<vscale x 4 x i1> %pg, <vscale x 4 x i32> %base) {
246; CHECK-LABEL: gld1w_s_imm_offset_out_of_range:
247; CHECK:       // %bb.0:
248; CHECK-NEXT:    mov w8, #125
249; CHECK-NEXT:    ld1w { z0.s }, p0/z, [x8, z0.s, uxtw]
250; CHECK-NEXT:    ret
251  %load = call <vscale x 4 x i32> @llvm.aarch64.sve.ld1.gather.scalar.offset.nxv4i32.nxv4i32(<vscale x 4 x i1> %pg,
252                                                                                             <vscale x 4 x i32> %base,
253                                                                                             i64 125)
254  ret <vscale x 4 x i32> %load
255}
256
257define <vscale x 2 x i64> @gld1w_d_imm_offset_out_of_range(<vscale x 2 x i1> %pg, <vscale x 2 x i64> %base) {
258; CHECK-LABEL: gld1w_d_imm_offset_out_of_range:
259; CHECK:       // %bb.0:
260; CHECK-NEXT:    mov w8, #125
261; CHECK-NEXT:    ld1w { z0.d }, p0/z, [x8, z0.d]
262; CHECK-NEXT:    ret
263  %load = call <vscale x 2 x i32> @llvm.aarch64.sve.ld1.gather.scalar.offset.nxv2i32.nxv2i64(<vscale x 2 x i1> %pg,
264                                                                                             <vscale x 2 x i64> %base,
265                                                                                             i64 125)
266  %res = zext <vscale x 2 x i32> %load to <vscale x 2 x i64>
267  ret <vscale x 2 x i64> %res
268}
269
270define <vscale x 4 x float> @gld1w_s_imm_offset_out_of_range_float(<vscale x 4 x i1> %pg, <vscale x 4 x i32> %base) {
271; CHECK-LABEL: gld1w_s_imm_offset_out_of_range_float:
272; CHECK:       // %bb.0:
273; CHECK-NEXT:    mov w8, #125
274; CHECK-NEXT:    ld1w { z0.s }, p0/z, [x8, z0.s, uxtw]
275; CHECK-NEXT:    ret
276  %load = call <vscale x 4 x float> @llvm.aarch64.sve.ld1.gather.scalar.offset.nxv4f32.nxv4i32(<vscale x 4 x i1> %pg,
277                                                                                               <vscale x 4 x i32> %base,
278                                                                                               i64 125)
279  ret <vscale x 4 x float> %load
280}
281
282; LD1D
283define <vscale x 2 x i64> @gld1d_d_imm_offset_out_of_range(<vscale x 2 x i1> %pg, <vscale x 2 x i64> %base) {
284; CHECK-LABEL: gld1d_d_imm_offset_out_of_range:
285; CHECK:       // %bb.0:
286; CHECK-NEXT:    mov w8, #249
287; CHECK-NEXT:    ld1d { z0.d }, p0/z, [x8, z0.d]
288; CHECK-NEXT:    ret
289  %load = call <vscale x 2 x i64> @llvm.aarch64.sve.ld1.gather.scalar.offset.nxv2i64.nxv2i64(<vscale x 2 x i1> %pg,
290                                                                                             <vscale x 2 x i64> %base,
291                                                                                             i64 249)
292  ret <vscale x 2 x i64> %load
293}
294
295define <vscale x 2 x double> @gld1d_d_imm_offset_out_of_range_double(<vscale x 2 x i1> %pg, <vscale x 2 x i64> %base) {
296; CHECK-LABEL: gld1d_d_imm_offset_out_of_range_double:
297; CHECK:       // %bb.0:
298; CHECK-NEXT:    mov w8, #249
299; CHECK-NEXT:    ld1d { z0.d }, p0/z, [x8, z0.d]
300; CHECK-NEXT:    ret
301  %load = call <vscale x 2 x double> @llvm.aarch64.sve.ld1.gather.scalar.offset.nxv2f64.nxv2i64(<vscale x 2 x i1> %pg,
302                                                                                                <vscale x 2 x i64> %base,
303                                                                                                i64 249)
304  ret <vscale x 2 x double> %load
305}
306
307;
308; LD1SB, LD1SW, LD1SH: vector base + out of range immediate offset
309;     e.g. ld1sb { z0.s }, p0/z, [x8, z0.s, uxtw]
310;
311
312; LD1SB
313define <vscale x 4 x i32> @gld1sb_s_imm_offset_out_of_range(<vscale x 4 x i1> %pg, <vscale x 4 x i32> %base) {
314; CHECK-LABEL: gld1sb_s_imm_offset_out_of_range:
315; CHECK:       // %bb.0:
316; CHECK-NEXT:    mov w8, #32
317; CHECK-NEXT:    ld1sb { z0.s }, p0/z, [x8, z0.s, uxtw]
318; CHECK-NEXT:    ret
319  %load = call <vscale x 4 x i8> @llvm.aarch64.sve.ld1.gather.scalar.offset.nxv4i8.nxv4i32(<vscale x 4 x i1> %pg,
320                                                                                           <vscale x 4 x i32> %base,
321                                                                                           i64 32)
322  %res = sext <vscale x 4 x i8> %load to <vscale x 4 x i32>
323  ret <vscale x 4 x i32> %res
324}
325
326define <vscale x 2 x i64> @gld1sb_d_imm_offset_out_of_range(<vscale x 2 x i1> %pg, <vscale x 2 x i64> %base) {
327; CHECK-LABEL: gld1sb_d_imm_offset_out_of_range:
328; CHECK:       // %bb.0:
329; CHECK-NEXT:    mov w8, #32
330; CHECK-NEXT:    ld1sb { z0.d }, p0/z, [x8, z0.d]
331; CHECK-NEXT:    ret
332  %load = call <vscale x 2 x i8> @llvm.aarch64.sve.ld1.gather.scalar.offset.nxv2i8.nxv2i64(<vscale x 2 x i1> %pg,
333                                                                                           <vscale x 2 x i64> %base,
334                                                                                           i64 32)
335  %res = sext <vscale x 2 x i8> %load to <vscale x 2 x i64>
336  ret <vscale x 2 x i64> %res
337}
338
339; LD1SH
340define <vscale x 4 x i32> @gld1sh_s_imm_offset_out_of_range(<vscale x 4 x i1> %pg, <vscale x 4 x i32> %base) {
341; CHECK-LABEL: gld1sh_s_imm_offset_out_of_range:
342; CHECK:       // %bb.0:
343; CHECK-NEXT:    mov w8, #63
344; CHECK-NEXT:    ld1sh { z0.s }, p0/z, [x8, z0.s, uxtw]
345; CHECK-NEXT:    ret
346  %load = call <vscale x 4 x i16> @llvm.aarch64.sve.ld1.gather.scalar.offset.nxv4i16.nxv4i32(<vscale x 4 x i1> %pg,
347                                                                                             <vscale x 4 x i32> %base,
348                                                                                             i64 63)
349  %res = sext <vscale x 4 x i16> %load to <vscale x 4 x i32>
350  ret <vscale x 4 x i32> %res
351}
352
353define <vscale x 2 x i64> @gld1sh_d_imm_offset_out_of_range(<vscale x 2 x i1> %pg, <vscale x 2 x i64> %base) {
354; CHECK-LABEL: gld1sh_d_imm_offset_out_of_range:
355; CHECK:       // %bb.0:
356; CHECK-NEXT:    mov w8, #63
357; CHECK-NEXT:    ld1sh { z0.d }, p0/z, [x8, z0.d]
358; CHECK-NEXT:    ret
359  %load = call <vscale x 2 x i16> @llvm.aarch64.sve.ld1.gather.scalar.offset.nxv2i16.nxv2i64(<vscale x 2 x i1> %pg,
360                                                                                             <vscale x 2 x i64> %base,
361                                                                                             i64 63)
362  %res = sext <vscale x 2 x i16> %load to <vscale x 2 x i64>
363  ret <vscale x 2 x i64> %res
364}
365
366; LD1SW
367define <vscale x 2 x i64> @gld1sw_d_imm_offset_out_of_range(<vscale x 2 x i1> %pg, <vscale x 2 x i64> %base) {
368; CHECK-LABEL: gld1sw_d_imm_offset_out_of_range:
369; CHECK:       // %bb.0:
370; CHECK-NEXT:    mov w8, #125
371; CHECK-NEXT:    ld1sw { z0.d }, p0/z, [x8, z0.d]
372; CHECK-NEXT:    ret
373  %load = call <vscale x 2 x i32> @llvm.aarch64.sve.ld1.gather.scalar.offset.nxv2i32.nxv2i64(<vscale x 2 x i1> %pg,
374                                                                                             <vscale x 2 x i64> %base,
375                                                                                             i64 125)
376  %res = sext <vscale x 2 x i32> %load to <vscale x 2 x i64>
377  ret <vscale x 2 x i64> %res
378}
379
380; LD1B/LD1SB
381declare <vscale x 4 x i8> @llvm.aarch64.sve.ld1.gather.scalar.offset.nxv4i8.nxv4i32(<vscale x 4 x i1>, <vscale x 4 x i32>, i64)
382declare <vscale x 2 x i8> @llvm.aarch64.sve.ld1.gather.scalar.offset.nxv2i8.nxv2i64(<vscale x 2 x i1>, <vscale x 2 x i64>, i64)
383
384; LD1H/LD1SH
385declare <vscale x 4 x i16> @llvm.aarch64.sve.ld1.gather.scalar.offset.nxv4i16.nxv4i32(<vscale x 4 x i1>, <vscale x 4 x i32>, i64)
386declare <vscale x 2 x i16> @llvm.aarch64.sve.ld1.gather.scalar.offset.nxv2i16.nxv2i64(<vscale x 2 x i1>, <vscale x 2 x i64>, i64)
387
388; LD1W/LD1SW
389declare <vscale x 4 x i32> @llvm.aarch64.sve.ld1.gather.scalar.offset.nxv4i32.nxv4i32(<vscale x 4 x i1>, <vscale x 4 x i32>, i64)
390declare <vscale x 2 x i32> @llvm.aarch64.sve.ld1.gather.scalar.offset.nxv2i32.nxv2i64(<vscale x 2 x i1>, <vscale x 2 x i64>, i64)
391
392declare <vscale x 4 x float> @llvm.aarch64.sve.ld1.gather.scalar.offset.nxv4f32.nxv4i32(<vscale x 4 x i1>, <vscale x 4 x i32>, i64)
393
394; LD1D
395declare <vscale x 2 x i64> @llvm.aarch64.sve.ld1.gather.scalar.offset.nxv2i64.nxv2i64(<vscale x 2 x i1>, <vscale x 2 x i64>, i64)
396
397declare <vscale x 2 x double> @llvm.aarch64.sve.ld1.gather.scalar.offset.nxv2f64.nxv2i64(<vscale x 2 x i1>, <vscale x 2 x i64>, i64)
398