1; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
2; RUN: llc -mtriple=aarch64-linux-gnu -mattr=+sve2 < %s | FileCheck %s
3
4;
5; LDNT1B, LDNT1W, LDNT1H, LDNT1D: base + 64-bit unscaled offsets
6;   e.g. ldnt1h { z0.d }, p0/z, [z0.d, x0]
7;
8
9define <vscale x 2 x i64> @gldnt1b_d(<vscale x 2 x i1> %pg, ptr %base, <vscale x 2 x i64> %b) {
10; CHECK-LABEL: gldnt1b_d:
11; CHECK:       // %bb.0:
12; CHECK-NEXT:    ldnt1b { z0.d }, p0/z, [z0.d, x0]
13; CHECK-NEXT:    ret
14  %load = call <vscale x 2 x i8> @llvm.aarch64.sve.ldnt1.gather.nxv2i8(<vscale x 2 x i1> %pg,
15                                                                       ptr %base,
16                                                                       <vscale x 2 x i64> %b)
17  %res = zext <vscale x 2 x i8> %load to <vscale x 2 x i64>
18  ret <vscale x 2 x i64> %res
19}
20
21define <vscale x 2 x i64> @gldnt1h_d(<vscale x 2 x i1> %pg, ptr %base, <vscale x 2 x i64> %b) {
22; CHECK-LABEL: gldnt1h_d:
23; CHECK:       // %bb.0:
24; CHECK-NEXT:    ldnt1h { z0.d }, p0/z, [z0.d, x0]
25; CHECK-NEXT:    ret
26  %load = call <vscale x 2 x i16> @llvm.aarch64.sve.ldnt1.gather.nxv2i16(<vscale x 2 x i1> %pg,
27                                                                         ptr %base,
28                                                                         <vscale x 2 x i64> %b)
29  %res = zext <vscale x 2 x i16> %load to <vscale x 2 x i64>
30  ret <vscale x 2 x i64> %res
31}
32
33define <vscale x 2 x i64> @gldnt1w_d(<vscale x 2 x i1> %pg, ptr %base, <vscale x 2 x i64> %offsets) {
34; CHECK-LABEL: gldnt1w_d:
35; CHECK:       // %bb.0:
36; CHECK-NEXT:    ldnt1w { z0.d }, p0/z, [z0.d, x0]
37; CHECK-NEXT:    ret
38  %load = call <vscale x 2 x i32> @llvm.aarch64.sve.ldnt1.gather.nxv2i32(<vscale x 2 x i1> %pg,
39                                                                         ptr %base,
40                                                                         <vscale x 2 x i64> %offsets)
41  %res = zext <vscale x 2 x i32> %load to <vscale x 2 x i64>
42  ret <vscale x 2 x i64> %res
43}
44
45define <vscale x 2 x i64> @gldnt1d_d(<vscale x 2 x i1> %pg, ptr %base, <vscale x 2 x i64> %b) {
46; CHECK-LABEL: gldnt1d_d:
47; CHECK:       // %bb.0:
48; CHECK-NEXT:    ldnt1d { z0.d }, p0/z, [z0.d, x0]
49; CHECK-NEXT:    ret
50  %load = call <vscale x 2 x i64> @llvm.aarch64.sve.ldnt1.gather.nxv2i64(<vscale x 2 x i1> %pg,
51                                                                         ptr %base,
52                                                                         <vscale x 2 x i64> %b)
53  ret <vscale x 2 x i64> %load
54}
55
56define <vscale x 2 x double> @gldnt1d_d_double(<vscale x 2 x i1> %pg, ptr %base, <vscale x 2 x i64> %b) {
57; CHECK-LABEL: gldnt1d_d_double:
58; CHECK:       // %bb.0:
59; CHECK-NEXT:    ldnt1d { z0.d }, p0/z, [z0.d, x0]
60; CHECK-NEXT:    ret
61  %load = call <vscale x 2 x double> @llvm.aarch64.sve.ldnt1.gather.nxv2f64(<vscale x 2 x i1> %pg,
62                                                                            ptr %base,
63                                                                            <vscale x 2 x i64> %b)
64  ret <vscale x 2 x double> %load
65}
66
67;
68; LDNT1SB, LDNT1SW, LDNT1SH: base + 64-bit unscaled offsets
69;   e.g. ldnt1sh { z0.d }, p0/z, [z0.d, x0]
70;
71
72define <vscale x 2 x i64> @gldnt1sb_d(<vscale x 2 x i1> %pg, ptr %base, <vscale x 2 x i64> %b) {
73; CHECK-LABEL: gldnt1sb_d:
74; CHECK:       // %bb.0:
75; CHECK-NEXT:    ldnt1sb { z0.d }, p0/z, [z0.d, x0]
76; CHECK-NEXT:    ret
77  %load = call <vscale x 2 x i8> @llvm.aarch64.sve.ldnt1.gather.nxv2i8(<vscale x 2 x i1> %pg,
78                                                                       ptr %base,
79                                                                       <vscale x 2 x i64> %b)
80  %res = sext <vscale x 2 x i8> %load to <vscale x 2 x i64>
81  ret <vscale x 2 x i64> %res
82}
83
84define <vscale x 2 x i64> @gldnt1sh_d(<vscale x 2 x i1> %pg, ptr %base, <vscale x 2 x i64> %b) {
85; CHECK-LABEL: gldnt1sh_d:
86; CHECK:       // %bb.0:
87; CHECK-NEXT:    ldnt1sh { z0.d }, p0/z, [z0.d, x0]
88; CHECK-NEXT:    ret
89  %load = call <vscale x 2 x i16> @llvm.aarch64.sve.ldnt1.gather.nxv2i16(<vscale x 2 x i1> %pg,
90                                                                         ptr %base,
91                                                                         <vscale x 2 x i64> %b)
92  %res = sext <vscale x 2 x i16> %load to <vscale x 2 x i64>
93  ret <vscale x 2 x i64> %res
94}
95
96define <vscale x 2 x i64> @gldnt1sw_d(<vscale x 2 x i1> %pg, ptr %base, <vscale x 2 x i64> %offsets) {
97; CHECK-LABEL: gldnt1sw_d:
98; CHECK:       // %bb.0:
99; CHECK-NEXT:    ldnt1sw { z0.d }, p0/z, [z0.d, x0]
100; CHECK-NEXT:    ret
101  %load = call <vscale x 2 x i32> @llvm.aarch64.sve.ldnt1.gather.nxv2i32(<vscale x 2 x i1> %pg,
102                                                                         ptr %base,
103                                                                         <vscale x 2 x i64> %offsets)
104  %res = sext <vscale x 2 x i32> %load to <vscale x 2 x i64>
105  ret <vscale x 2 x i64> %res
106}
107
108declare <vscale x 2 x i8> @llvm.aarch64.sve.ldnt1.gather.nxv2i8(<vscale x 2 x i1>, ptr, <vscale x 2 x i64>)
109declare <vscale x 2 x i16> @llvm.aarch64.sve.ldnt1.gather.nxv2i16(<vscale x 2 x i1>, ptr, <vscale x 2 x i64>)
110declare <vscale x 2 x i32> @llvm.aarch64.sve.ldnt1.gather.nxv2i32(<vscale x 2 x i1>, ptr, <vscale x 2 x i64>)
111declare <vscale x 2 x i64> @llvm.aarch64.sve.ldnt1.gather.nxv2i64(<vscale x 2 x i1>, ptr, <vscale x 2 x i64>)
112declare <vscale x 2 x double> @llvm.aarch64.sve.ldnt1.gather.nxv2f64(<vscale x 2 x i1>, ptr, <vscale x 2 x i64>)
113