Debian Codegraph @

clang/test/CodeGen/AArch64/sve2p1-intrinsics/acle_sve2p1_load_struct.c

raw
1 // NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py
2 // REQUIRES: aarch64-registered-target
3 // RUN: %clang_cc1 -triple aarch64 -target-feature +sve -target-feature +sve2 -target-feature +sve2p1 -disable-O0-optnone -Werror -Wall -emit-llvm -o - %s | opt -S -p mem2reg,instcombine,tailcallelim | FileCheck %s
4 // RUN: %clang_cc1 -triple aarch64 -target-feature +sme -target-feature +sme2 -target-feature +sme2p1 -disable-O0-optnone -Werror -Wall -emit-llvm -o - %s | opt -S -p mem2reg,instcombine,tailcallelim | FileCheck %s
5 // RUN: %clang_cc1 -triple aarch64 -target-feature +sve -target-feature +sme -target-feature +sve2p1 -disable-O0-optnone -Werror -Wall -emit-llvm -o - %s | opt -S -p mem2reg,instcombine,tailcallelim | FileCheck %s
6 // RUN: %clang_cc1 -triple aarch64 -target-feature +sve -target-feature +sme -target-feature +sme2p1 -disable-O0-optnone -Werror -Wall -emit-llvm -o - %s | opt -S -p mem2reg,instcombine,tailcallelim | FileCheck %s
7 // RUN: %clang_cc1 -triple aarch64 -target-feature +sve -target-feature +sve2 -target-feature +sve2p1 -disable-O0-optnone -Werror -Wall -emit-llvm -o - -x c++ %s | opt -S -p mem2reg,instcombine,tailcallelim | FileCheck %s -check-prefix=CPP-CHECK
8 // RUN: %clang_cc1 -DSVE_OVERLOADED_FORMS -triple aarch64 -target-feature +sve -target-feature +sve2 -target-feature +sve2p1 -disable-O0-optnone -Werror -Wall -emit-llvm -o - %s | opt -S -p mem2reg,instcombine,tailcallelim | FileCheck %s
9 // RUN: %clang_cc1 -DSVE_OVERLOADED_FORMS -triple aarch64 -target-feature +sve -target-feature +sve2 -target-feature +sve2p1 -disable-O0-optnone -Werror -Wall -emit-llvm -o - -x c++ %s | opt -S -p mem2reg,instcombine,tailcallelim | FileCheck %s -check-prefix=CPP-CHECK
10 // RUN: %clang_cc1 -triple aarch64 -target-feature +sve -target-feature +sve2 -target-feature +sve2p1 -S -disable-O0-optnone -Werror -Wall -o /dev/null %s
11
12 #include <arm_sve.h>
13
14 #ifdef SVE_OVERLOADED_FORMS
15 // A simple used,unused... macro, long enough to represent any SVE builtin.
16 #define SVE_ACLE_FUNC(A1,A2_UNUSED,A3,A4_UNUSED) A1##A3
17 #else
18 #define SVE_ACLE_FUNC(A1,A2,A3,A4) A1##A2##A3##A4
19 #endif
20
21 #if defined(__ARM_FEATURE_SME) && defined(__ARM_FEATURE_SVE)
22 #define ATTR __arm_streaming_compatible
23 #elif defined(__ARM_FEATURE_SME)
24 #define ATTR __arm_streaming
25 #else
26 #define ATTR
27 #endif
28
29 // CHECK-LABEL: @test_svld2q_u8(
30 // CHECK-NEXT: entry:
31 // CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld2q.sret.nxv16i8(<vscale x 16 x i1> [[PG:%.*]], ptr [[BASE:%.*]])
32 // CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]]
33 //
34 // CPP-CHECK-LABEL: @_Z14test_svld2q_u8u10__SVBool_tPKh(
35 // CPP-CHECK-NEXT: entry:
36 // CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld2q.sret.nxv16i8(<vscale x 16 x i1> [[PG:%.*]], ptr [[BASE:%.*]])
37 // CPP-CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]]
38 //
39 svuint8x2_t test_svld2q_u8(svbool_t pg, const uint8_t *base) ATTR
40 {
41 return SVE_ACLE_FUNC(svld2q,,_u8,)(pg, base);
42 }
43
44 // CHECK-LABEL: @test_svld2q_s8(
45 // CHECK-NEXT: entry:
46 // CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld2q.sret.nxv16i8(<vscale x 16 x i1> [[PG:%.*]], ptr [[BASE:%.*]])
47 // CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]]
48 //
49 // CPP-CHECK-LABEL: @_Z14test_svld2q_s8u10__SVBool_tPKa(
50 // CPP-CHECK-NEXT: entry:
51 // CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld2q.sret.nxv16i8(<vscale x 16 x i1> [[PG:%.*]], ptr [[BASE:%.*]])
52 // CPP-CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]]
53 //
54 svint8x2_t test_svld2q_s8(svbool_t pg, const int8_t *base) ATTR
55 {
56 return SVE_ACLE_FUNC(svld2q,,_s8,)(pg, base);
57 }
58 // CHECK-LABEL: @test_svld2q_u16(
59 // CHECK-NEXT: entry:
60 // CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
61 // CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.ld2q.sret.nxv8i16(<vscale x 8 x i1> [[TMP0]], ptr [[BASE:%.*]])
62 // CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]]
63 //
64 // CPP-CHECK-LABEL: @_Z15test_svld2q_u16u10__SVBool_tPKt(
65 // CPP-CHECK-NEXT: entry:
66 // CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
67 // CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.ld2q.sret.nxv8i16(<vscale x 8 x i1> [[TMP0]], ptr [[BASE:%.*]])
68 // CPP-CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]]
69 //
70 svuint16x2_t test_svld2q_u16(svbool_t pg, const uint16_t *base) ATTR
71 {
72 return SVE_ACLE_FUNC(svld2q,,_u16,)(pg, base);
73 }
74
75 // CHECK-LABEL: @test_svld2q_s16(
76 // CHECK-NEXT: entry:
77 // CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
78 // CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.ld2q.sret.nxv8i16(<vscale x 8 x i1> [[TMP0]], ptr [[BASE:%.*]])
79 // CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]]
80 //
81 // CPP-CHECK-LABEL: @_Z15test_svld2q_s16u10__SVBool_tPKs(
82 // CPP-CHECK-NEXT: entry:
83 // CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
84 // CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.ld2q.sret.nxv8i16(<vscale x 8 x i1> [[TMP0]], ptr [[BASE:%.*]])
85 // CPP-CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]]
86 //
87 svint16x2_t test_svld2q_s16(svbool_t pg, const int16_t *base) ATTR
88 {
89 return SVE_ACLE_FUNC(svld2q,,_s16,)(pg, base);
90 }
91
92 // CHECK-LABEL: @test_svld2q_u32(
93 // CHECK-NEXT: entry:
94 // CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
95 // CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.ld2q.sret.nxv4i32(<vscale x 4 x i1> [[TMP0]], ptr [[BASE:%.*]])
96 // CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]]
97 //
98 // CPP-CHECK-LABEL: @_Z15test_svld2q_u32u10__SVBool_tPKj(
99 // CPP-CHECK-NEXT: entry:
100 // CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
101 // CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.ld2q.sret.nxv4i32(<vscale x 4 x i1> [[TMP0]], ptr [[BASE:%.*]])
102 // CPP-CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]]
103 //
104 svuint32x2_t test_svld2q_u32(svbool_t pg, const uint32_t *base) ATTR
105 {
106 return SVE_ACLE_FUNC(svld2q,,_u32,)(pg, base);
107 }
108
109 // CHECK-LABEL: @test_svld2q_s32(
110 // CHECK-NEXT: entry:
111 // CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
112 // CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.ld2q.sret.nxv4i32(<vscale x 4 x i1> [[TMP0]], ptr [[BASE:%.*]])
113 // CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]]
114 //
115 // CPP-CHECK-LABEL: @_Z15test_svld2q_s32u10__SVBool_tPKi(
116 // CPP-CHECK-NEXT: entry:
117 // CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
118 // CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.ld2q.sret.nxv4i32(<vscale x 4 x i1> [[TMP0]], ptr [[BASE:%.*]])
119 // CPP-CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]]
120 //
121 svint32x2_t test_svld2q_s32(svbool_t pg, const int32_t *base) ATTR
122 {
123 return SVE_ACLE_FUNC(svld2q,,_s32,)(pg, base);
124 }
125
126 // CHECK-LABEL: @test_svld2q_u64(
127 // CHECK-NEXT: entry:
128 // CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
129 // CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.ld2q.sret.nxv2i64(<vscale x 2 x i1> [[TMP0]], ptr [[BASE:%.*]])
130 // CHECK-NEXT: ret { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]]
131 //
132 // CPP-CHECK-LABEL: @_Z15test_svld2q_u64u10__SVBool_tPKm(
133 // CPP-CHECK-NEXT: entry:
134 // CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
135 // CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.ld2q.sret.nxv2i64(<vscale x 2 x i1> [[TMP0]], ptr [[BASE:%.*]])
136 // CPP-CHECK-NEXT: ret { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]]
137 //
138 svuint64x2_t test_svld2q_u64(svbool_t pg, const uint64_t *base) ATTR
139 {
140 return SVE_ACLE_FUNC(svld2q,,_u64,)(pg, base);
141 }
142
143 // CHECK-LABEL: @test_svld2q_s64(
144 // CHECK-NEXT: entry:
145 // CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
146 // CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.ld2q.sret.nxv2i64(<vscale x 2 x i1> [[TMP0]], ptr [[BASE:%.*]])
147 // CHECK-NEXT: ret { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]]
148 //
149 // CPP-CHECK-LABEL: @_Z15test_svld2q_s64u10__SVBool_tPKl(
150 // CPP-CHECK-NEXT: entry:
151 // CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
152 // CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.ld2q.sret.nxv2i64(<vscale x 2 x i1> [[TMP0]], ptr [[BASE:%.*]])
153 // CPP-CHECK-NEXT: ret { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]]
154 //
155 svint64x2_t test_svld2q_s64(svbool_t pg, const int64_t *base) ATTR
156 {
157 return SVE_ACLE_FUNC(svld2q,,_s64,)(pg, base);
158 }
159
160 // CHECK-LABEL: @test_svld2q_f16(
161 // CHECK-NEXT: entry:
162 // CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
163 // CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sve.ld2q.sret.nxv8f16(<vscale x 8 x i1> [[TMP0]], ptr [[BASE:%.*]])
164 // CHECK-NEXT: ret { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP1]]
165 //
166 // CPP-CHECK-LABEL: @_Z15test_svld2q_f16u10__SVBool_tPKDh(
167 // CPP-CHECK-NEXT: entry:
168 // CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
169 // CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sve.ld2q.sret.nxv8f16(<vscale x 8 x i1> [[TMP0]], ptr [[BASE:%.*]])
170 // CPP-CHECK-NEXT: ret { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP1]]
171 //
172 svfloat16x2_t test_svld2q_f16(svbool_t pg, const float16_t *base) ATTR
173 {
174 return SVE_ACLE_FUNC(svld2q,,_f16,)(pg, base);
175 }
176
177 // CHECK-LABEL: @test_svld2q_bf16(
178 // CHECK-NEXT: entry:
179 // CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
180 // CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } @llvm.aarch64.sve.ld2q.sret.nxv8bf16(<vscale x 8 x i1> [[TMP0]], ptr [[BASE:%.*]])
181 // CHECK-NEXT: ret { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP1]]
182 //
183 // CPP-CHECK-LABEL: @_Z16test_svld2q_bf16u10__SVBool_tPKu6__bf16(
184 // CPP-CHECK-NEXT: entry:
185 // CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
186 // CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } @llvm.aarch64.sve.ld2q.sret.nxv8bf16(<vscale x 8 x i1> [[TMP0]], ptr [[BASE:%.*]])
187 // CPP-CHECK-NEXT: ret { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP1]]
188 //
189 svbfloat16x2_t test_svld2q_bf16(svbool_t pg, const bfloat16_t *base) ATTR
190 {
191 return SVE_ACLE_FUNC(svld2q,,_bf16,)(pg, base);
192 }
193
194 // CHECK-LABEL: @test_svld2q_f32(
195 // CHECK-NEXT: entry:
196 // CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
197 // CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.ld2q.sret.nxv4f32(<vscale x 4 x i1> [[TMP0]], ptr [[BASE:%.*]])
198 // CHECK-NEXT: ret { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP1]]
199 //
200 // CPP-CHECK-LABEL: @_Z15test_svld2q_f32u10__SVBool_tPKf(
201 // CPP-CHECK-NEXT: entry:
202 // CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
203 // CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.ld2q.sret.nxv4f32(<vscale x 4 x i1> [[TMP0]], ptr [[BASE:%.*]])
204 // CPP-CHECK-NEXT: ret { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP1]]
205 //
206 svfloat32x2_t test_svld2q_f32(svbool_t pg, const float32_t *base) ATTR
207 {
208 return SVE_ACLE_FUNC(svld2q,,_f32,)(pg, base);
209 }
210
211 // CHECK-LABEL: @test_svld2q_f64(
212 // CHECK-NEXT: entry:
213 // CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
214 // CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 2 x double>, <vscale x 2 x double> } @llvm.aarch64.sve.ld2q.sret.nxv2f64(<vscale x 2 x i1> [[TMP0]], ptr [[BASE:%.*]])
215 // CHECK-NEXT: ret { <vscale x 2 x double>, <vscale x 2 x double> } [[TMP1]]
216 //
217 // CPP-CHECK-LABEL: @_Z15test_svld2q_f64u10__SVBool_tPKd(
218 // CPP-CHECK-NEXT: entry:
219 // CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
220 // CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 2 x double>, <vscale x 2 x double> } @llvm.aarch64.sve.ld2q.sret.nxv2f64(<vscale x 2 x i1> [[TMP0]], ptr [[BASE:%.*]])
221 // CPP-CHECK-NEXT: ret { <vscale x 2 x double>, <vscale x 2 x double> } [[TMP1]]
222 //
223 svfloat64x2_t test_svld2q_f64(svbool_t pg, const float64_t *base) ATTR
224 {
225 return SVE_ACLE_FUNC(svld2q,,_f64,)(pg, base);
226 }
227
228 // CHECK-LABEL: @test_svld2q_mf8(
229 // CHECK-NEXT: entry:
230 // CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld2q.sret.nxv16i8(<vscale x 16 x i1> [[PG:%.*]], ptr [[BASE:%.*]])
231 // CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]]
232 //
233 // CPP-CHECK-LABEL: @_Z15test_svld2q_mf8u10__SVBool_tPKu6__mfp8(
234 // CPP-CHECK-NEXT: entry:
235 // CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld2q.sret.nxv16i8(<vscale x 16 x i1> [[PG:%.*]], ptr [[BASE:%.*]])
236 // CPP-CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]]
237 //
238 svmfloat8x2_t test_svld2q_mf8(svbool_t pg, const mfloat8_t *base) ATTR
239 {
240 return SVE_ACLE_FUNC(svld2q,,_mf8,)(pg, base);
241 }
242
243 // CHECK-LABEL: @test_svld2q_vnum_u8(
244 // CHECK-NEXT: entry:
245 // CHECK-NEXT: [[TMP0:%.*]] = tail call i64 @llvm.vscale.i64()
246 // CHECK-NEXT: [[TMP1:%.*]] = shl nuw nsw i64 [[TMP0]], 4
247 // CHECK-NEXT: [[DOTIDX:%.*]] = mul i64 [[VNUM:%.*]], [[TMP1]]
248 // CHECK-NEXT: [[TMP2:%.*]] = getelementptr i8, ptr [[BASE:%.*]], i64 [[DOTIDX]]
249 // CHECK-NEXT: [[TMP3:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld2q.sret.nxv16i8(<vscale x 16 x i1> [[PG:%.*]], ptr [[TMP2]])
250 // CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP3]]
251 //
252 // CPP-CHECK-LABEL: @_Z19test_svld2q_vnum_u8u10__SVBool_tPKhl(
253 // CPP-CHECK-NEXT: entry:
254 // CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call i64 @llvm.vscale.i64()
255 // CPP-CHECK-NEXT: [[TMP1:%.*]] = shl nuw nsw i64 [[TMP0]], 4
256 // CPP-CHECK-NEXT: [[DOTIDX:%.*]] = mul i64 [[VNUM:%.*]], [[TMP1]]
257 // CPP-CHECK-NEXT: [[TMP2:%.*]] = getelementptr i8, ptr [[BASE:%.*]], i64 [[DOTIDX]]
258 // CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld2q.sret.nxv16i8(<vscale x 16 x i1> [[PG:%.*]], ptr [[TMP2]])
259 // CPP-CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP3]]
260 //
261 svuint8x2_t test_svld2q_vnum_u8(svbool_t pg, const uint8_t *base, int64_t vnum) ATTR
262 {
263 return SVE_ACLE_FUNC(svld2q_vnum_,,u8,)(pg, base, vnum);
264 }
265
266 // CHECK-LABEL: @test_svld2q_vnum_s8(
267 // CHECK-NEXT: entry:
268 // CHECK-NEXT: [[TMP0:%.*]] = tail call i64 @llvm.vscale.i64()
269 // CHECK-NEXT: [[TMP1:%.*]] = shl nuw nsw i64 [[TMP0]], 4
270 // CHECK-NEXT: [[DOTIDX:%.*]] = mul i64 [[VNUM:%.*]], [[TMP1]]
271 // CHECK-NEXT: [[TMP2:%.*]] = getelementptr i8, ptr [[BASE:%.*]], i64 [[DOTIDX]]
272 // CHECK-NEXT: [[TMP3:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld2q.sret.nxv16i8(<vscale x 16 x i1> [[PG:%.*]], ptr [[TMP2]])
273 // CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP3]]
274 //
275 // CPP-CHECK-LABEL: @_Z19test_svld2q_vnum_s8u10__SVBool_tPKal(
276 // CPP-CHECK-NEXT: entry:
277 // CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call i64 @llvm.vscale.i64()
278 // CPP-CHECK-NEXT: [[TMP1:%.*]] = shl nuw nsw i64 [[TMP0]], 4
279 // CPP-CHECK-NEXT: [[DOTIDX:%.*]] = mul i64 [[VNUM:%.*]], [[TMP1]]
280 // CPP-CHECK-NEXT: [[TMP2:%.*]] = getelementptr i8, ptr [[BASE:%.*]], i64 [[DOTIDX]]
281 // CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld2q.sret.nxv16i8(<vscale x 16 x i1> [[PG:%.*]], ptr [[TMP2]])
282 // CPP-CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP3]]
283 //
284 svint8x2_t test_svld2q_vnum_s8(svbool_t pg, const int8_t *base, int64_t vnum) ATTR
285 {
286 return SVE_ACLE_FUNC(svld2q_vnum,,_s8,)(pg, base, vnum);
287 }
288 // CHECK-LABEL: @test_svld2q_vnum_u16(
289 // CHECK-NEXT: entry:
290 // CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
291 // CHECK-NEXT: [[TMP1:%.*]] = tail call i64 @llvm.vscale.i64()
292 // CHECK-NEXT: [[TMP2:%.*]] = shl nuw nsw i64 [[TMP1]], 4
293 // CHECK-NEXT: [[DOTIDX:%.*]] = mul i64 [[VNUM:%.*]], [[TMP2]]
294 // CHECK-NEXT: [[TMP3:%.*]] = getelementptr i8, ptr [[BASE:%.*]], i64 [[DOTIDX]]
295 // CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.ld2q.sret.nxv8i16(<vscale x 8 x i1> [[TMP0]], ptr [[TMP3]])
296 // CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP4]]
297 //
298 // CPP-CHECK-LABEL: @_Z20test_svld2q_vnum_u16u10__SVBool_tPKtl(
299 // CPP-CHECK-NEXT: entry:
300 // CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
301 // CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call i64 @llvm.vscale.i64()
302 // CPP-CHECK-NEXT: [[TMP2:%.*]] = shl nuw nsw i64 [[TMP1]], 4
303 // CPP-CHECK-NEXT: [[DOTIDX:%.*]] = mul i64 [[VNUM:%.*]], [[TMP2]]
304 // CPP-CHECK-NEXT: [[TMP3:%.*]] = getelementptr i8, ptr [[BASE:%.*]], i64 [[DOTIDX]]
305 // CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.ld2q.sret.nxv8i16(<vscale x 8 x i1> [[TMP0]], ptr [[TMP3]])
306 // CPP-CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP4]]
307 //
308 svuint16x2_t test_svld2q_vnum_u16(svbool_t pg, const uint16_t *base, int64_t vnum) ATTR
309 {
310 return SVE_ACLE_FUNC(svld2q_vnum,,_u16,)(pg, base, vnum);
311 }
312
313 // CHECK-LABEL: @test_svld2q_vnum_s16(
314 // CHECK-NEXT: entry:
315 // CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
316 // CHECK-NEXT: [[TMP1:%.*]] = tail call i64 @llvm.vscale.i64()
317 // CHECK-NEXT: [[TMP2:%.*]] = shl nuw nsw i64 [[TMP1]], 4
318 // CHECK-NEXT: [[DOTIDX:%.*]] = mul i64 [[VNUM:%.*]], [[TMP2]]
319 // CHECK-NEXT: [[TMP3:%.*]] = getelementptr i8, ptr [[BASE:%.*]], i64 [[DOTIDX]]
320 // CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.ld2q.sret.nxv8i16(<vscale x 8 x i1> [[TMP0]], ptr [[TMP3]])
321 // CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP4]]
322 //
323 // CPP-CHECK-LABEL: @_Z20test_svld2q_vnum_s16u10__SVBool_tPKsl(
324 // CPP-CHECK-NEXT: entry:
325 // CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
326 // CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call i64 @llvm.vscale.i64()
327 // CPP-CHECK-NEXT: [[TMP2:%.*]] = shl nuw nsw i64 [[TMP1]], 4
328 // CPP-CHECK-NEXT: [[DOTIDX:%.*]] = mul i64 [[VNUM:%.*]], [[TMP2]]
329 // CPP-CHECK-NEXT: [[TMP3:%.*]] = getelementptr i8, ptr [[BASE:%.*]], i64 [[DOTIDX]]
330 // CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.ld2q.sret.nxv8i16(<vscale x 8 x i1> [[TMP0]], ptr [[TMP3]])
331 // CPP-CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP4]]
332 //
333 svint16x2_t test_svld2q_vnum_s16(svbool_t pg, const int16_t *base, int64_t vnum) ATTR
334 {
335 return SVE_ACLE_FUNC(svld2q_vnum,,_s16,)(pg, base, vnum);
336 }
337
338 // CHECK-LABEL: @test_svld2q_vnum_u32(
339 // CHECK-NEXT: entry:
340 // CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
341 // CHECK-NEXT: [[TMP1:%.*]] = tail call i64 @llvm.vscale.i64()
342 // CHECK-NEXT: [[TMP2:%.*]] = shl nuw nsw i64 [[TMP1]], 4
343 // CHECK-NEXT: [[DOTIDX:%.*]] = mul i64 [[VNUM:%.*]], [[TMP2]]
344 // CHECK-NEXT: [[TMP3:%.*]] = getelementptr i8, ptr [[BASE:%.*]], i64 [[DOTIDX]]
345 // CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.ld2q.sret.nxv4i32(<vscale x 4 x i1> [[TMP0]], ptr [[TMP3]])
346 // CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP4]]
347 //
348 // CPP-CHECK-LABEL: @_Z20test_svld2q_vnum_u32u10__SVBool_tPKjl(
349 // CPP-CHECK-NEXT: entry:
350 // CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
351 // CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call i64 @llvm.vscale.i64()
352 // CPP-CHECK-NEXT: [[TMP2:%.*]] = shl nuw nsw i64 [[TMP1]], 4
353 // CPP-CHECK-NEXT: [[DOTIDX:%.*]] = mul i64 [[VNUM:%.*]], [[TMP2]]
354 // CPP-CHECK-NEXT: [[TMP3:%.*]] = getelementptr i8, ptr [[BASE:%.*]], i64 [[DOTIDX]]
355 // CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.ld2q.sret.nxv4i32(<vscale x 4 x i1> [[TMP0]], ptr [[TMP3]])
356 // CPP-CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP4]]
357 //
358 svuint32x2_t test_svld2q_vnum_u32(svbool_t pg, const uint32_t *base, int64_t vnum) ATTR
359 {
360 return SVE_ACLE_FUNC(svld2q_vnum,,_u32,)(pg, base, vnum);
361 }
362
363 // CHECK-LABEL: @test_svld2q_vnum_s32(
364 // CHECK-NEXT: entry:
365 // CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
366 // CHECK-NEXT: [[TMP1:%.*]] = tail call i64 @llvm.vscale.i64()
367 // CHECK-NEXT: [[TMP2:%.*]] = shl nuw nsw i64 [[TMP1]], 4
368 // CHECK-NEXT: [[DOTIDX:%.*]] = mul i64 [[VNUM:%.*]], [[TMP2]]
369 // CHECK-NEXT: [[TMP3:%.*]] = getelementptr i8, ptr [[BASE:%.*]], i64 [[DOTIDX]]
370 // CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.ld2q.sret.nxv4i32(<vscale x 4 x i1> [[TMP0]], ptr [[TMP3]])
371 // CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP4]]
372 //
373 // CPP-CHECK-LABEL: @_Z20test_svld2q_vnum_s32u10__SVBool_tPKil(
374 // CPP-CHECK-NEXT: entry:
375 // CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
376 // CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call i64 @llvm.vscale.i64()
377 // CPP-CHECK-NEXT: [[TMP2:%.*]] = shl nuw nsw i64 [[TMP1]], 4
378 // CPP-CHECK-NEXT: [[DOTIDX:%.*]] = mul i64 [[VNUM:%.*]], [[TMP2]]
379 // CPP-CHECK-NEXT: [[TMP3:%.*]] = getelementptr i8, ptr [[BASE:%.*]], i64 [[DOTIDX]]
380 // CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.ld2q.sret.nxv4i32(<vscale x 4 x i1> [[TMP0]], ptr [[TMP3]])
381 // CPP-CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP4]]
382 //
383 svint32x2_t test_svld2q_vnum_s32(svbool_t pg, const int32_t *base, int64_t vnum) ATTR
384 {
385 return SVE_ACLE_FUNC(svld2q_vnum,,_s32,)(pg, base, vnum);
386 }
387
388 // CHECK-LABEL: @test_svld2q_vnum_u64(
389 // CHECK-NEXT: entry:
390 // CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
391 // CHECK-NEXT: [[TMP1:%.*]] = tail call i64 @llvm.vscale.i64()
392 // CHECK-NEXT: [[TMP2:%.*]] = shl nuw nsw i64 [[TMP1]], 4
393 // CHECK-NEXT: [[DOTIDX:%.*]] = mul i64 [[VNUM:%.*]], [[TMP2]]
394 // CHECK-NEXT: [[TMP3:%.*]] = getelementptr i8, ptr [[BASE:%.*]], i64 [[DOTIDX]]
395 // CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.ld2q.sret.nxv2i64(<vscale x 2 x i1> [[TMP0]], ptr [[TMP3]])
396 // CHECK-NEXT: ret { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP4]]
397 //
398 // CPP-CHECK-LABEL: @_Z20test_svld2q_vnum_u64u10__SVBool_tPKml(
399 // CPP-CHECK-NEXT: entry:
400 // CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
401 // CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call i64 @llvm.vscale.i64()
402 // CPP-CHECK-NEXT: [[TMP2:%.*]] = shl nuw nsw i64 [[TMP1]], 4
403 // CPP-CHECK-NEXT: [[DOTIDX:%.*]] = mul i64 [[VNUM:%.*]], [[TMP2]]
404 // CPP-CHECK-NEXT: [[TMP3:%.*]] = getelementptr i8, ptr [[BASE:%.*]], i64 [[DOTIDX]]
405 // CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.ld2q.sret.nxv2i64(<vscale x 2 x i1> [[TMP0]], ptr [[TMP3]])
406 // CPP-CHECK-NEXT: ret { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP4]]
407 //
408 svuint64x2_t test_svld2q_vnum_u64(svbool_t pg, const uint64_t *base, int64_t vnum) ATTR
409 {
410 return SVE_ACLE_FUNC(svld2q_vnum,,_u64,)(pg, base, vnum);
411 }
412
413 // CHECK-LABEL: @test_svld2q_vnum_s64(
414 // CHECK-NEXT: entry:
415 // CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
416 // CHECK-NEXT: [[TMP1:%.*]] = tail call i64 @llvm.vscale.i64()
417 // CHECK-NEXT: [[TMP2:%.*]] = shl nuw nsw i64 [[TMP1]], 4
418 // CHECK-NEXT: [[DOTIDX:%.*]] = mul i64 [[VNUM:%.*]], [[TMP2]]
419 // CHECK-NEXT: [[TMP3:%.*]] = getelementptr i8, ptr [[BASE:%.*]], i64 [[DOTIDX]]
420 // CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.ld2q.sret.nxv2i64(<vscale x 2 x i1> [[TMP0]], ptr [[TMP3]])
421 // CHECK-NEXT: ret { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP4]]
422 //
423 // CPP-CHECK-LABEL: @_Z20test_svld2q_vnum_s64u10__SVBool_tPKll(
424 // CPP-CHECK-NEXT: entry:
425 // CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
426 // CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call i64 @llvm.vscale.i64()
427 // CPP-CHECK-NEXT: [[TMP2:%.*]] = shl nuw nsw i64 [[TMP1]], 4
428 // CPP-CHECK-NEXT: [[DOTIDX:%.*]] = mul i64 [[VNUM:%.*]], [[TMP2]]
429 // CPP-CHECK-NEXT: [[TMP3:%.*]] = getelementptr i8, ptr [[BASE:%.*]], i64 [[DOTIDX]]
430 // CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.ld2q.sret.nxv2i64(<vscale x 2 x i1> [[TMP0]], ptr [[TMP3]])
431 // CPP-CHECK-NEXT: ret { <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP4]]
432 //
433 svint64x2_t test_svld2q_vnum_s64(svbool_t pg, const int64_t *base, int64_t vnum) ATTR
434 {
435 return SVE_ACLE_FUNC(svld2q_vnum,,_s64,)(pg, base, vnum);
436 }
437
438 // CHECK-LABEL: @test_svld2q_vnum_f16(
439 // CHECK-NEXT: entry:
440 // CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
441 // CHECK-NEXT: [[TMP1:%.*]] = tail call i64 @llvm.vscale.i64()
442 // CHECK-NEXT: [[TMP2:%.*]] = shl nuw nsw i64 [[TMP1]], 4
443 // CHECK-NEXT: [[DOTIDX:%.*]] = mul i64 [[VNUM:%.*]], [[TMP2]]
444 // CHECK-NEXT: [[TMP3:%.*]] = getelementptr i8, ptr [[BASE:%.*]], i64 [[DOTIDX]]
445 // CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sve.ld2q.sret.nxv8f16(<vscale x 8 x i1> [[TMP0]], ptr [[TMP3]])
446 // CHECK-NEXT: ret { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP4]]
447 //
448 // CPP-CHECK-LABEL: @_Z20test_svld2q_vnum_f16u10__SVBool_tPKDhl(
449 // CPP-CHECK-NEXT: entry:
450 // CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
451 // CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call i64 @llvm.vscale.i64()
452 // CPP-CHECK-NEXT: [[TMP2:%.*]] = shl nuw nsw i64 [[TMP1]], 4
453 // CPP-CHECK-NEXT: [[DOTIDX:%.*]] = mul i64 [[VNUM:%.*]], [[TMP2]]
454 // CPP-CHECK-NEXT: [[TMP3:%.*]] = getelementptr i8, ptr [[BASE:%.*]], i64 [[DOTIDX]]
455 // CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sve.ld2q.sret.nxv8f16(<vscale x 8 x i1> [[TMP0]], ptr [[TMP3]])
456 // CPP-CHECK-NEXT: ret { <vscale x 8 x half>, <vscale x 8 x half> } [[TMP4]]
457 //
458 svfloat16x2_t test_svld2q_vnum_f16(svbool_t pg, const float16_t *base, int64_t vnum) ATTR
459 {
460 return SVE_ACLE_FUNC(svld2q_vnum,,_f16,)(pg, base, vnum);
461 }
462
463 // CHECK-LABEL: @test_svld2q_vnum_bf16(
464 // CHECK-NEXT: entry:
465 // CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
466 // CHECK-NEXT: [[TMP1:%.*]] = tail call i64 @llvm.vscale.i64()
467 // CHECK-NEXT: [[TMP2:%.*]] = shl nuw nsw i64 [[TMP1]], 4
468 // CHECK-NEXT: [[DOTIDX:%.*]] = mul i64 [[VNUM:%.*]], [[TMP2]]
469 // CHECK-NEXT: [[TMP3:%.*]] = getelementptr i8, ptr [[BASE:%.*]], i64 [[DOTIDX]]
470 // CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } @llvm.aarch64.sve.ld2q.sret.nxv8bf16(<vscale x 8 x i1> [[TMP0]], ptr [[TMP3]])
471 // CHECK-NEXT: ret { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP4]]
472 //
473 // CPP-CHECK-LABEL: @_Z21test_svld2q_vnum_bf16u10__SVBool_tPKu6__bf16l(
474 // CPP-CHECK-NEXT: entry:
475 // CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
476 // CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call i64 @llvm.vscale.i64()
477 // CPP-CHECK-NEXT: [[TMP2:%.*]] = shl nuw nsw i64 [[TMP1]], 4
478 // CPP-CHECK-NEXT: [[DOTIDX:%.*]] = mul i64 [[VNUM:%.*]], [[TMP2]]
479 // CPP-CHECK-NEXT: [[TMP3:%.*]] = getelementptr i8, ptr [[BASE:%.*]], i64 [[DOTIDX]]
480 // CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } @llvm.aarch64.sve.ld2q.sret.nxv8bf16(<vscale x 8 x i1> [[TMP0]], ptr [[TMP3]])
481 // CPP-CHECK-NEXT: ret { <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP4]]
482 //
483 svbfloat16x2_t test_svld2q_vnum_bf16(svbool_t pg, const bfloat16_t *base, int64_t vnum) ATTR
484 {
485 return SVE_ACLE_FUNC(svld2q_vnum,,_bf16,)(pg, base, vnum);
486 }
487
488 // CHECK-LABEL: @test_svld2q_vnum_f32(
489 // CHECK-NEXT: entry:
490 // CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
491 // CHECK-NEXT: [[TMP1:%.*]] = tail call i64 @llvm.vscale.i64()
492 // CHECK-NEXT: [[TMP2:%.*]] = shl nuw nsw i64 [[TMP1]], 4
493 // CHECK-NEXT: [[DOTIDX:%.*]] = mul i64 [[VNUM:%.*]], [[TMP2]]
494 // CHECK-NEXT: [[TMP3:%.*]] = getelementptr i8, ptr [[BASE:%.*]], i64 [[DOTIDX]]
495 // CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.ld2q.sret.nxv4f32(<vscale x 4 x i1> [[TMP0]], ptr [[TMP3]])
496 // CHECK-NEXT: ret { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP4]]
497 //
498 // CPP-CHECK-LABEL: @_Z20test_svld2q_vnum_f32u10__SVBool_tPKfl(
499 // CPP-CHECK-NEXT: entry:
500 // CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
501 // CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call i64 @llvm.vscale.i64()
502 // CPP-CHECK-NEXT: [[TMP2:%.*]] = shl nuw nsw i64 [[TMP1]], 4
503 // CPP-CHECK-NEXT: [[DOTIDX:%.*]] = mul i64 [[VNUM:%.*]], [[TMP2]]
504 // CPP-CHECK-NEXT: [[TMP3:%.*]] = getelementptr i8, ptr [[BASE:%.*]], i64 [[DOTIDX]]
505 // CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.ld2q.sret.nxv4f32(<vscale x 4 x i1> [[TMP0]], ptr [[TMP3]])
506 // CPP-CHECK-NEXT: ret { <vscale x 4 x float>, <vscale x 4 x float> } [[TMP4]]
507 //
508 svfloat32x2_t test_svld2q_vnum_f32(svbool_t pg, const float32_t *base, int64_t vnum) ATTR
509 {
510 return SVE_ACLE_FUNC(svld2q_vnum,,_f32,)(pg, base, vnum);
511 }
512
513 // CHECK-LABEL: @test_svld2q_vnum_f64(
514 // CHECK-NEXT: entry:
515 // CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
516 // CHECK-NEXT: [[TMP1:%.*]] = tail call i64 @llvm.vscale.i64()
517 // CHECK-NEXT: [[TMP2:%.*]] = shl nuw nsw i64 [[TMP1]], 4
518 // CHECK-NEXT: [[DOTIDX:%.*]] = mul i64 [[VNUM:%.*]], [[TMP2]]
519 // CHECK-NEXT: [[TMP3:%.*]] = getelementptr i8, ptr [[BASE:%.*]], i64 [[DOTIDX]]
520 // CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 2 x double>, <vscale x 2 x double> } @llvm.aarch64.sve.ld2q.sret.nxv2f64(<vscale x 2 x i1> [[TMP0]], ptr [[TMP3]])
521 // CHECK-NEXT: ret { <vscale x 2 x double>, <vscale x 2 x double> } [[TMP4]]
522 //
523 // CPP-CHECK-LABEL: @_Z20test_svld2q_vnum_f64u10__SVBool_tPKdl(
524 // CPP-CHECK-NEXT: entry:
525 // CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
526 // CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call i64 @llvm.vscale.i64()
527 // CPP-CHECK-NEXT: [[TMP2:%.*]] = shl nuw nsw i64 [[TMP1]], 4
528 // CPP-CHECK-NEXT: [[DOTIDX:%.*]] = mul i64 [[VNUM:%.*]], [[TMP2]]
529 // CPP-CHECK-NEXT: [[TMP3:%.*]] = getelementptr i8, ptr [[BASE:%.*]], i64 [[DOTIDX]]
530 // CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 2 x double>, <vscale x 2 x double> } @llvm.aarch64.sve.ld2q.sret.nxv2f64(<vscale x 2 x i1> [[TMP0]], ptr [[TMP3]])
531 // CPP-CHECK-NEXT: ret { <vscale x 2 x double>, <vscale x 2 x double> } [[TMP4]]
532 //
533 svfloat64x2_t test_svld2q_vnum_f64(svbool_t pg, const float64_t *base, int64_t vnum) ATTR
534 {
535 return SVE_ACLE_FUNC(svld2q_vnum,,_f64,)(pg, base, vnum);
536 }
537
538 // CHECK-LABEL: @test_svld2q_vnum_mf8(
539 // CHECK-NEXT: entry:
540 // CHECK-NEXT: [[TMP0:%.*]] = tail call i64 @llvm.vscale.i64()
541 // CHECK-NEXT: [[TMP1:%.*]] = shl nuw nsw i64 [[TMP0]], 4
542 // CHECK-NEXT: [[DOTIDX:%.*]] = mul i64 [[VNUM:%.*]], [[TMP1]]
543 // CHECK-NEXT: [[TMP2:%.*]] = getelementptr i8, ptr [[BASE:%.*]], i64 [[DOTIDX]]
544 // CHECK-NEXT: [[TMP3:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld2q.sret.nxv16i8(<vscale x 16 x i1> [[PG:%.*]], ptr [[TMP2]])
545 // CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP3]]
546 //
547 // CPP-CHECK-LABEL: @_Z20test_svld2q_vnum_mf8u10__SVBool_tPKu6__mfp8l(
548 // CPP-CHECK-NEXT: entry:
549 // CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call i64 @llvm.vscale.i64()
550 // CPP-CHECK-NEXT: [[TMP1:%.*]] = shl nuw nsw i64 [[TMP0]], 4
551 // CPP-CHECK-NEXT: [[DOTIDX:%.*]] = mul i64 [[VNUM:%.*]], [[TMP1]]
552 // CPP-CHECK-NEXT: [[TMP2:%.*]] = getelementptr i8, ptr [[BASE:%.*]], i64 [[DOTIDX]]
553 // CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld2q.sret.nxv16i8(<vscale x 16 x i1> [[PG:%.*]], ptr [[TMP2]])
554 // CPP-CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP3]]
555 //
556 svmfloat8x2_t test_svld2q_vnum_mf8(svbool_t pg, const mfloat8_t *base, int64_t vnum) ATTR
557 {
558 return SVE_ACLE_FUNC(svld2q_vnum,,_mf8,)(pg, base, vnum);
559 }
560
561 // CHECK-LABEL: @test_svld3q_u8(
562 // CHECK-NEXT: entry:
563 // CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld3q.sret.nxv16i8(<vscale x 16 x i1> [[PG:%.*]], ptr [[BASE:%.*]])
564 // CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]]
565 //
566 // CPP-CHECK-LABEL: @_Z14test_svld3q_u8u10__SVBool_tPKh(
567 // CPP-CHECK-NEXT: entry:
568 // CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld3q.sret.nxv16i8(<vscale x 16 x i1> [[PG:%.*]], ptr [[BASE:%.*]])
569 // CPP-CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]]
570 //
571 svuint8x3_t test_svld3q_u8(svbool_t pg, const uint8_t *base) ATTR
572 {
573 return SVE_ACLE_FUNC(svld3q,,_u8,)(pg, base);
574 }
575
576 // CHECK-LABEL: @test_svld3q_s8(
577 // CHECK-NEXT: entry:
578 // CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld3q.sret.nxv16i8(<vscale x 16 x i1> [[PG:%.*]], ptr [[BASE:%.*]])
579 // CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]]
580 //
581 // CPP-CHECK-LABEL: @_Z14test_svld3q_s8u10__SVBool_tPKa(
582 // CPP-CHECK-NEXT: entry:
583 // CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld3q.sret.nxv16i8(<vscale x 16 x i1> [[PG:%.*]], ptr [[BASE:%.*]])
584 // CPP-CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]]
585 //
586 svint8x3_t test_svld3q_s8(svbool_t pg, const int8_t *base) ATTR
587 {
588 return SVE_ACLE_FUNC(svld3q,,_s8,)(pg, base);
589 }
590
591 // CHECK-LABEL: @test_svld3q_u16(
592 // CHECK-NEXT: entry:
593 // CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
594 // CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.ld3q.sret.nxv8i16(<vscale x 8 x i1> [[TMP0]], ptr [[BASE:%.*]])
595 // CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]]
596 //
597 // CPP-CHECK-LABEL: @_Z15test_svld3q_u16u10__SVBool_tPKt(
598 // CPP-CHECK-NEXT: entry:
599 // CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
600 // CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.ld3q.sret.nxv8i16(<vscale x 8 x i1> [[TMP0]], ptr [[BASE:%.*]])
601 // CPP-CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]]
602 //
603 svuint16x3_t test_svld3q_u16(svbool_t pg, const uint16_t *base) ATTR
604 {
605 return SVE_ACLE_FUNC(svld3q,,_u16,)(pg, base);
606 }
607
608 // CHECK-LABEL: @test_svld3q_s16(
609 // CHECK-NEXT: entry:
610 // CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
611 // CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.ld3q.sret.nxv8i16(<vscale x 8 x i1> [[TMP0]], ptr [[BASE:%.*]])
612 // CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]]
613 //
614 // CPP-CHECK-LABEL: @_Z15test_svld3q_s16u10__SVBool_tPKs(
615 // CPP-CHECK-NEXT: entry:
616 // CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
617 // CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.ld3q.sret.nxv8i16(<vscale x 8 x i1> [[TMP0]], ptr [[BASE:%.*]])
618 // CPP-CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]]
619 //
620 svint16x3_t test_svld3q_s16(svbool_t pg, const int16_t *base) ATTR
621 {
622 return SVE_ACLE_FUNC(svld3q,,_s16,)(pg, base);
623 }
624
625 // CHECK-LABEL: @test_svld3q_u32(
626 // CHECK-NEXT: entry:
627 // CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
628 // CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.ld3q.sret.nxv4i32(<vscale x 4 x i1> [[TMP0]], ptr [[BASE:%.*]])
629 // CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]]
630 //
631 // CPP-CHECK-LABEL: @_Z15test_svld3q_u32u10__SVBool_tPKj(
632 // CPP-CHECK-NEXT: entry:
633 // CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
634 // CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.ld3q.sret.nxv4i32(<vscale x 4 x i1> [[TMP0]], ptr [[BASE:%.*]])
635 // CPP-CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]]
636 //
637 svuint32x3_t test_svld3q_u32(svbool_t pg, const uint32_t *base) ATTR
638 {
639 return SVE_ACLE_FUNC(svld3q,,_u32,)(pg, base);
640 }
641
642 // CHECK-LABEL: @test_svld3q_s32(
643 // CHECK-NEXT: entry:
644 // CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
645 // CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.ld3q.sret.nxv4i32(<vscale x 4 x i1> [[TMP0]], ptr [[BASE:%.*]])
646 // CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]]
647 //
648 // CPP-CHECK-LABEL: @_Z15test_svld3q_s32u10__SVBool_tPKi(
649 // CPP-CHECK-NEXT: entry:
650 // CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
651 // CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.ld3q.sret.nxv4i32(<vscale x 4 x i1> [[TMP0]], ptr [[BASE:%.*]])
652 // CPP-CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]]
653 //
654 svint32x3_t test_svld3q_s32(svbool_t pg, const int32_t *base) ATTR
655 {
656 return SVE_ACLE_FUNC(svld3q,,_s32,)(pg, base);
657 }
658
659 // CHECK-LABEL: @test_svld3q_u64(
660 // CHECK-NEXT: entry:
661 // CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
662 // CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.ld3q.sret.nxv2i64(<vscale x 2 x i1> [[TMP0]], ptr [[BASE:%.*]])
663 // CHECK-NEXT: ret { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]]
664 //
665 // CPP-CHECK-LABEL: @_Z15test_svld3q_u64u10__SVBool_tPKm(
666 // CPP-CHECK-NEXT: entry:
667 // CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
668 // CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.ld3q.sret.nxv2i64(<vscale x 2 x i1> [[TMP0]], ptr [[BASE:%.*]])
669 // CPP-CHECK-NEXT: ret { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]]
670 //
671 svuint64x3_t test_svld3q_u64(svbool_t pg, const uint64_t *base) ATTR
672 {
673 return SVE_ACLE_FUNC(svld3q,,_u64,)(pg, base);
674 }
675
676 // CHECK-LABEL: @test_svld3q_s64(
677 // CHECK-NEXT: entry:
678 // CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
679 // CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.ld3q.sret.nxv2i64(<vscale x 2 x i1> [[TMP0]], ptr [[BASE:%.*]])
680 // CHECK-NEXT: ret { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]]
681 //
682 // CPP-CHECK-LABEL: @_Z15test_svld3q_s64u10__SVBool_tPKl(
683 // CPP-CHECK-NEXT: entry:
684 // CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
685 // CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.ld3q.sret.nxv2i64(<vscale x 2 x i1> [[TMP0]], ptr [[BASE:%.*]])
686 // CPP-CHECK-NEXT: ret { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]]
687 //
688 svint64x3_t test_svld3q_s64(svbool_t pg, const int64_t *base) ATTR
689 {
690 return SVE_ACLE_FUNC(svld3q,,_s64,)(pg, base);
691 }
692
693 // CHECK-LABEL: @test_svld3q_f16(
694 // CHECK-NEXT: entry:
695 // CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
696 // CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sve.ld3q.sret.nxv8f16(<vscale x 8 x i1> [[TMP0]], ptr [[BASE:%.*]])
697 // CHECK-NEXT: ret { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP1]]
698 //
699 // CPP-CHECK-LABEL: @_Z15test_svld3q_f16u10__SVBool_tPKDh(
700 // CPP-CHECK-NEXT: entry:
701 // CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
702 // CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sve.ld3q.sret.nxv8f16(<vscale x 8 x i1> [[TMP0]], ptr [[BASE:%.*]])
703 // CPP-CHECK-NEXT: ret { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP1]]
704 //
705 svfloat16x3_t test_svld3q_f16(svbool_t pg, const float16_t *base) ATTR
706 {
707 return SVE_ACLE_FUNC(svld3q,,_f16,)(pg, base);
708 }
709
710 // CHECK-LABEL: @test_svld3q_bf16(
711 // CHECK-NEXT: entry:
712 // CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
713 // CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } @llvm.aarch64.sve.ld3q.sret.nxv8bf16(<vscale x 8 x i1> [[TMP0]], ptr [[BASE:%.*]])
714 // CHECK-NEXT: ret { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP1]]
715 //
716 // CPP-CHECK-LABEL: @_Z16test_svld3q_bf16u10__SVBool_tPKu6__bf16(
717 // CPP-CHECK-NEXT: entry:
718 // CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
719 // CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } @llvm.aarch64.sve.ld3q.sret.nxv8bf16(<vscale x 8 x i1> [[TMP0]], ptr [[BASE:%.*]])
720 // CPP-CHECK-NEXT: ret { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP1]]
721 //
722 svbfloat16x3_t test_svld3q_bf16(svbool_t pg, const bfloat16_t *base) ATTR
723 {
724 return SVE_ACLE_FUNC(svld3q,,_bf16,)(pg, base);
725 }
726
727 // CHECK-LABEL: @test_svld3q_f32(
728 // CHECK-NEXT: entry:
729 // CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
730 // CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.ld3q.sret.nxv4f32(<vscale x 4 x i1> [[TMP0]], ptr [[BASE:%.*]])
731 // CHECK-NEXT: ret { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP1]]
732 //
733 // CPP-CHECK-LABEL: @_Z15test_svld3q_f32u10__SVBool_tPKf(
734 // CPP-CHECK-NEXT: entry:
735 // CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
736 // CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.ld3q.sret.nxv4f32(<vscale x 4 x i1> [[TMP0]], ptr [[BASE:%.*]])
737 // CPP-CHECK-NEXT: ret { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP1]]
738 //
739 svfloat32x3_t test_svld3q_f32(svbool_t pg, const float32_t *base) ATTR
740 {
741 return SVE_ACLE_FUNC(svld3q,,_f32,)(pg, base);
742 }
743
744 // CHECK-LABEL: @test_svld3q_f64(
745 // CHECK-NEXT: entry:
746 // CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
747 // CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } @llvm.aarch64.sve.ld3q.sret.nxv2f64(<vscale x 2 x i1> [[TMP0]], ptr [[BASE:%.*]])
748 // CHECK-NEXT: ret { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP1]]
749 //
750 // CPP-CHECK-LABEL: @_Z15test_svld3q_f64u10__SVBool_tPKd(
751 // CPP-CHECK-NEXT: entry:
752 // CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
753 // CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } @llvm.aarch64.sve.ld3q.sret.nxv2f64(<vscale x 2 x i1> [[TMP0]], ptr [[BASE:%.*]])
754 // CPP-CHECK-NEXT: ret { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP1]]
755 //
756 svfloat64x3_t test_svld3q_f64(svbool_t pg, const float64_t *base) ATTR
757 {
758 return SVE_ACLE_FUNC(svld3q,,_f64,)(pg, base);
759 }
760
761 // CHECK-LABEL: @test_svld3q_mf8(
762 // CHECK-NEXT: entry:
763 // CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld3q.sret.nxv16i8(<vscale x 16 x i1> [[PG:%.*]], ptr [[BASE:%.*]])
764 // CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]]
765 //
766 // CPP-CHECK-LABEL: @_Z15test_svld3q_mf8u10__SVBool_tPKu6__mfp8(
767 // CPP-CHECK-NEXT: entry:
768 // CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld3q.sret.nxv16i8(<vscale x 16 x i1> [[PG:%.*]], ptr [[BASE:%.*]])
769 // CPP-CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]]
770 //
771 svmfloat8x3_t test_svld3q_mf8(svbool_t pg, const mfloat8_t *base) ATTR
772 {
773 return SVE_ACLE_FUNC(svld3q,,_mf8,)(pg, base);
774 }
775
776 // CHECK-LABEL: @test_svld3q_vnum_u8(
777 // CHECK-NEXT: entry:
778 // CHECK-NEXT: [[TMP0:%.*]] = tail call i64 @llvm.vscale.i64()
779 // CHECK-NEXT: [[TMP1:%.*]] = shl nuw nsw i64 [[TMP0]], 4
780 // CHECK-NEXT: [[DOTIDX:%.*]] = mul i64 [[VNUM:%.*]], [[TMP1]]
781 // CHECK-NEXT: [[TMP2:%.*]] = getelementptr i8, ptr [[BASE:%.*]], i64 [[DOTIDX]]
782 // CHECK-NEXT: [[TMP3:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld3q.sret.nxv16i8(<vscale x 16 x i1> [[PG:%.*]], ptr [[TMP2]])
783 // CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP3]]
784 //
785 // CPP-CHECK-LABEL: @_Z19test_svld3q_vnum_u8u10__SVBool_tPKhl(
786 // CPP-CHECK-NEXT: entry:
787 // CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call i64 @llvm.vscale.i64()
788 // CPP-CHECK-NEXT: [[TMP1:%.*]] = shl nuw nsw i64 [[TMP0]], 4
789 // CPP-CHECK-NEXT: [[DOTIDX:%.*]] = mul i64 [[VNUM:%.*]], [[TMP1]]
790 // CPP-CHECK-NEXT: [[TMP2:%.*]] = getelementptr i8, ptr [[BASE:%.*]], i64 [[DOTIDX]]
791 // CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld3q.sret.nxv16i8(<vscale x 16 x i1> [[PG:%.*]], ptr [[TMP2]])
792 // CPP-CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP3]]
793 //
794 svuint8x3_t test_svld3q_vnum_u8(svbool_t pg, const uint8_t *base, int64_t vnum) ATTR
795 {
796 return SVE_ACLE_FUNC(svld3q_vnum_,,u8,)(pg, base, vnum);
797 }
798
799 // CHECK-LABEL: @test_svld3q_vnum_s8(
800 // CHECK-NEXT: entry:
801 // CHECK-NEXT: [[TMP0:%.*]] = tail call i64 @llvm.vscale.i64()
802 // CHECK-NEXT: [[TMP1:%.*]] = shl nuw nsw i64 [[TMP0]], 4
803 // CHECK-NEXT: [[DOTIDX:%.*]] = mul i64 [[VNUM:%.*]], [[TMP1]]
804 // CHECK-NEXT: [[TMP2:%.*]] = getelementptr i8, ptr [[BASE:%.*]], i64 [[DOTIDX]]
805 // CHECK-NEXT: [[TMP3:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld3q.sret.nxv16i8(<vscale x 16 x i1> [[PG:%.*]], ptr [[TMP2]])
806 // CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP3]]
807 //
808 // CPP-CHECK-LABEL: @_Z19test_svld3q_vnum_s8u10__SVBool_tPKal(
809 // CPP-CHECK-NEXT: entry:
810 // CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call i64 @llvm.vscale.i64()
811 // CPP-CHECK-NEXT: [[TMP1:%.*]] = shl nuw nsw i64 [[TMP0]], 4
812 // CPP-CHECK-NEXT: [[DOTIDX:%.*]] = mul i64 [[VNUM:%.*]], [[TMP1]]
813 // CPP-CHECK-NEXT: [[TMP2:%.*]] = getelementptr i8, ptr [[BASE:%.*]], i64 [[DOTIDX]]
814 // CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld3q.sret.nxv16i8(<vscale x 16 x i1> [[PG:%.*]], ptr [[TMP2]])
815 // CPP-CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP3]]
816 //
817 svint8x3_t test_svld3q_vnum_s8(svbool_t pg, const int8_t *base, int64_t vnum) ATTR
818 {
819 return SVE_ACLE_FUNC(svld3q_vnum,,_s8,)(pg, base, vnum);
820 }
821
822 // CHECK-LABEL: @test_svld3q_vnum_u16(
823 // CHECK-NEXT: entry:
824 // CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
825 // CHECK-NEXT: [[TMP1:%.*]] = tail call i64 @llvm.vscale.i64()
826 // CHECK-NEXT: [[TMP2:%.*]] = shl nuw nsw i64 [[TMP1]], 4
827 // CHECK-NEXT: [[DOTIDX:%.*]] = mul i64 [[VNUM:%.*]], [[TMP2]]
828 // CHECK-NEXT: [[TMP3:%.*]] = getelementptr i8, ptr [[BASE:%.*]], i64 [[DOTIDX]]
829 // CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.ld3q.sret.nxv8i16(<vscale x 8 x i1> [[TMP0]], ptr [[TMP3]])
830 // CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP4]]
831 //
832 // CPP-CHECK-LABEL: @_Z20test_svld3q_vnum_u16u10__SVBool_tPKtl(
833 // CPP-CHECK-NEXT: entry:
834 // CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
835 // CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call i64 @llvm.vscale.i64()
836 // CPP-CHECK-NEXT: [[TMP2:%.*]] = shl nuw nsw i64 [[TMP1]], 4
837 // CPP-CHECK-NEXT: [[DOTIDX:%.*]] = mul i64 [[VNUM:%.*]], [[TMP2]]
838 // CPP-CHECK-NEXT: [[TMP3:%.*]] = getelementptr i8, ptr [[BASE:%.*]], i64 [[DOTIDX]]
839 // CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.ld3q.sret.nxv8i16(<vscale x 8 x i1> [[TMP0]], ptr [[TMP3]])
840 // CPP-CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP4]]
841 //
842 svuint16x3_t test_svld3q_vnum_u16(svbool_t pg, const uint16_t *base, int64_t vnum) ATTR
843 {
844 return SVE_ACLE_FUNC(svld3q_vnum,,_u16,)(pg, base, vnum);
845 }
846
847 // CHECK-LABEL: @test_svld3q_vnum_s16(
848 // CHECK-NEXT: entry:
849 // CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
850 // CHECK-NEXT: [[TMP1:%.*]] = tail call i64 @llvm.vscale.i64()
851 // CHECK-NEXT: [[TMP2:%.*]] = shl nuw nsw i64 [[TMP1]], 4
852 // CHECK-NEXT: [[DOTIDX:%.*]] = mul i64 [[VNUM:%.*]], [[TMP2]]
853 // CHECK-NEXT: [[TMP3:%.*]] = getelementptr i8, ptr [[BASE:%.*]], i64 [[DOTIDX]]
854 // CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.ld3q.sret.nxv8i16(<vscale x 8 x i1> [[TMP0]], ptr [[TMP3]])
855 // CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP4]]
856 //
857 // CPP-CHECK-LABEL: @_Z20test_svld3q_vnum_s16u10__SVBool_tPKsl(
858 // CPP-CHECK-NEXT: entry:
859 // CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
860 // CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call i64 @llvm.vscale.i64()
861 // CPP-CHECK-NEXT: [[TMP2:%.*]] = shl nuw nsw i64 [[TMP1]], 4
862 // CPP-CHECK-NEXT: [[DOTIDX:%.*]] = mul i64 [[VNUM:%.*]], [[TMP2]]
863 // CPP-CHECK-NEXT: [[TMP3:%.*]] = getelementptr i8, ptr [[BASE:%.*]], i64 [[DOTIDX]]
864 // CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.ld3q.sret.nxv8i16(<vscale x 8 x i1> [[TMP0]], ptr [[TMP3]])
865 // CPP-CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP4]]
866 //
867 svint16x3_t test_svld3q_vnum_s16(svbool_t pg, const int16_t *base, int64_t vnum) ATTR
868 {
869 return SVE_ACLE_FUNC(svld3q_vnum,,_s16,)(pg, base, vnum);
870 }
871
872 // CHECK-LABEL: @test_svld3q_vnum_u32(
873 // CHECK-NEXT: entry:
874 // CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
875 // CHECK-NEXT: [[TMP1:%.*]] = tail call i64 @llvm.vscale.i64()
876 // CHECK-NEXT: [[TMP2:%.*]] = shl nuw nsw i64 [[TMP1]], 4
877 // CHECK-NEXT: [[DOTIDX:%.*]] = mul i64 [[VNUM:%.*]], [[TMP2]]
878 // CHECK-NEXT: [[TMP3:%.*]] = getelementptr i8, ptr [[BASE:%.*]], i64 [[DOTIDX]]
879 // CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.ld3q.sret.nxv4i32(<vscale x 4 x i1> [[TMP0]], ptr [[TMP3]])
880 // CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP4]]
881 //
882 // CPP-CHECK-LABEL: @_Z20test_svld3q_vnum_u32u10__SVBool_tPKjl(
883 // CPP-CHECK-NEXT: entry:
884 // CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
885 // CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call i64 @llvm.vscale.i64()
886 // CPP-CHECK-NEXT: [[TMP2:%.*]] = shl nuw nsw i64 [[TMP1]], 4
887 // CPP-CHECK-NEXT: [[DOTIDX:%.*]] = mul i64 [[VNUM:%.*]], [[TMP2]]
888 // CPP-CHECK-NEXT: [[TMP3:%.*]] = getelementptr i8, ptr [[BASE:%.*]], i64 [[DOTIDX]]
889 // CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.ld3q.sret.nxv4i32(<vscale x 4 x i1> [[TMP0]], ptr [[TMP3]])
890 // CPP-CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP4]]
891 //
892 svuint32x3_t test_svld3q_vnum_u32(svbool_t pg, const uint32_t *base, int64_t vnum) ATTR
893 {
894 return SVE_ACLE_FUNC(svld3q_vnum,,_u32,)(pg, base, vnum);
895 }
896
897 // CHECK-LABEL: @test_svld3q_vnum_s32(
898 // CHECK-NEXT: entry:
899 // CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
900 // CHECK-NEXT: [[TMP1:%.*]] = tail call i64 @llvm.vscale.i64()
901 // CHECK-NEXT: [[TMP2:%.*]] = shl nuw nsw i64 [[TMP1]], 4
902 // CHECK-NEXT: [[DOTIDX:%.*]] = mul i64 [[VNUM:%.*]], [[TMP2]]
903 // CHECK-NEXT: [[TMP3:%.*]] = getelementptr i8, ptr [[BASE:%.*]], i64 [[DOTIDX]]
904 // CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.ld3q.sret.nxv4i32(<vscale x 4 x i1> [[TMP0]], ptr [[TMP3]])
905 // CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP4]]
906 //
907 // CPP-CHECK-LABEL: @_Z20test_svld3q_vnum_s32u10__SVBool_tPKil(
908 // CPP-CHECK-NEXT: entry:
909 // CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
910 // CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call i64 @llvm.vscale.i64()
911 // CPP-CHECK-NEXT: [[TMP2:%.*]] = shl nuw nsw i64 [[TMP1]], 4
912 // CPP-CHECK-NEXT: [[DOTIDX:%.*]] = mul i64 [[VNUM:%.*]], [[TMP2]]
913 // CPP-CHECK-NEXT: [[TMP3:%.*]] = getelementptr i8, ptr [[BASE:%.*]], i64 [[DOTIDX]]
914 // CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.ld3q.sret.nxv4i32(<vscale x 4 x i1> [[TMP0]], ptr [[TMP3]])
915 // CPP-CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP4]]
916 //
917 svint32x3_t test_svld3q_vnum_s32(svbool_t pg, const int32_t *base, int64_t vnum) ATTR
918 {
919 return SVE_ACLE_FUNC(svld3q_vnum,,_s32,)(pg, base, vnum);
920 }
921
922 // CHECK-LABEL: @test_svld3q_vnum_u64(
923 // CHECK-NEXT: entry:
924 // CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
925 // CHECK-NEXT: [[TMP1:%.*]] = tail call i64 @llvm.vscale.i64()
926 // CHECK-NEXT: [[TMP2:%.*]] = shl nuw nsw i64 [[TMP1]], 4
927 // CHECK-NEXT: [[DOTIDX:%.*]] = mul i64 [[VNUM:%.*]], [[TMP2]]
928 // CHECK-NEXT: [[TMP3:%.*]] = getelementptr i8, ptr [[BASE:%.*]], i64 [[DOTIDX]]
929 // CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.ld3q.sret.nxv2i64(<vscale x 2 x i1> [[TMP0]], ptr [[TMP3]])
930 // CHECK-NEXT: ret { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP4]]
931 //
932 // CPP-CHECK-LABEL: @_Z20test_svld3q_vnum_u64u10__SVBool_tPKml(
933 // CPP-CHECK-NEXT: entry:
934 // CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
935 // CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call i64 @llvm.vscale.i64()
936 // CPP-CHECK-NEXT: [[TMP2:%.*]] = shl nuw nsw i64 [[TMP1]], 4
937 // CPP-CHECK-NEXT: [[DOTIDX:%.*]] = mul i64 [[VNUM:%.*]], [[TMP2]]
938 // CPP-CHECK-NEXT: [[TMP3:%.*]] = getelementptr i8, ptr [[BASE:%.*]], i64 [[DOTIDX]]
939 // CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.ld3q.sret.nxv2i64(<vscale x 2 x i1> [[TMP0]], ptr [[TMP3]])
940 // CPP-CHECK-NEXT: ret { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP4]]
941 //
942 svuint64x3_t test_svld3q_vnum_u64(svbool_t pg, const uint64_t *base, int64_t vnum) ATTR
943 {
944 return SVE_ACLE_FUNC(svld3q_vnum,,_u64,)(pg, base, vnum);
945 }
946
947 // CHECK-LABEL: @test_svld3q_vnum_s64(
948 // CHECK-NEXT: entry:
949 // CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
950 // CHECK-NEXT: [[TMP1:%.*]] = tail call i64 @llvm.vscale.i64()
951 // CHECK-NEXT: [[TMP2:%.*]] = shl nuw nsw i64 [[TMP1]], 4
952 // CHECK-NEXT: [[DOTIDX:%.*]] = mul i64 [[VNUM:%.*]], [[TMP2]]
953 // CHECK-NEXT: [[TMP3:%.*]] = getelementptr i8, ptr [[BASE:%.*]], i64 [[DOTIDX]]
954 // CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.ld3q.sret.nxv2i64(<vscale x 2 x i1> [[TMP0]], ptr [[TMP3]])
955 // CHECK-NEXT: ret { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP4]]
956 //
957 // CPP-CHECK-LABEL: @_Z20test_svld3q_vnum_s64u10__SVBool_tPKll(
958 // CPP-CHECK-NEXT: entry:
959 // CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
960 // CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call i64 @llvm.vscale.i64()
961 // CPP-CHECK-NEXT: [[TMP2:%.*]] = shl nuw nsw i64 [[TMP1]], 4
962 // CPP-CHECK-NEXT: [[DOTIDX:%.*]] = mul i64 [[VNUM:%.*]], [[TMP2]]
963 // CPP-CHECK-NEXT: [[TMP3:%.*]] = getelementptr i8, ptr [[BASE:%.*]], i64 [[DOTIDX]]
964 // CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.ld3q.sret.nxv2i64(<vscale x 2 x i1> [[TMP0]], ptr [[TMP3]])
965 // CPP-CHECK-NEXT: ret { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP4]]
966 //
967 svint64x3_t test_svld3q_vnum_s64(svbool_t pg, const int64_t *base, int64_t vnum) ATTR
968 {
969 return SVE_ACLE_FUNC(svld3q_vnum,,_s64,)(pg, base, vnum);
970 }
971
972 // CHECK-LABEL: @test_svld3q_vnum_f16(
973 // CHECK-NEXT: entry:
974 // CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
975 // CHECK-NEXT: [[TMP1:%.*]] = tail call i64 @llvm.vscale.i64()
976 // CHECK-NEXT: [[TMP2:%.*]] = shl nuw nsw i64 [[TMP1]], 4
977 // CHECK-NEXT: [[DOTIDX:%.*]] = mul i64 [[VNUM:%.*]], [[TMP2]]
978 // CHECK-NEXT: [[TMP3:%.*]] = getelementptr i8, ptr [[BASE:%.*]], i64 [[DOTIDX]]
979 // CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sve.ld3q.sret.nxv8f16(<vscale x 8 x i1> [[TMP0]], ptr [[TMP3]])
980 // CHECK-NEXT: ret { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP4]]
981 //
982 // CPP-CHECK-LABEL: @_Z20test_svld3q_vnum_f16u10__SVBool_tPKDhl(
983 // CPP-CHECK-NEXT: entry:
984 // CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
985 // CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call i64 @llvm.vscale.i64()
986 // CPP-CHECK-NEXT: [[TMP2:%.*]] = shl nuw nsw i64 [[TMP1]], 4
987 // CPP-CHECK-NEXT: [[DOTIDX:%.*]] = mul i64 [[VNUM:%.*]], [[TMP2]]
988 // CPP-CHECK-NEXT: [[TMP3:%.*]] = getelementptr i8, ptr [[BASE:%.*]], i64 [[DOTIDX]]
989 // CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sve.ld3q.sret.nxv8f16(<vscale x 8 x i1> [[TMP0]], ptr [[TMP3]])
990 // CPP-CHECK-NEXT: ret { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP4]]
991 //
992 svfloat16x3_t test_svld3q_vnum_f16(svbool_t pg, const float16_t *base, int64_t vnum) ATTR
993 {
994 return SVE_ACLE_FUNC(svld3q_vnum,,_f16,)(pg, base, vnum);
995 }
996
997 // CHECK-LABEL: @test_svld3q_vnum_bf16(
998 // CHECK-NEXT: entry:
999 // CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
1000 // CHECK-NEXT: [[TMP1:%.*]] = tail call i64 @llvm.vscale.i64()
1001 // CHECK-NEXT: [[TMP2:%.*]] = shl nuw nsw i64 [[TMP1]], 4
1002 // CHECK-NEXT: [[DOTIDX:%.*]] = mul i64 [[VNUM:%.*]], [[TMP2]]
1003 // CHECK-NEXT: [[TMP3:%.*]] = getelementptr i8, ptr [[BASE:%.*]], i64 [[DOTIDX]]
1004 // CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } @llvm.aarch64.sve.ld3q.sret.nxv8bf16(<vscale x 8 x i1> [[TMP0]], ptr [[TMP3]])
1005 // CHECK-NEXT: ret { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP4]]
1006 //
1007 // CPP-CHECK-LABEL: @_Z21test_svld3q_vnum_bf16u10__SVBool_tPKu6__bf16l(
1008 // CPP-CHECK-NEXT: entry:
1009 // CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
1010 // CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call i64 @llvm.vscale.i64()
1011 // CPP-CHECK-NEXT: [[TMP2:%.*]] = shl nuw nsw i64 [[TMP1]], 4
1012 // CPP-CHECK-NEXT: [[DOTIDX:%.*]] = mul i64 [[VNUM:%.*]], [[TMP2]]
1013 // CPP-CHECK-NEXT: [[TMP3:%.*]] = getelementptr i8, ptr [[BASE:%.*]], i64 [[DOTIDX]]
1014 // CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } @llvm.aarch64.sve.ld3q.sret.nxv8bf16(<vscale x 8 x i1> [[TMP0]], ptr [[TMP3]])
1015 // CPP-CHECK-NEXT: ret { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP4]]
1016 //
1017 svbfloat16x3_t test_svld3q_vnum_bf16(svbool_t pg, const bfloat16_t *base, int64_t vnum) ATTR
1018 {
1019 return SVE_ACLE_FUNC(svld3q_vnum,,_bf16,)(pg, base, vnum);
1020 }
1021
1022 // CHECK-LABEL: @test_svld3q_vnum_f32(
1023 // CHECK-NEXT: entry:
1024 // CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
1025 // CHECK-NEXT: [[TMP1:%.*]] = tail call i64 @llvm.vscale.i64()
1026 // CHECK-NEXT: [[TMP2:%.*]] = shl nuw nsw i64 [[TMP1]], 4
1027 // CHECK-NEXT: [[DOTIDX:%.*]] = mul i64 [[VNUM:%.*]], [[TMP2]]
1028 // CHECK-NEXT: [[TMP3:%.*]] = getelementptr i8, ptr [[BASE:%.*]], i64 [[DOTIDX]]
1029 // CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.ld3q.sret.nxv4f32(<vscale x 4 x i1> [[TMP0]], ptr [[TMP3]])
1030 // CHECK-NEXT: ret { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP4]]
1031 //
1032 // CPP-CHECK-LABEL: @_Z20test_svld3q_vnum_f32u10__SVBool_tPKfl(
1033 // CPP-CHECK-NEXT: entry:
1034 // CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
1035 // CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call i64 @llvm.vscale.i64()
1036 // CPP-CHECK-NEXT: [[TMP2:%.*]] = shl nuw nsw i64 [[TMP1]], 4
1037 // CPP-CHECK-NEXT: [[DOTIDX:%.*]] = mul i64 [[VNUM:%.*]], [[TMP2]]
1038 // CPP-CHECK-NEXT: [[TMP3:%.*]] = getelementptr i8, ptr [[BASE:%.*]], i64 [[DOTIDX]]
1039 // CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.ld3q.sret.nxv4f32(<vscale x 4 x i1> [[TMP0]], ptr [[TMP3]])
1040 // CPP-CHECK-NEXT: ret { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP4]]
1041 //
1042 svfloat32x3_t test_svld3q_vnum_f32(svbool_t pg, const float32_t *base, int64_t vnum) ATTR
1043 {
1044 return SVE_ACLE_FUNC(svld3q_vnum,,_f32,)(pg, base, vnum);
1045 }
1046
1047 // CHECK-LABEL: @test_svld3q_vnum_f64(
1048 // CHECK-NEXT: entry:
1049 // CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
1050 // CHECK-NEXT: [[TMP1:%.*]] = tail call i64 @llvm.vscale.i64()
1051 // CHECK-NEXT: [[TMP2:%.*]] = shl nuw nsw i64 [[TMP1]], 4
1052 // CHECK-NEXT: [[DOTIDX:%.*]] = mul i64 [[VNUM:%.*]], [[TMP2]]
1053 // CHECK-NEXT: [[TMP3:%.*]] = getelementptr i8, ptr [[BASE:%.*]], i64 [[DOTIDX]]
1054 // CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } @llvm.aarch64.sve.ld3q.sret.nxv2f64(<vscale x 2 x i1> [[TMP0]], ptr [[TMP3]])
1055 // CHECK-NEXT: ret { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP4]]
1056 //
1057 // CPP-CHECK-LABEL: @_Z20test_svld3q_vnum_f64u10__SVBool_tPKdl(
1058 // CPP-CHECK-NEXT: entry:
1059 // CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
1060 // CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call i64 @llvm.vscale.i64()
1061 // CPP-CHECK-NEXT: [[TMP2:%.*]] = shl nuw nsw i64 [[TMP1]], 4
1062 // CPP-CHECK-NEXT: [[DOTIDX:%.*]] = mul i64 [[VNUM:%.*]], [[TMP2]]
1063 // CPP-CHECK-NEXT: [[TMP3:%.*]] = getelementptr i8, ptr [[BASE:%.*]], i64 [[DOTIDX]]
1064 // CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } @llvm.aarch64.sve.ld3q.sret.nxv2f64(<vscale x 2 x i1> [[TMP0]], ptr [[TMP3]])
1065 // CPP-CHECK-NEXT: ret { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP4]]
1066 //
1067 svfloat64x3_t test_svld3q_vnum_f64(svbool_t pg, const float64_t *base, int64_t vnum) ATTR
1068 {
1069 return SVE_ACLE_FUNC(svld3q_vnum,,_f64,)(pg, base, vnum);
1070 }
1071
1072 // CHECK-LABEL: @test_svld3q_vnum_mf8(
1073 // CHECK-NEXT: entry:
1074 // CHECK-NEXT: [[TMP0:%.*]] = tail call i64 @llvm.vscale.i64()
1075 // CHECK-NEXT: [[TMP1:%.*]] = shl nuw nsw i64 [[TMP0]], 4
1076 // CHECK-NEXT: [[DOTIDX:%.*]] = mul i64 [[VNUM:%.*]], [[TMP1]]
1077 // CHECK-NEXT: [[TMP2:%.*]] = getelementptr i8, ptr [[BASE:%.*]], i64 [[DOTIDX]]
1078 // CHECK-NEXT: [[TMP3:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld3q.sret.nxv16i8(<vscale x 16 x i1> [[PG:%.*]], ptr [[TMP2]])
1079 // CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP3]]
1080 //
1081 // CPP-CHECK-LABEL: @_Z20test_svld3q_vnum_mf8u10__SVBool_tPKu6__mfp8l(
1082 // CPP-CHECK-NEXT: entry:
1083 // CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call i64 @llvm.vscale.i64()
1084 // CPP-CHECK-NEXT: [[TMP1:%.*]] = shl nuw nsw i64 [[TMP0]], 4
1085 // CPP-CHECK-NEXT: [[DOTIDX:%.*]] = mul i64 [[VNUM:%.*]], [[TMP1]]
1086 // CPP-CHECK-NEXT: [[TMP2:%.*]] = getelementptr i8, ptr [[BASE:%.*]], i64 [[DOTIDX]]
1087 // CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld3q.sret.nxv16i8(<vscale x 16 x i1> [[PG:%.*]], ptr [[TMP2]])
1088 // CPP-CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP3]]
1089 //
1090 svmfloat8x3_t test_svld3q_vnum_mf8(svbool_t pg, const mfloat8_t *base, int64_t vnum) ATTR
1091 {
1092 return SVE_ACLE_FUNC(svld3q_vnum,,_mf8,)(pg, base, vnum);
1093 }
1094
1095 // CHECK-LABEL: @test_svld4q_u8(
1096 // CHECK-NEXT: entry:
1097 // CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld4q.sret.nxv16i8(<vscale x 16 x i1> [[PG:%.*]], ptr [[BASE:%.*]])
1098 // CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]]
1099 //
1100 // CPP-CHECK-LABEL: @_Z14test_svld4q_u8u10__SVBool_tPKh(
1101 // CPP-CHECK-NEXT: entry:
1102 // CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld4q.sret.nxv16i8(<vscale x 16 x i1> [[PG:%.*]], ptr [[BASE:%.*]])
1103 // CPP-CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]]
1104 //
1105 svuint8x4_t test_svld4q_u8(svbool_t pg, const uint8_t *base) ATTR
1106 {
1107 return SVE_ACLE_FUNC(svld4q,,_u8,)(pg, base);
1108 }
1109
1110 // CHECK-LABEL: @test_svld4q_u16(
1111 // CHECK-NEXT: entry:
1112 // CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
1113 // CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.ld4q.sret.nxv8i16(<vscale x 8 x i1> [[TMP0]], ptr [[BASE:%.*]])
1114 // CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]]
1115 //
1116 // CPP-CHECK-LABEL: @_Z15test_svld4q_u16u10__SVBool_tPKt(
1117 // CPP-CHECK-NEXT: entry:
1118 // CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
1119 // CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.ld4q.sret.nxv8i16(<vscale x 8 x i1> [[TMP0]], ptr [[BASE:%.*]])
1120 // CPP-CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]]
1121 //
1122 svuint16x4_t test_svld4q_u16(svbool_t pg, const uint16_t *base) ATTR
1123 {
1124 return SVE_ACLE_FUNC(svld4q,,_u16,)(pg, base);
1125 }
1126
1127 // CHECK-LABEL: @test_svld4q_s16(
1128 // CHECK-NEXT: entry:
1129 // CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
1130 // CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.ld4q.sret.nxv8i16(<vscale x 8 x i1> [[TMP0]], ptr [[BASE:%.*]])
1131 // CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]]
1132 //
1133 // CPP-CHECK-LABEL: @_Z15test_svld4q_s16u10__SVBool_tPKs(
1134 // CPP-CHECK-NEXT: entry:
1135 // CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
1136 // CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.ld4q.sret.nxv8i16(<vscale x 8 x i1> [[TMP0]], ptr [[BASE:%.*]])
1137 // CPP-CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP1]]
1138 //
1139 svint16x4_t test_svld4q_s16(svbool_t pg, const int16_t *base) ATTR
1140 {
1141 return SVE_ACLE_FUNC(svld4q,,_s16,)(pg, base);
1142 }
1143
1144 // CHECK-LABEL: @test_svld4q_u32(
1145 // CHECK-NEXT: entry:
1146 // CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
1147 // CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.ld4q.sret.nxv4i32(<vscale x 4 x i1> [[TMP0]], ptr [[BASE:%.*]])
1148 // CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]]
1149 //
1150 // CPP-CHECK-LABEL: @_Z15test_svld4q_u32u10__SVBool_tPKj(
1151 // CPP-CHECK-NEXT: entry:
1152 // CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
1153 // CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.ld4q.sret.nxv4i32(<vscale x 4 x i1> [[TMP0]], ptr [[BASE:%.*]])
1154 // CPP-CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]]
1155 //
1156 svuint32x4_t test_svld4q_u32(svbool_t pg, const uint32_t *base) ATTR
1157 {
1158 return SVE_ACLE_FUNC(svld4q,,_u32,)(pg, base);
1159 }
1160
1161 // CHECK-LABEL: @test_svld4q_s32(
1162 // CHECK-NEXT: entry:
1163 // CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
1164 // CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.ld4q.sret.nxv4i32(<vscale x 4 x i1> [[TMP0]], ptr [[BASE:%.*]])
1165 // CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]]
1166 //
1167 // CPP-CHECK-LABEL: @_Z15test_svld4q_s32u10__SVBool_tPKi(
1168 // CPP-CHECK-NEXT: entry:
1169 // CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
1170 // CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.ld4q.sret.nxv4i32(<vscale x 4 x i1> [[TMP0]], ptr [[BASE:%.*]])
1171 // CPP-CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP1]]
1172 //
1173 svint32x4_t test_svld4q_s32(svbool_t pg, const int32_t *base) ATTR
1174 {
1175 return SVE_ACLE_FUNC(svld4q,,_s32,)(pg, base);
1176 }
1177
1178 // CHECK-LABEL: @test_svld4q_u64(
1179 // CHECK-NEXT: entry:
1180 // CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
1181 // CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.ld4q.sret.nxv2i64(<vscale x 2 x i1> [[TMP0]], ptr [[BASE:%.*]])
1182 // CHECK-NEXT: ret { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]]
1183 //
1184 // CPP-CHECK-LABEL: @_Z15test_svld4q_u64u10__SVBool_tPKm(
1185 // CPP-CHECK-NEXT: entry:
1186 // CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
1187 // CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.ld4q.sret.nxv2i64(<vscale x 2 x i1> [[TMP0]], ptr [[BASE:%.*]])
1188 // CPP-CHECK-NEXT: ret { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]]
1189 //
1190 svuint64x4_t test_svld4q_u64(svbool_t pg, const uint64_t *base) ATTR
1191 {
1192 return SVE_ACLE_FUNC(svld4q,,_u64,)(pg, base);
1193 }
1194
1195 // CHECK-LABEL: @test_svld4q_s64(
1196 // CHECK-NEXT: entry:
1197 // CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
1198 // CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.ld4q.sret.nxv2i64(<vscale x 2 x i1> [[TMP0]], ptr [[BASE:%.*]])
1199 // CHECK-NEXT: ret { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]]
1200 //
1201 // CPP-CHECK-LABEL: @_Z15test_svld4q_s64u10__SVBool_tPKl(
1202 // CPP-CHECK-NEXT: entry:
1203 // CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
1204 // CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.ld4q.sret.nxv2i64(<vscale x 2 x i1> [[TMP0]], ptr [[BASE:%.*]])
1205 // CPP-CHECK-NEXT: ret { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP1]]
1206 //
1207 svint64x4_t test_svld4q_s64(svbool_t pg, const int64_t *base) ATTR
1208 {
1209 return SVE_ACLE_FUNC(svld4q,,_s64,)(pg, base);
1210 }
1211
1212 // CHECK-LABEL: @test_svld4q_f16(
1213 // CHECK-NEXT: entry:
1214 // CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
1215 // CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sve.ld4q.sret.nxv8f16(<vscale x 8 x i1> [[TMP0]], ptr [[BASE:%.*]])
1216 // CHECK-NEXT: ret { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP1]]
1217 //
1218 // CPP-CHECK-LABEL: @_Z15test_svld4q_f16u10__SVBool_tPKDh(
1219 // CPP-CHECK-NEXT: entry:
1220 // CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
1221 // CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sve.ld4q.sret.nxv8f16(<vscale x 8 x i1> [[TMP0]], ptr [[BASE:%.*]])
1222 // CPP-CHECK-NEXT: ret { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP1]]
1223 //
1224 svfloat16x4_t test_svld4q_f16(svbool_t pg, const float16_t *base) ATTR
1225 {
1226 return SVE_ACLE_FUNC(svld4q,,_f16,)(pg, base);
1227 }
1228
1229 // CHECK-LABEL: @test_svld4q_bf16(
1230 // CHECK-NEXT: entry:
1231 // CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
1232 // CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } @llvm.aarch64.sve.ld4q.sret.nxv8bf16(<vscale x 8 x i1> [[TMP0]], ptr [[BASE:%.*]])
1233 // CHECK-NEXT: ret { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP1]]
1234 //
1235 // CPP-CHECK-LABEL: @_Z16test_svld4q_bf16u10__SVBool_tPKu6__bf16(
1236 // CPP-CHECK-NEXT: entry:
1237 // CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
1238 // CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } @llvm.aarch64.sve.ld4q.sret.nxv8bf16(<vscale x 8 x i1> [[TMP0]], ptr [[BASE:%.*]])
1239 // CPP-CHECK-NEXT: ret { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP1]]
1240 //
1241 svbfloat16x4_t test_svld4q_bf16(svbool_t pg, const bfloat16_t *base) ATTR
1242 {
1243 return SVE_ACLE_FUNC(svld4q,,_bf16,)(pg, base);
1244 }
1245
1246 // CHECK-LABEL: @test_svld4q_f32(
1247 // CHECK-NEXT: entry:
1248 // CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
1249 // CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.ld4q.sret.nxv4f32(<vscale x 4 x i1> [[TMP0]], ptr [[BASE:%.*]])
1250 // CHECK-NEXT: ret { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP1]]
1251 //
1252 // CPP-CHECK-LABEL: @_Z15test_svld4q_f32u10__SVBool_tPKf(
1253 // CPP-CHECK-NEXT: entry:
1254 // CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
1255 // CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.ld4q.sret.nxv4f32(<vscale x 4 x i1> [[TMP0]], ptr [[BASE:%.*]])
1256 // CPP-CHECK-NEXT: ret { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP1]]
1257 //
1258 svfloat32x4_t test_svld4q_f32(svbool_t pg, const float32_t *base) ATTR
1259 {
1260 return SVE_ACLE_FUNC(svld4q,,_f32,)(pg, base);
1261 }
1262
1263 // CHECK-LABEL: @test_svld4q_f64(
1264 // CHECK-NEXT: entry:
1265 // CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
1266 // CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } @llvm.aarch64.sve.ld4q.sret.nxv2f64(<vscale x 2 x i1> [[TMP0]], ptr [[BASE:%.*]])
1267 // CHECK-NEXT: ret { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP1]]
1268 //
1269 // CPP-CHECK-LABEL: @_Z15test_svld4q_f64u10__SVBool_tPKd(
1270 // CPP-CHECK-NEXT: entry:
1271 // CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
1272 // CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } @llvm.aarch64.sve.ld4q.sret.nxv2f64(<vscale x 2 x i1> [[TMP0]], ptr [[BASE:%.*]])
1273 // CPP-CHECK-NEXT: ret { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP1]]
1274 //
1275 svfloat64x4_t test_svld4q_f64(svbool_t pg, const float64_t *base) ATTR
1276 {
1277 return SVE_ACLE_FUNC(svld4q,,_f64,)(pg, base);
1278 }
1279
1280 // CHECK-LABEL: @test_svld4q_mf8(
1281 // CHECK-NEXT: entry:
1282 // CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld4q.sret.nxv16i8(<vscale x 16 x i1> [[PG:%.*]], ptr [[BASE:%.*]])
1283 // CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]]
1284 //
1285 // CPP-CHECK-LABEL: @_Z15test_svld4q_mf8u10__SVBool_tPKu6__mfp8(
1286 // CPP-CHECK-NEXT: entry:
1287 // CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld4q.sret.nxv16i8(<vscale x 16 x i1> [[PG:%.*]], ptr [[BASE:%.*]])
1288 // CPP-CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP0]]
1289 //
1290 svmfloat8x4_t test_svld4q_mf8(svbool_t pg, const mfloat8_t *base) ATTR
1291 {
1292 return SVE_ACLE_FUNC(svld4q,,_mf8,)(pg, base);
1293 }
1294
1295 // CHECK-LABEL: @test_svld4q_vnum_u8(
1296 // CHECK-NEXT: entry:
1297 // CHECK-NEXT: [[TMP0:%.*]] = tail call i64 @llvm.vscale.i64()
1298 // CHECK-NEXT: [[TMP1:%.*]] = shl nuw nsw i64 [[TMP0]], 4
1299 // CHECK-NEXT: [[DOTIDX:%.*]] = mul i64 [[VNUM:%.*]], [[TMP1]]
1300 // CHECK-NEXT: [[TMP2:%.*]] = getelementptr i8, ptr [[BASE:%.*]], i64 [[DOTIDX]]
1301 // CHECK-NEXT: [[TMP3:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld4q.sret.nxv16i8(<vscale x 16 x i1> [[PG:%.*]], ptr [[TMP2]])
1302 // CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP3]]
1303 //
1304 // CPP-CHECK-LABEL: @_Z19test_svld4q_vnum_u8u10__SVBool_tPKhl(
1305 // CPP-CHECK-NEXT: entry:
1306 // CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call i64 @llvm.vscale.i64()
1307 // CPP-CHECK-NEXT: [[TMP1:%.*]] = shl nuw nsw i64 [[TMP0]], 4
1308 // CPP-CHECK-NEXT: [[DOTIDX:%.*]] = mul i64 [[VNUM:%.*]], [[TMP1]]
1309 // CPP-CHECK-NEXT: [[TMP2:%.*]] = getelementptr i8, ptr [[BASE:%.*]], i64 [[DOTIDX]]
1310 // CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld4q.sret.nxv16i8(<vscale x 16 x i1> [[PG:%.*]], ptr [[TMP2]])
1311 // CPP-CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP3]]
1312 //
1313 svuint8x4_t test_svld4q_vnum_u8(svbool_t pg, const uint8_t *base, int64_t vnum) ATTR
1314 {
1315 return SVE_ACLE_FUNC(svld4q_vnum_,,u8,)(pg, base, vnum);
1316 }
1317
1318 // CHECK-LABEL: @test_svld4q_vnum_s8(
1319 // CHECK-NEXT: entry:
1320 // CHECK-NEXT: [[TMP0:%.*]] = tail call i64 @llvm.vscale.i64()
1321 // CHECK-NEXT: [[TMP1:%.*]] = shl nuw nsw i64 [[TMP0]], 4
1322 // CHECK-NEXT: [[DOTIDX:%.*]] = mul i64 [[VNUM:%.*]], [[TMP1]]
1323 // CHECK-NEXT: [[TMP2:%.*]] = getelementptr i8, ptr [[BASE:%.*]], i64 [[DOTIDX]]
1324 // CHECK-NEXT: [[TMP3:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld4q.sret.nxv16i8(<vscale x 16 x i1> [[PG:%.*]], ptr [[TMP2]])
1325 // CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP3]]
1326 //
1327 // CPP-CHECK-LABEL: @_Z19test_svld4q_vnum_s8u10__SVBool_tPKal(
1328 // CPP-CHECK-NEXT: entry:
1329 // CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call i64 @llvm.vscale.i64()
1330 // CPP-CHECK-NEXT: [[TMP1:%.*]] = shl nuw nsw i64 [[TMP0]], 4
1331 // CPP-CHECK-NEXT: [[DOTIDX:%.*]] = mul i64 [[VNUM:%.*]], [[TMP1]]
1332 // CPP-CHECK-NEXT: [[TMP2:%.*]] = getelementptr i8, ptr [[BASE:%.*]], i64 [[DOTIDX]]
1333 // CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld4q.sret.nxv16i8(<vscale x 16 x i1> [[PG:%.*]], ptr [[TMP2]])
1334 // CPP-CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP3]]
1335 //
1336 svint8x4_t test_svld4q_vnum_s8(svbool_t pg, const int8_t *base, int64_t vnum) ATTR
1337 {
1338 return SVE_ACLE_FUNC(svld4q_vnum,,_s8,)(pg, base, vnum);
1339 }
1340 // CHECK-LABEL: @test_svld4q_vnum_u16(
1341 // CHECK-NEXT: entry:
1342 // CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
1343 // CHECK-NEXT: [[TMP1:%.*]] = tail call i64 @llvm.vscale.i64()
1344 // CHECK-NEXT: [[TMP2:%.*]] = shl nuw nsw i64 [[TMP1]], 4
1345 // CHECK-NEXT: [[DOTIDX:%.*]] = mul i64 [[VNUM:%.*]], [[TMP2]]
1346 // CHECK-NEXT: [[TMP3:%.*]] = getelementptr i8, ptr [[BASE:%.*]], i64 [[DOTIDX]]
1347 // CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.ld4q.sret.nxv8i16(<vscale x 8 x i1> [[TMP0]], ptr [[TMP3]])
1348 // CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP4]]
1349 //
1350 // CPP-CHECK-LABEL: @_Z20test_svld4q_vnum_u16u10__SVBool_tPKtl(
1351 // CPP-CHECK-NEXT: entry:
1352 // CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
1353 // CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call i64 @llvm.vscale.i64()
1354 // CPP-CHECK-NEXT: [[TMP2:%.*]] = shl nuw nsw i64 [[TMP1]], 4
1355 // CPP-CHECK-NEXT: [[DOTIDX:%.*]] = mul i64 [[VNUM:%.*]], [[TMP2]]
1356 // CPP-CHECK-NEXT: [[TMP3:%.*]] = getelementptr i8, ptr [[BASE:%.*]], i64 [[DOTIDX]]
1357 // CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.ld4q.sret.nxv8i16(<vscale x 8 x i1> [[TMP0]], ptr [[TMP3]])
1358 // CPP-CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP4]]
1359 //
1360 svuint16x4_t test_svld4q_vnum_u16(svbool_t pg, const uint16_t *base, int64_t vnum) ATTR
1361 {
1362 return SVE_ACLE_FUNC(svld4q_vnum,,_u16,)(pg, base, vnum);
1363 }
1364
1365 // CHECK-LABEL: @test_svld4q_vnum_s16(
1366 // CHECK-NEXT: entry:
1367 // CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
1368 // CHECK-NEXT: [[TMP1:%.*]] = tail call i64 @llvm.vscale.i64()
1369 // CHECK-NEXT: [[TMP2:%.*]] = shl nuw nsw i64 [[TMP1]], 4
1370 // CHECK-NEXT: [[DOTIDX:%.*]] = mul i64 [[VNUM:%.*]], [[TMP2]]
1371 // CHECK-NEXT: [[TMP3:%.*]] = getelementptr i8, ptr [[BASE:%.*]], i64 [[DOTIDX]]
1372 // CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.ld4q.sret.nxv8i16(<vscale x 8 x i1> [[TMP0]], ptr [[TMP3]])
1373 // CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP4]]
1374 //
1375 // CPP-CHECK-LABEL: @_Z20test_svld4q_vnum_s16u10__SVBool_tPKsl(
1376 // CPP-CHECK-NEXT: entry:
1377 // CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
1378 // CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call i64 @llvm.vscale.i64()
1379 // CPP-CHECK-NEXT: [[TMP2:%.*]] = shl nuw nsw i64 [[TMP1]], 4
1380 // CPP-CHECK-NEXT: [[DOTIDX:%.*]] = mul i64 [[VNUM:%.*]], [[TMP2]]
1381 // CPP-CHECK-NEXT: [[TMP3:%.*]] = getelementptr i8, ptr [[BASE:%.*]], i64 [[DOTIDX]]
1382 // CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } @llvm.aarch64.sve.ld4q.sret.nxv8i16(<vscale x 8 x i1> [[TMP0]], ptr [[TMP3]])
1383 // CPP-CHECK-NEXT: ret { <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16>, <vscale x 8 x i16> } [[TMP4]]
1384 //
1385 svint16x4_t test_svld4q_vnum_s16(svbool_t pg, const int16_t *base, int64_t vnum) ATTR
1386 {
1387 return SVE_ACLE_FUNC(svld4q_vnum,,_s16,)(pg, base, vnum);
1388 }
1389
1390 // CHECK-LABEL: @test_svld4q_vnum_u32(
1391 // CHECK-NEXT: entry:
1392 // CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
1393 // CHECK-NEXT: [[TMP1:%.*]] = tail call i64 @llvm.vscale.i64()
1394 // CHECK-NEXT: [[TMP2:%.*]] = shl nuw nsw i64 [[TMP1]], 4
1395 // CHECK-NEXT: [[DOTIDX:%.*]] = mul i64 [[VNUM:%.*]], [[TMP2]]
1396 // CHECK-NEXT: [[TMP3:%.*]] = getelementptr i8, ptr [[BASE:%.*]], i64 [[DOTIDX]]
1397 // CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.ld4q.sret.nxv4i32(<vscale x 4 x i1> [[TMP0]], ptr [[TMP3]])
1398 // CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP4]]
1399 //
1400 // CPP-CHECK-LABEL: @_Z20test_svld4q_vnum_u32u10__SVBool_tPKjl(
1401 // CPP-CHECK-NEXT: entry:
1402 // CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
1403 // CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call i64 @llvm.vscale.i64()
1404 // CPP-CHECK-NEXT: [[TMP2:%.*]] = shl nuw nsw i64 [[TMP1]], 4
1405 // CPP-CHECK-NEXT: [[DOTIDX:%.*]] = mul i64 [[VNUM:%.*]], [[TMP2]]
1406 // CPP-CHECK-NEXT: [[TMP3:%.*]] = getelementptr i8, ptr [[BASE:%.*]], i64 [[DOTIDX]]
1407 // CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.ld4q.sret.nxv4i32(<vscale x 4 x i1> [[TMP0]], ptr [[TMP3]])
1408 // CPP-CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP4]]
1409 //
1410 svuint32x4_t test_svld4q_vnum_u32(svbool_t pg, const uint32_t *base, int64_t vnum) ATTR
1411 {
1412 return SVE_ACLE_FUNC(svld4q_vnum,,_u32,)(pg, base, vnum);
1413 }
1414
1415 // CHECK-LABEL: @test_svld4q_vnum_s32(
1416 // CHECK-NEXT: entry:
1417 // CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
1418 // CHECK-NEXT: [[TMP1:%.*]] = tail call i64 @llvm.vscale.i64()
1419 // CHECK-NEXT: [[TMP2:%.*]] = shl nuw nsw i64 [[TMP1]], 4
1420 // CHECK-NEXT: [[DOTIDX:%.*]] = mul i64 [[VNUM:%.*]], [[TMP2]]
1421 // CHECK-NEXT: [[TMP3:%.*]] = getelementptr i8, ptr [[BASE:%.*]], i64 [[DOTIDX]]
1422 // CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.ld4q.sret.nxv4i32(<vscale x 4 x i1> [[TMP0]], ptr [[TMP3]])
1423 // CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP4]]
1424 //
1425 // CPP-CHECK-LABEL: @_Z20test_svld4q_vnum_s32u10__SVBool_tPKil(
1426 // CPP-CHECK-NEXT: entry:
1427 // CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
1428 // CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call i64 @llvm.vscale.i64()
1429 // CPP-CHECK-NEXT: [[TMP2:%.*]] = shl nuw nsw i64 [[TMP1]], 4
1430 // CPP-CHECK-NEXT: [[DOTIDX:%.*]] = mul i64 [[VNUM:%.*]], [[TMP2]]
1431 // CPP-CHECK-NEXT: [[TMP3:%.*]] = getelementptr i8, ptr [[BASE:%.*]], i64 [[DOTIDX]]
1432 // CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.aarch64.sve.ld4q.sret.nxv4i32(<vscale x 4 x i1> [[TMP0]], ptr [[TMP3]])
1433 // CPP-CHECK-NEXT: ret { <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32> } [[TMP4]]
1434 //
1435 svint32x4_t test_svld4q_vnum_s32(svbool_t pg, const int32_t *base, int64_t vnum) ATTR
1436 {
1437 return SVE_ACLE_FUNC(svld4q_vnum,,_s32,)(pg, base, vnum);
1438 }
1439
1440 // CHECK-LABEL: @test_svld4q_vnum_u64(
1441 // CHECK-NEXT: entry:
1442 // CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
1443 // CHECK-NEXT: [[TMP1:%.*]] = tail call i64 @llvm.vscale.i64()
1444 // CHECK-NEXT: [[TMP2:%.*]] = shl nuw nsw i64 [[TMP1]], 4
1445 // CHECK-NEXT: [[DOTIDX:%.*]] = mul i64 [[VNUM:%.*]], [[TMP2]]
1446 // CHECK-NEXT: [[TMP3:%.*]] = getelementptr i8, ptr [[BASE:%.*]], i64 [[DOTIDX]]
1447 // CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.ld4q.sret.nxv2i64(<vscale x 2 x i1> [[TMP0]], ptr [[TMP3]])
1448 // CHECK-NEXT: ret { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP4]]
1449 //
1450 // CPP-CHECK-LABEL: @_Z20test_svld4q_vnum_u64u10__SVBool_tPKml(
1451 // CPP-CHECK-NEXT: entry:
1452 // CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
1453 // CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call i64 @llvm.vscale.i64()
1454 // CPP-CHECK-NEXT: [[TMP2:%.*]] = shl nuw nsw i64 [[TMP1]], 4
1455 // CPP-CHECK-NEXT: [[DOTIDX:%.*]] = mul i64 [[VNUM:%.*]], [[TMP2]]
1456 // CPP-CHECK-NEXT: [[TMP3:%.*]] = getelementptr i8, ptr [[BASE:%.*]], i64 [[DOTIDX]]
1457 // CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.ld4q.sret.nxv2i64(<vscale x 2 x i1> [[TMP0]], ptr [[TMP3]])
1458 // CPP-CHECK-NEXT: ret { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP4]]
1459 //
1460 svuint64x4_t test_svld4q_vnum_u64(svbool_t pg, const uint64_t *base, int64_t vnum) ATTR
1461 {
1462 return SVE_ACLE_FUNC(svld4q_vnum,,_u64,)(pg, base, vnum);
1463 }
1464
1465 // CHECK-LABEL: @test_svld4q_vnum_s64(
1466 // CHECK-NEXT: entry:
1467 // CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
1468 // CHECK-NEXT: [[TMP1:%.*]] = tail call i64 @llvm.vscale.i64()
1469 // CHECK-NEXT: [[TMP2:%.*]] = shl nuw nsw i64 [[TMP1]], 4
1470 // CHECK-NEXT: [[DOTIDX:%.*]] = mul i64 [[VNUM:%.*]], [[TMP2]]
1471 // CHECK-NEXT: [[TMP3:%.*]] = getelementptr i8, ptr [[BASE:%.*]], i64 [[DOTIDX]]
1472 // CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.ld4q.sret.nxv2i64(<vscale x 2 x i1> [[TMP0]], ptr [[TMP3]])
1473 // CHECK-NEXT: ret { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP4]]
1474 //
1475 // CPP-CHECK-LABEL: @_Z20test_svld4q_vnum_s64u10__SVBool_tPKll(
1476 // CPP-CHECK-NEXT: entry:
1477 // CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
1478 // CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call i64 @llvm.vscale.i64()
1479 // CPP-CHECK-NEXT: [[TMP2:%.*]] = shl nuw nsw i64 [[TMP1]], 4
1480 // CPP-CHECK-NEXT: [[DOTIDX:%.*]] = mul i64 [[VNUM:%.*]], [[TMP2]]
1481 // CPP-CHECK-NEXT: [[TMP3:%.*]] = getelementptr i8, ptr [[BASE:%.*]], i64 [[DOTIDX]]
1482 // CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.aarch64.sve.ld4q.sret.nxv2i64(<vscale x 2 x i1> [[TMP0]], ptr [[TMP3]])
1483 // CPP-CHECK-NEXT: ret { <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64> } [[TMP4]]
1484 //
1485 svint64x4_t test_svld4q_vnum_s64(svbool_t pg, const int64_t *base, int64_t vnum) ATTR
1486 {
1487 return SVE_ACLE_FUNC(svld4q_vnum,,_s64,)(pg, base, vnum);
1488 }
1489
1490 // CHECK-LABEL: @test_svld4q_vnum_f16(
1491 // CHECK-NEXT: entry:
1492 // CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
1493 // CHECK-NEXT: [[TMP1:%.*]] = tail call i64 @llvm.vscale.i64()
1494 // CHECK-NEXT: [[TMP2:%.*]] = shl nuw nsw i64 [[TMP1]], 4
1495 // CHECK-NEXT: [[DOTIDX:%.*]] = mul i64 [[VNUM:%.*]], [[TMP2]]
1496 // CHECK-NEXT: [[TMP3:%.*]] = getelementptr i8, ptr [[BASE:%.*]], i64 [[DOTIDX]]
1497 // CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sve.ld4q.sret.nxv8f16(<vscale x 8 x i1> [[TMP0]], ptr [[TMP3]])
1498 // CHECK-NEXT: ret { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP4]]
1499 //
1500 // CPP-CHECK-LABEL: @_Z20test_svld4q_vnum_f16u10__SVBool_tPKDhl(
1501 // CPP-CHECK-NEXT: entry:
1502 // CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
1503 // CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call i64 @llvm.vscale.i64()
1504 // CPP-CHECK-NEXT: [[TMP2:%.*]] = shl nuw nsw i64 [[TMP1]], 4
1505 // CPP-CHECK-NEXT: [[DOTIDX:%.*]] = mul i64 [[VNUM:%.*]], [[TMP2]]
1506 // CPP-CHECK-NEXT: [[TMP3:%.*]] = getelementptr i8, ptr [[BASE:%.*]], i64 [[DOTIDX]]
1507 // CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } @llvm.aarch64.sve.ld4q.sret.nxv8f16(<vscale x 8 x i1> [[TMP0]], ptr [[TMP3]])
1508 // CPP-CHECK-NEXT: ret { <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half>, <vscale x 8 x half> } [[TMP4]]
1509 //
1510 svfloat16x4_t test_svld4q_vnum_f16(svbool_t pg, const float16_t *base, int64_t vnum) ATTR
1511 {
1512 return SVE_ACLE_FUNC(svld4q_vnum,,_f16,)(pg, base, vnum);
1513 }
1514
1515 // CHECK-LABEL: @test_svld4q_vnum_bf16(
1516 // CHECK-NEXT: entry:
1517 // CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
1518 // CHECK-NEXT: [[TMP1:%.*]] = tail call i64 @llvm.vscale.i64()
1519 // CHECK-NEXT: [[TMP2:%.*]] = shl nuw nsw i64 [[TMP1]], 4
1520 // CHECK-NEXT: [[DOTIDX:%.*]] = mul i64 [[VNUM:%.*]], [[TMP2]]
1521 // CHECK-NEXT: [[TMP3:%.*]] = getelementptr i8, ptr [[BASE:%.*]], i64 [[DOTIDX]]
1522 // CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } @llvm.aarch64.sve.ld4q.sret.nxv8bf16(<vscale x 8 x i1> [[TMP0]], ptr [[TMP3]])
1523 // CHECK-NEXT: ret { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP4]]
1524 //
1525 // CPP-CHECK-LABEL: @_Z21test_svld4q_vnum_bf16u10__SVBool_tPKu6__bf16l(
1526 // CPP-CHECK-NEXT: entry:
1527 // CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
1528 // CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call i64 @llvm.vscale.i64()
1529 // CPP-CHECK-NEXT: [[TMP2:%.*]] = shl nuw nsw i64 [[TMP1]], 4
1530 // CPP-CHECK-NEXT: [[DOTIDX:%.*]] = mul i64 [[VNUM:%.*]], [[TMP2]]
1531 // CPP-CHECK-NEXT: [[TMP3:%.*]] = getelementptr i8, ptr [[BASE:%.*]], i64 [[DOTIDX]]
1532 // CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } @llvm.aarch64.sve.ld4q.sret.nxv8bf16(<vscale x 8 x i1> [[TMP0]], ptr [[TMP3]])
1533 // CPP-CHECK-NEXT: ret { <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat> } [[TMP4]]
1534 //
1535 svbfloat16x4_t test_svld4q_vnum_bf16(svbool_t pg, const bfloat16_t *base, int64_t vnum) ATTR
1536 {
1537 return SVE_ACLE_FUNC(svld4q_vnum,,_bf16,)(pg, base, vnum);
1538 }
1539
1540 // CHECK-LABEL: @test_svld4q_vnum_f32(
1541 // CHECK-NEXT: entry:
1542 // CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
1543 // CHECK-NEXT: [[TMP1:%.*]] = tail call i64 @llvm.vscale.i64()
1544 // CHECK-NEXT: [[TMP2:%.*]] = shl nuw nsw i64 [[TMP1]], 4
1545 // CHECK-NEXT: [[DOTIDX:%.*]] = mul i64 [[VNUM:%.*]], [[TMP2]]
1546 // CHECK-NEXT: [[TMP3:%.*]] = getelementptr i8, ptr [[BASE:%.*]], i64 [[DOTIDX]]
1547 // CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.ld4q.sret.nxv4f32(<vscale x 4 x i1> [[TMP0]], ptr [[TMP3]])
1548 // CHECK-NEXT: ret { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP4]]
1549 //
1550 // CPP-CHECK-LABEL: @_Z20test_svld4q_vnum_f32u10__SVBool_tPKfl(
1551 // CPP-CHECK-NEXT: entry:
1552 // CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
1553 // CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call i64 @llvm.vscale.i64()
1554 // CPP-CHECK-NEXT: [[TMP2:%.*]] = shl nuw nsw i64 [[TMP1]], 4
1555 // CPP-CHECK-NEXT: [[DOTIDX:%.*]] = mul i64 [[VNUM:%.*]], [[TMP2]]
1556 // CPP-CHECK-NEXT: [[TMP3:%.*]] = getelementptr i8, ptr [[BASE:%.*]], i64 [[DOTIDX]]
1557 // CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } @llvm.aarch64.sve.ld4q.sret.nxv4f32(<vscale x 4 x i1> [[TMP0]], ptr [[TMP3]])
1558 // CPP-CHECK-NEXT: ret { <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float>, <vscale x 4 x float> } [[TMP4]]
1559 //
1560 svfloat32x4_t test_svld4q_vnum_f32(svbool_t pg, const float32_t *base, int64_t vnum) ATTR
1561 {
1562 return SVE_ACLE_FUNC(svld4q_vnum,,_f32,)(pg, base, vnum);
1563 }
1564
1565 // CHECK-LABEL: @test_svld4q_vnum_f64(
1566 // CHECK-NEXT: entry:
1567 // CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
1568 // CHECK-NEXT: [[TMP1:%.*]] = tail call i64 @llvm.vscale.i64()
1569 // CHECK-NEXT: [[TMP2:%.*]] = shl nuw nsw i64 [[TMP1]], 4
1570 // CHECK-NEXT: [[DOTIDX:%.*]] = mul i64 [[VNUM:%.*]], [[TMP2]]
1571 // CHECK-NEXT: [[TMP3:%.*]] = getelementptr i8, ptr [[BASE:%.*]], i64 [[DOTIDX]]
1572 // CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } @llvm.aarch64.sve.ld4q.sret.nxv2f64(<vscale x 2 x i1> [[TMP0]], ptr [[TMP3]])
1573 // CHECK-NEXT: ret { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP4]]
1574 //
1575 // CPP-CHECK-LABEL: @_Z20test_svld4q_vnum_f64u10__SVBool_tPKdl(
1576 // CPP-CHECK-NEXT: entry:
1577 // CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
1578 // CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call i64 @llvm.vscale.i64()
1579 // CPP-CHECK-NEXT: [[TMP2:%.*]] = shl nuw nsw i64 [[TMP1]], 4
1580 // CPP-CHECK-NEXT: [[DOTIDX:%.*]] = mul i64 [[VNUM:%.*]], [[TMP2]]
1581 // CPP-CHECK-NEXT: [[TMP3:%.*]] = getelementptr i8, ptr [[BASE:%.*]], i64 [[DOTIDX]]
1582 // CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } @llvm.aarch64.sve.ld4q.sret.nxv2f64(<vscale x 2 x i1> [[TMP0]], ptr [[TMP3]])
1583 // CPP-CHECK-NEXT: ret { <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double>, <vscale x 2 x double> } [[TMP4]]
1584 //
1585 svfloat64x4_t test_svld4q_vnum_f64(svbool_t pg, const float64_t *base, int64_t vnum) ATTR
1586 {
1587 return SVE_ACLE_FUNC(svld4q_vnum,,_f64,)(pg, base, vnum);
1588 }
1589
1590 // CHECK-LABEL: @test_svld4q_vnum_mf8(
1591 // CHECK-NEXT: entry:
1592 // CHECK-NEXT: [[TMP0:%.*]] = tail call i64 @llvm.vscale.i64()
1593 // CHECK-NEXT: [[TMP1:%.*]] = shl nuw nsw i64 [[TMP0]], 4
1594 // CHECK-NEXT: [[DOTIDX:%.*]] = mul i64 [[VNUM:%.*]], [[TMP1]]
1595 // CHECK-NEXT: [[TMP2:%.*]] = getelementptr i8, ptr [[BASE:%.*]], i64 [[DOTIDX]]
1596 // CHECK-NEXT: [[TMP3:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld4q.sret.nxv16i8(<vscale x 16 x i1> [[PG:%.*]], ptr [[TMP2]])
1597 // CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP3]]
1598 //
1599 // CPP-CHECK-LABEL: @_Z20test_svld4q_vnum_mf8u10__SVBool_tPKu6__mfp8l(
1600 // CPP-CHECK-NEXT: entry:
1601 // CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call i64 @llvm.vscale.i64()
1602 // CPP-CHECK-NEXT: [[TMP1:%.*]] = shl nuw nsw i64 [[TMP0]], 4
1603 // CPP-CHECK-NEXT: [[DOTIDX:%.*]] = mul i64 [[VNUM:%.*]], [[TMP1]]
1604 // CPP-CHECK-NEXT: [[TMP2:%.*]] = getelementptr i8, ptr [[BASE:%.*]], i64 [[DOTIDX]]
1605 // CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.aarch64.sve.ld4q.sret.nxv16i8(<vscale x 16 x i1> [[PG:%.*]], ptr [[TMP2]])
1606 // CPP-CHECK-NEXT: ret { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[TMP3]]
1607 //
1608 svmfloat8x4_t test_svld4q_vnum_mf8(svbool_t pg, const mfloat8_t *base, int64_t vnum) ATTR
1609 {
1610 return SVE_ACLE_FUNC(svld4q_vnum,,_mf8,)(pg, base, vnum);
1611 }