1 // REQUIRES: aarch64-registered-target
2 // RUN: %clang_cc1 -triple aarch64-none-linux-gnu -target-feature +sve -fallow-half-arguments-and-returns -S -O1 -Werror -emit-llvm -o - %s | FileCheck %s
3 // RUN: %clang_cc1 -triple aarch64-none-linux-gnu -target-feature +sve -fallow-half-arguments-and-returns -S -O1 -Werror -emit-llvm -o - -x c++ %s | FileCheck %s
4 // RUN: %clang_cc1 -DSVE_OVERLOADED_FORMS -triple aarch64-none-linux-gnu -target-feature +sve -fallow-half-arguments-and-returns -S -O1 -Werror -emit-llvm -o - %s | FileCheck %s
5 // RUN: %clang_cc1 -DSVE_OVERLOADED_FORMS -triple aarch64-none-linux-gnu -target-feature +sve -fallow-half-arguments-and-returns -S -O1 -Werror -emit-llvm -o - -x c++ %s | FileCheck %s
6 // RUN: %clang_cc1 -triple aarch64-none-linux-gnu -target-feature +sve -fallow-half-arguments-and-returns -S -O1 -Werror -o - %s >/dev/null
7 #include <arm_sve.h>
8
9 #ifdef SVE_OVERLOADED_FORMS
10 // A simple used,unused... macro, long enough to represent any SVE builtin.
11 #define SVE_ACLE_FUNC(A1,A2_UNUSED,A3,A4_UNUSED) A1##A3
12 #else
13 #define SVE_ACLE_FUNC(A1,A2,A3,A4) A1##A2##A3##A4
14 #endif
15
test_svld1_s8(svbool_t pg,const int8_t * base)16 svint8_t test_svld1_s8(svbool_t pg, const int8_t *base)
17 {
18 // CHECK-LABEL: test_svld1_s8
19 // CHECK: %[[LOAD:.*]] = call <vscale x 16 x i8> @llvm.aarch64.sve.ld1.nxv16i8(<vscale x 16 x i1> %pg, i8* %base)
20 // CHECK: ret <vscale x 16 x i8> %[[LOAD]]
21 return SVE_ACLE_FUNC(svld1,_s8,,)(pg, base);
22 }
23
test_svld1_s16(svbool_t pg,const int16_t * base)24 svint16_t test_svld1_s16(svbool_t pg, const int16_t *base)
25 {
26 // CHECK-LABEL: test_svld1_s16
27 // CHECK: %[[PG:.*]] = call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> %pg)
28 // CHECK: %[[LOAD:.*]] = call <vscale x 8 x i16> @llvm.aarch64.sve.ld1.nxv8i16(<vscale x 8 x i1> %[[PG]], i16* %base)
29 // CHECK: ret <vscale x 8 x i16> %[[LOAD]]
30 return SVE_ACLE_FUNC(svld1,_s16,,)(pg, base);
31 }
32
test_svld1_s32(svbool_t pg,const int32_t * base)33 svint32_t test_svld1_s32(svbool_t pg, const int32_t *base)
34 {
35 // CHECK-LABEL: test_svld1_s32
36 // CHECK: %[[PG:.*]] = call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg)
37 // CHECK: %[[LOAD:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.ld1.nxv4i32(<vscale x 4 x i1> %[[PG]], i32* %base)
38 // CHECK: ret <vscale x 4 x i32> %[[LOAD]]
39 return SVE_ACLE_FUNC(svld1,_s32,,)(pg, base);
40 }
41
test_svld1_s64(svbool_t pg,const int64_t * base)42 svint64_t test_svld1_s64(svbool_t pg, const int64_t *base)
43 {
44 // CHECK-LABEL: test_svld1_s64
45 // CHECK: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg)
46 // CHECK: %[[LOAD:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.ld1.nxv2i64(<vscale x 2 x i1> %[[PG]], i64* %base)
47 // CHECK: ret <vscale x 2 x i64> %[[LOAD]]
48 return SVE_ACLE_FUNC(svld1,_s64,,)(pg, base);
49 }
50
test_svld1_u8(svbool_t pg,const uint8_t * base)51 svuint8_t test_svld1_u8(svbool_t pg, const uint8_t *base)
52 {
53 // CHECK-LABEL: test_svld1_u8
54 // CHECK: %[[LOAD:.*]] = call <vscale x 16 x i8> @llvm.aarch64.sve.ld1.nxv16i8(<vscale x 16 x i1> %pg, i8* %base)
55 // CHECK: ret <vscale x 16 x i8> %[[LOAD]]
56 return SVE_ACLE_FUNC(svld1,_u8,,)(pg, base);
57 }
58
test_svld1_u16(svbool_t pg,const uint16_t * base)59 svuint16_t test_svld1_u16(svbool_t pg, const uint16_t *base)
60 {
61 // CHECK-LABEL: test_svld1_u16
62 // CHECK: %[[PG:.*]] = call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> %pg)
63 // CHECK: %[[LOAD:.*]] = call <vscale x 8 x i16> @llvm.aarch64.sve.ld1.nxv8i16(<vscale x 8 x i1> %[[PG]], i16* %base)
64 // CHECK: ret <vscale x 8 x i16> %[[LOAD]]
65 return SVE_ACLE_FUNC(svld1,_u16,,)(pg, base);
66 }
67
test_svld1_u32(svbool_t pg,const uint32_t * base)68 svuint32_t test_svld1_u32(svbool_t pg, const uint32_t *base)
69 {
70 // CHECK-LABEL: test_svld1_u32
71 // CHECK: %[[PG:.*]] = call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg)
72 // CHECK: %[[LOAD:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.ld1.nxv4i32(<vscale x 4 x i1> %[[PG]], i32* %base)
73 // CHECK: ret <vscale x 4 x i32> %[[LOAD]]
74 return SVE_ACLE_FUNC(svld1,_u32,,)(pg, base);
75 }
76
test_svld1_u64(svbool_t pg,const uint64_t * base)77 svuint64_t test_svld1_u64(svbool_t pg, const uint64_t *base)
78 {
79 // CHECK-LABEL: test_svld1_u64
80 // CHECK: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg)
81 // CHECK: %[[LOAD:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.ld1.nxv2i64(<vscale x 2 x i1> %[[PG]], i64* %base)
82 // CHECK: ret <vscale x 2 x i64> %[[LOAD]]
83 return SVE_ACLE_FUNC(svld1,_u64,,)(pg, base);
84 }
85
test_svld1_f16(svbool_t pg,const float16_t * base)86 svfloat16_t test_svld1_f16(svbool_t pg, const float16_t *base)
87 {
88 // CHECK-LABEL: test_svld1_f16
89 // CHECK: %[[PG:.*]] = call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> %pg)
90 // CHECK: %[[LOAD:.*]] = call <vscale x 8 x half> @llvm.aarch64.sve.ld1.nxv8f16(<vscale x 8 x i1> %[[PG]], half* %base)
91 // CHECK: ret <vscale x 8 x half> %[[LOAD]]
92 return SVE_ACLE_FUNC(svld1,_f16,,)(pg, base);
93 }
94
test_svld1_f32(svbool_t pg,const float32_t * base)95 svfloat32_t test_svld1_f32(svbool_t pg, const float32_t *base)
96 {
97 // CHECK-LABEL: test_svld1_f32
98 // CHECK: %[[PG:.*]] = call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg)
99 // CHECK: %[[LOAD:.*]] = call <vscale x 4 x float> @llvm.aarch64.sve.ld1.nxv4f32(<vscale x 4 x i1> %[[PG]], float* %base)
100 // CHECK: ret <vscale x 4 x float> %[[LOAD]]
101 return SVE_ACLE_FUNC(svld1,_f32,,)(pg, base);
102 }
103
test_svld1_f64(svbool_t pg,const float64_t * base)104 svfloat64_t test_svld1_f64(svbool_t pg, const float64_t *base)
105 {
106 // CHECK-LABEL: test_svld1_f64
107 // CHECK: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg)
108 // CHECK: %[[LOAD:.*]] = call <vscale x 2 x double> @llvm.aarch64.sve.ld1.nxv2f64(<vscale x 2 x i1> %[[PG]], double* %base)
109 // CHECK: ret <vscale x 2 x double> %[[LOAD]]
110 return SVE_ACLE_FUNC(svld1,_f64,,)(pg, base);
111 }
112
test_svld1_vnum_s8(svbool_t pg,const int8_t * base,int64_t vnum)113 svint8_t test_svld1_vnum_s8(svbool_t pg, const int8_t *base, int64_t vnum)
114 {
115 // CHECK-LABEL: test_svld1_vnum_s8
116 // CHECK: %[[BITCAST:.*]] = bitcast i8* %base to <vscale x 16 x i8>*
117 // CHECK: %[[GEP:.*]] = getelementptr <vscale x 16 x i8>, <vscale x 16 x i8>* %[[BITCAST]], i64 %vnum, i64 0
118 // CHECK: %[[LOAD:.*]] = call <vscale x 16 x i8> @llvm.aarch64.sve.ld1.nxv16i8(<vscale x 16 x i1> %pg, i8* %[[GEP]])
119 // CHECK: ret <vscale x 16 x i8> %[[LOAD]]
120 return SVE_ACLE_FUNC(svld1_vnum,_s8,,)(pg, base, vnum);
121 }
122
test_svld1_vnum_s16(svbool_t pg,const int16_t * base,int64_t vnum)123 svint16_t test_svld1_vnum_s16(svbool_t pg, const int16_t *base, int64_t vnum)
124 {
125 // CHECK-LABEL: test_svld1_vnum_s16
126 // CHECK-DAG: %[[PG:.*]] = call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> %pg)
127 // CHECK-DAG: %[[BITCAST:.*]] = bitcast i16* %base to <vscale x 8 x i16>*
128 // CHECK-DAG: %[[GEP:.*]] = getelementptr <vscale x 8 x i16>, <vscale x 8 x i16>* %[[BITCAST]], i64 %vnum, i64 0
129 // CHECK: %[[LOAD:.*]] = call <vscale x 8 x i16> @llvm.aarch64.sve.ld1.nxv8i16(<vscale x 8 x i1> %[[PG]], i16* %[[GEP]])
130 // CHECK: ret <vscale x 8 x i16> %[[LOAD]]
131 return SVE_ACLE_FUNC(svld1_vnum,_s16,,)(pg, base, vnum);
132 }
133
test_svld1_vnum_s32(svbool_t pg,const int32_t * base,int64_t vnum)134 svint32_t test_svld1_vnum_s32(svbool_t pg, const int32_t *base, int64_t vnum)
135 {
136 // CHECK-LABEL: test_svld1_vnum_s32
137 // CHECK-DAG: %[[PG:.*]] = call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg)
138 // CHECK-DAG: %[[BITCAST:.*]] = bitcast i32* %base to <vscale x 4 x i32>*
139 // CHECK-DAG: %[[GEP:.*]] = getelementptr <vscale x 4 x i32>, <vscale x 4 x i32>* %[[BITCAST]], i64 %vnum, i64 0
140 // CHECK: %[[LOAD:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.ld1.nxv4i32(<vscale x 4 x i1> %[[PG]], i32* %[[GEP]])
141 // CHECK: ret <vscale x 4 x i32> %[[LOAD]]
142 return SVE_ACLE_FUNC(svld1_vnum,_s32,,)(pg, base, vnum);
143 }
144
test_svld1_vnum_s64(svbool_t pg,const int64_t * base,int64_t vnum)145 svint64_t test_svld1_vnum_s64(svbool_t pg, const int64_t *base, int64_t vnum)
146 {
147 // CHECK-LABEL: test_svld1_vnum_s64
148 // CHECK-DAG: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg)
149 // CHECK-DAG: %[[BITCAST:.*]] = bitcast i64* %base to <vscale x 2 x i64>*
150 // CHECK-DAG: %[[GEP:.*]] = getelementptr <vscale x 2 x i64>, <vscale x 2 x i64>* %[[BITCAST]], i64 %vnum, i64 0
151 // CHECK: %[[LOAD:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.ld1.nxv2i64(<vscale x 2 x i1> %[[PG]], i64* %[[GEP]])
152 // CHECK: ret <vscale x 2 x i64> %[[LOAD]]
153 return SVE_ACLE_FUNC(svld1_vnum,_s64,,)(pg, base, vnum);
154 }
155
test_svld1_vnum_u8(svbool_t pg,const uint8_t * base,int64_t vnum)156 svuint8_t test_svld1_vnum_u8(svbool_t pg, const uint8_t *base, int64_t vnum)
157 {
158 // CHECK-LABEL: test_svld1_vnum_u8
159 // CHECK: %[[BITCAST:.*]] = bitcast i8* %base to <vscale x 16 x i8>*
160 // CHECK: %[[GEP:.*]] = getelementptr <vscale x 16 x i8>, <vscale x 16 x i8>* %[[BITCAST]], i64 %vnum, i64 0
161 // CHECK: %[[LOAD:.*]] = call <vscale x 16 x i8> @llvm.aarch64.sve.ld1.nxv16i8(<vscale x 16 x i1> %pg, i8* %[[GEP]])
162 // CHECK: ret <vscale x 16 x i8> %[[LOAD]]
163 return SVE_ACLE_FUNC(svld1_vnum,_u8,,)(pg, base, vnum);
164 }
165
test_svld1_vnum_u16(svbool_t pg,const uint16_t * base,int64_t vnum)166 svuint16_t test_svld1_vnum_u16(svbool_t pg, const uint16_t *base, int64_t vnum)
167 {
168 // CHECK-LABEL: test_svld1_vnum_u16
169 // CHECK-DAG: %[[PG:.*]] = call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> %pg)
170 // CHECK-DAG: %[[BITCAST:.*]] = bitcast i16* %base to <vscale x 8 x i16>*
171 // CHECK-DAG: %[[GEP:.*]] = getelementptr <vscale x 8 x i16>, <vscale x 8 x i16>* %[[BITCAST]], i64 %vnum, i64 0
172 // CHECK: %[[LOAD:.*]] = call <vscale x 8 x i16> @llvm.aarch64.sve.ld1.nxv8i16(<vscale x 8 x i1> %[[PG]], i16* %[[GEP]])
173 // CHECK: ret <vscale x 8 x i16> %[[LOAD]]
174 return SVE_ACLE_FUNC(svld1_vnum,_u16,,)(pg, base, vnum);
175 }
176
test_svld1_vnum_u32(svbool_t pg,const uint32_t * base,int64_t vnum)177 svuint32_t test_svld1_vnum_u32(svbool_t pg, const uint32_t *base, int64_t vnum)
178 {
179 // CHECK-LABEL: test_svld1_vnum_u32
180 // CHECK-DAG: %[[PG:.*]] = call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg)
181 // CHECK-DAG: %[[BITCAST:.*]] = bitcast i32* %base to <vscale x 4 x i32>*
182 // CHECK-DAG: %[[GEP:.*]] = getelementptr <vscale x 4 x i32>, <vscale x 4 x i32>* %[[BITCAST]], i64 %vnum, i64 0
183 // CHECK: %[[LOAD:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.ld1.nxv4i32(<vscale x 4 x i1> %[[PG]], i32* %[[GEP]])
184 // CHECK: ret <vscale x 4 x i32> %[[LOAD]]
185 return SVE_ACLE_FUNC(svld1_vnum,_u32,,)(pg, base, vnum);
186 }
187
test_svld1_vnum_u64(svbool_t pg,const uint64_t * base,int64_t vnum)188 svuint64_t test_svld1_vnum_u64(svbool_t pg, const uint64_t *base, int64_t vnum)
189 {
190 // CHECK-LABEL: test_svld1_vnum_u64
191 // CHECK-DAG: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg)
192 // CHECK-DAG: %[[BITCAST:.*]] = bitcast i64* %base to <vscale x 2 x i64>*
193 // CHECK-DAG: %[[GEP:.*]] = getelementptr <vscale x 2 x i64>, <vscale x 2 x i64>* %[[BITCAST]], i64 %vnum, i64 0
194 // CHECK: %[[LOAD:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.ld1.nxv2i64(<vscale x 2 x i1> %[[PG]], i64* %[[GEP]])
195 // CHECK: ret <vscale x 2 x i64> %[[LOAD]]
196 return SVE_ACLE_FUNC(svld1_vnum,_u64,,)(pg, base, vnum);
197 }
198
test_svld1_vnum_f16(svbool_t pg,const float16_t * base,int64_t vnum)199 svfloat16_t test_svld1_vnum_f16(svbool_t pg, const float16_t *base, int64_t vnum)
200 {
201 // CHECK-LABEL: test_svld1_vnum_f16
202 // CHECK-DAG: %[[PG:.*]] = call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> %pg)
203 // CHECK-DAG: %[[BITCAST:.*]] = bitcast half* %base to <vscale x 8 x half>*
204 // CHECK-DAG: %[[GEP:.*]] = getelementptr <vscale x 8 x half>, <vscale x 8 x half>* %[[BITCAST]], i64 %vnum, i64 0
205 // CHECK: %[[LOAD:.*]] = call <vscale x 8 x half> @llvm.aarch64.sve.ld1.nxv8f16(<vscale x 8 x i1> %[[PG]], half* %[[GEP]])
206 // CHECK: ret <vscale x 8 x half> %[[LOAD]]
207 return SVE_ACLE_FUNC(svld1_vnum,_f16,,)(pg, base, vnum);
208 }
209
test_svld1_vnum_f32(svbool_t pg,const float32_t * base,int64_t vnum)210 svfloat32_t test_svld1_vnum_f32(svbool_t pg, const float32_t *base, int64_t vnum)
211 {
212 // CHECK-LABEL: test_svld1_vnum_f32
213 // CHECK-DAG: %[[PG:.*]] = call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg)
214 // CHECK-DAG: %[[BITCAST:.*]] = bitcast float* %base to <vscale x 4 x float>*
215 // CHECK-DAG: %[[GEP:.*]] = getelementptr <vscale x 4 x float>, <vscale x 4 x float>* %[[BITCAST]], i64 %vnum, i64 0
216 // CHECK: %[[LOAD:.*]] = call <vscale x 4 x float> @llvm.aarch64.sve.ld1.nxv4f32(<vscale x 4 x i1> %[[PG]], float* %[[GEP]])
217 // CHECK: ret <vscale x 4 x float> %[[LOAD]]
218 return SVE_ACLE_FUNC(svld1_vnum,_f32,,)(pg, base, vnum);
219 }
220
test_svld1_vnum_f64(svbool_t pg,const float64_t * base,int64_t vnum)221 svfloat64_t test_svld1_vnum_f64(svbool_t pg, const float64_t *base, int64_t vnum)
222 {
223 // CHECK-LABEL: test_svld1_vnum_f64
224 // CHECK-DAG: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg)
225 // CHECK-DAG: %[[BITCAST:.*]] = bitcast double* %base to <vscale x 2 x double>*
226 // CHECK-DAG: %[[GEP:.*]] = getelementptr <vscale x 2 x double>, <vscale x 2 x double>* %[[BITCAST]], i64 %vnum, i64 0
227 // CHECK: %[[LOAD:.*]] = call <vscale x 2 x double> @llvm.aarch64.sve.ld1.nxv2f64(<vscale x 2 x i1> %[[PG]], double* %[[GEP]])
228 // CHECK: ret <vscale x 2 x double> %[[LOAD]]
229 return SVE_ACLE_FUNC(svld1_vnum,_f64,,)(pg, base, vnum);
230 }
231
test_svld1_gather_u32base_s32(svbool_t pg,svuint32_t bases)232 svint32_t test_svld1_gather_u32base_s32(svbool_t pg, svuint32_t bases) {
233 // CHECK-LABEL: test_svld1_gather_u32base_s32
234 // CHECK: %[[PG:.*]] = call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg)
235 // CHECK: %[[LOAD:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.ld1.gather.scalar.offset.nxv4i32.nxv4i32(<vscale x 4 x i1> %[[PG]], <vscale x 4 x i32> %bases, i64 0)
236 // CHECK: ret <vscale x 4 x i32> %[[LOAD]]
237 return SVE_ACLE_FUNC(svld1_gather, _u32base, _s32, )(pg, bases);
238 }
239
test_svld1_gather_u64base_s64(svbool_t pg,svuint64_t bases)240 svint64_t test_svld1_gather_u64base_s64(svbool_t pg, svuint64_t bases) {
241 // CHECK-LABEL: test_svld1_gather_u64base_s64
242 // CHECK: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg)
243 // CHECK: %[[LOAD:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.ld1.gather.scalar.offset.nxv2i64.nxv2i64(<vscale x 2 x i1> %[[PG]], <vscale x 2 x i64> %bases, i64 0)
244 // CHECK: ret <vscale x 2 x i64> %[[LOAD]]
245 return SVE_ACLE_FUNC(svld1_gather, _u64base, _s64, )(pg, bases);
246 }
247
test_svld1_gather_u32base_u32(svbool_t pg,svuint32_t bases)248 svuint32_t test_svld1_gather_u32base_u32(svbool_t pg, svuint32_t bases) {
249 // CHECK-LABEL: test_svld1_gather_u32base_u32
250 // CHECK: %[[PG:.*]] = call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg)
251 // CHECK: %[[LOAD:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.ld1.gather.scalar.offset.nxv4i32.nxv4i32(<vscale x 4 x i1> %[[PG]], <vscale x 4 x i32> %bases, i64 0)
252 // CHECK: ret <vscale x 4 x i32> %[[LOAD]]
253 return SVE_ACLE_FUNC(svld1_gather, _u32base, _u32, )(pg, bases);
254 }
255
test_svld1_gather_u64base_u64(svbool_t pg,svuint64_t bases)256 svuint64_t test_svld1_gather_u64base_u64(svbool_t pg, svuint64_t bases) {
257 // CHECK-LABEL: test_svld1_gather_u64base_u64
258 // CHECK: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg)
259 // CHECK: %[[LOAD:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.ld1.gather.scalar.offset.nxv2i64.nxv2i64(<vscale x 2 x i1> %[[PG]], <vscale x 2 x i64> %bases, i64 0)
260 // CHECK: ret <vscale x 2 x i64> %[[LOAD]]
261 return SVE_ACLE_FUNC(svld1_gather, _u64base, _u64, )(pg, bases);
262 }
263
test_svld1_gather_u32base_f32(svbool_t pg,svuint32_t bases)264 svfloat32_t test_svld1_gather_u32base_f32(svbool_t pg, svuint32_t bases) {
265 // CHECK-LABEL: test_svld1_gather_u32base_f32
266 // CHECK: %[[PG:.*]] = call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg)
267 // CHECK: %[[LOAD:.*]] = call <vscale x 4 x float> @llvm.aarch64.sve.ld1.gather.scalar.offset.nxv4f32.nxv4i32(<vscale x 4 x i1> %[[PG]], <vscale x 4 x i32> %bases, i64 0)
268 // CHECK: ret <vscale x 4 x float> %[[LOAD]]
269 return SVE_ACLE_FUNC(svld1_gather, _u32base, _f32, )(pg, bases);
270 }
271
test_svld1_gather_u64base_f64(svbool_t pg,svuint64_t bases)272 svfloat64_t test_svld1_gather_u64base_f64(svbool_t pg, svuint64_t bases) {
273 // CHECK-LABEL: test_svld1_gather_u64base_f64
274 // CHECK: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg)
275 // CHECK: %[[LOAD:.*]] = call <vscale x 2 x double> @llvm.aarch64.sve.ld1.gather.scalar.offset.nxv2f64.nxv2i64(<vscale x 2 x i1> %[[PG]], <vscale x 2 x i64> %bases, i64 0)
276 // CHECK: ret <vscale x 2 x double> %[[LOAD]]
277 return SVE_ACLE_FUNC(svld1_gather, _u64base, _f64, )(pg, bases);
278 }
279
test_svld1_gather_s32offset_s32(svbool_t pg,const int32_t * base,svint32_t offsets)280 svint32_t test_svld1_gather_s32offset_s32(svbool_t pg, const int32_t *base, svint32_t offsets) {
281 // CHECK-LABEL: test_svld1_gather_s32offset_s32
282 // CHECK: %[[PG:.*]] = call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg)
283 // CHECK: %[[LOAD:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.ld1.gather.sxtw.nxv4i32(<vscale x 4 x i1> %[[PG]], i32* %base, <vscale x 4 x i32> %offsets)
284 // CHECK: ret <vscale x 4 x i32> %[[LOAD]]
285 return SVE_ACLE_FUNC(svld1_gather_, s32, offset, _s32)(pg, base, offsets);
286 }
287
test_svld1_gather_s64offset_s64(svbool_t pg,const int64_t * base,svint64_t offsets)288 svint64_t test_svld1_gather_s64offset_s64(svbool_t pg, const int64_t *base, svint64_t offsets) {
289 // CHECK-LABEL: test_svld1_gather_s64offset_s64
290 // CHECK: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg)
291 // CHECK: %[[LOAD:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.ld1.gather.nxv2i64(<vscale x 2 x i1> %[[PG]], i64* %base, <vscale x 2 x i64> %offsets)
292 // CHECK: ret <vscale x 2 x i64> %[[LOAD]]
293 return SVE_ACLE_FUNC(svld1_gather_, s64, offset, _s64)(pg, base, offsets);
294 }
295
test_svld1_gather_s32offset_u32(svbool_t pg,const uint32_t * base,svint32_t offsets)296 svuint32_t test_svld1_gather_s32offset_u32(svbool_t pg, const uint32_t *base, svint32_t offsets) {
297 // CHECK-LABEL: test_svld1_gather_s32offset_u32
298 // CHECK: %[[PG:.*]] = call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg)
299 // CHECK: %[[LOAD:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.ld1.gather.sxtw.nxv4i32(<vscale x 4 x i1> %[[PG]], i32* %base, <vscale x 4 x i32> %offsets)
300 // CHECK: ret <vscale x 4 x i32> %[[LOAD]]
301 return SVE_ACLE_FUNC(svld1_gather_, s32, offset, _u32)(pg, base, offsets);
302 }
303
test_svld1_gather_s64offset_u64(svbool_t pg,const uint64_t * base,svint64_t offsets)304 svuint64_t test_svld1_gather_s64offset_u64(svbool_t pg, const uint64_t *base, svint64_t offsets) {
305 // CHECK-LABEL: test_svld1_gather_s64offset_u64
306 // CHECK: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg)
307 // CHECK: %[[LOAD:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.ld1.gather.nxv2i64(<vscale x 2 x i1> %[[PG]], i64* %base, <vscale x 2 x i64> %offsets)
308 // CHECK: ret <vscale x 2 x i64> %[[LOAD]]
309 return SVE_ACLE_FUNC(svld1_gather_, s64, offset, _u64)(pg, base, offsets);
310 }
311
test_svld1_gather_s32offset_f32(svbool_t pg,const float32_t * base,svint32_t offsets)312 svfloat32_t test_svld1_gather_s32offset_f32(svbool_t pg, const float32_t *base, svint32_t offsets) {
313 // CHECK-LABEL: test_svld1_gather_s32offset_f32
314 // CHECK: %[[PG:.*]] = call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg)
315 // CHECK: %[[LOAD:.*]] = call <vscale x 4 x float> @llvm.aarch64.sve.ld1.gather.sxtw.nxv4f32(<vscale x 4 x i1> %[[PG]], float* %base, <vscale x 4 x i32> %offsets)
316 // CHECK: ret <vscale x 4 x float> %[[LOAD]]
317 return SVE_ACLE_FUNC(svld1_gather_, s32, offset, _f32)(pg, base, offsets);
318 }
319
test_svld1_gather_s64offset_f64(svbool_t pg,const float64_t * base,svint64_t offsets)320 svfloat64_t test_svld1_gather_s64offset_f64(svbool_t pg, const float64_t *base, svint64_t offsets) {
321 // CHECK-LABEL: test_svld1_gather_s64offset_f64
322 // CHECK: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg)
323 // CHECK: %[[LOAD:.*]] = call <vscale x 2 x double> @llvm.aarch64.sve.ld1.gather.nxv2f64(<vscale x 2 x i1> %[[PG]], double* %base, <vscale x 2 x i64> %offsets)
324 // CHECK: ret <vscale x 2 x double> %[[LOAD]]
325 return SVE_ACLE_FUNC(svld1_gather_, s64, offset, _f64)(pg, base, offsets);
326 }
327
test_svld1_gather_u32offset_s32(svbool_t pg,const int32_t * base,svuint32_t offsets)328 svint32_t test_svld1_gather_u32offset_s32(svbool_t pg, const int32_t *base, svuint32_t offsets) {
329 // CHECK-LABEL: test_svld1_gather_u32offset_s32
330 // CHECK: %[[PG:.*]] = call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg)
331 // CHECK: %[[LOAD:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.ld1.gather.uxtw.nxv4i32(<vscale x 4 x i1> %[[PG]], i32* %base, <vscale x 4 x i32> %offsets)
332 // CHECK: ret <vscale x 4 x i32> %[[LOAD]]
333 return SVE_ACLE_FUNC(svld1_gather_, u32, offset, _s32)(pg, base, offsets);
334 }
335
test_svld1_gather_u64offset_s64(svbool_t pg,const int64_t * base,svuint64_t offsets)336 svint64_t test_svld1_gather_u64offset_s64(svbool_t pg, const int64_t *base, svuint64_t offsets) {
337 // CHECK-LABEL: test_svld1_gather_u64offset_s64
338 // CHECK: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg)
339 // CHECK: %[[LOAD:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.ld1.gather.nxv2i64(<vscale x 2 x i1> %[[PG]], i64* %base, <vscale x 2 x i64> %offsets)
340 // CHECK: ret <vscale x 2 x i64> %[[LOAD]]
341 return SVE_ACLE_FUNC(svld1_gather_, u64, offset, _s64)(pg, base, offsets);
342 }
343
test_svld1_gather_u32offset_u32(svbool_t pg,const uint32_t * base,svuint32_t offsets)344 svuint32_t test_svld1_gather_u32offset_u32(svbool_t pg, const uint32_t *base, svuint32_t offsets) {
345 // CHECK-LABEL: test_svld1_gather_u32offset_u32
346 // CHECK: %[[PG:.*]] = call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg)
347 // CHECK: %[[LOAD:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.ld1.gather.uxtw.nxv4i32(<vscale x 4 x i1> %[[PG]], i32* %base, <vscale x 4 x i32> %offsets)
348 // CHECK: ret <vscale x 4 x i32> %[[LOAD]]
349 return SVE_ACLE_FUNC(svld1_gather_, u32, offset, _u32)(pg, base, offsets);
350 }
351
test_svld1_gather_u64offset_u64(svbool_t pg,const uint64_t * base,svuint64_t offsets)352 svuint64_t test_svld1_gather_u64offset_u64(svbool_t pg, const uint64_t *base, svuint64_t offsets) {
353 // CHECK-LABEL: test_svld1_gather_u64offset_u64
354 // CHECK: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg)
355 // CHECK: %[[LOAD:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.ld1.gather.nxv2i64(<vscale x 2 x i1> %[[PG]], i64* %base, <vscale x 2 x i64> %offsets)
356 // CHECK: ret <vscale x 2 x i64> %[[LOAD]]
357 return SVE_ACLE_FUNC(svld1_gather_, u64, offset, _u64)(pg, base, offsets);
358 }
359
test_svld1_gather_u32offset_f32(svbool_t pg,const float32_t * base,svuint32_t offsets)360 svfloat32_t test_svld1_gather_u32offset_f32(svbool_t pg, const float32_t *base, svuint32_t offsets) {
361 // CHECK-LABEL: test_svld1_gather_u32offset_f32
362 // CHECK: %[[PG:.*]] = call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg)
363 // CHECK: %[[LOAD:.*]] = call <vscale x 4 x float> @llvm.aarch64.sve.ld1.gather.uxtw.nxv4f32(<vscale x 4 x i1> %[[PG]], float* %base, <vscale x 4 x i32> %offsets)
364 // CHECK: ret <vscale x 4 x float> %[[LOAD]]
365 return SVE_ACLE_FUNC(svld1_gather_, u32, offset, _f32)(pg, base, offsets);
366 }
367
test_svld1_gather_u64offset_f64(svbool_t pg,const float64_t * base,svuint64_t offsets)368 svfloat64_t test_svld1_gather_u64offset_f64(svbool_t pg, const float64_t *base, svuint64_t offsets) {
369 // CHECK-LABEL: test_svld1_gather_u64offset_f64
370 // CHECK: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg)
371 // CHECK: %[[LOAD:.*]] = call <vscale x 2 x double> @llvm.aarch64.sve.ld1.gather.nxv2f64(<vscale x 2 x i1> %[[PG]], double* %base, <vscale x 2 x i64> %offsets)
372 // CHECK: ret <vscale x 2 x double> %[[LOAD]]
373 return SVE_ACLE_FUNC(svld1_gather_, u64, offset, _f64)(pg, base, offsets);
374 }
375
test_svld1_gather_u32base_offset_s32(svbool_t pg,svuint32_t bases,int64_t offset)376 svint32_t test_svld1_gather_u32base_offset_s32(svbool_t pg, svuint32_t bases, int64_t offset) {
377 // CHECK-LABEL: test_svld1_gather_u32base_offset_s32
378 // CHECK: %[[PG:.*]] = call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg)
379 // CHECK: %[[LOAD:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.ld1.gather.scalar.offset.nxv4i32.nxv4i32(<vscale x 4 x i1> %[[PG]], <vscale x 4 x i32> %bases, i64 %offset)
380 // CHECK: ret <vscale x 4 x i32> %[[LOAD]]
381 return SVE_ACLE_FUNC(svld1_gather, _u32base, _offset_s32, )(pg, bases, offset);
382 }
383
test_svld1_gather_u64base_offset_s64(svbool_t pg,svuint64_t bases,int64_t offset)384 svint64_t test_svld1_gather_u64base_offset_s64(svbool_t pg, svuint64_t bases, int64_t offset) {
385 // CHECK-LABEL: test_svld1_gather_u64base_offset_s64
386 // CHECK: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg)
387 // CHECK: %[[LOAD:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.ld1.gather.scalar.offset.nxv2i64.nxv2i64(<vscale x 2 x i1> %[[PG]], <vscale x 2 x i64> %bases, i64 %offset)
388 // CHECK: ret <vscale x 2 x i64> %[[LOAD]]
389 return SVE_ACLE_FUNC(svld1_gather, _u64base, _offset_s64, )(pg, bases, offset);
390 }
391
test_svld1_gather_u32base_offset_u32(svbool_t pg,svuint32_t bases,int64_t offset)392 svuint32_t test_svld1_gather_u32base_offset_u32(svbool_t pg, svuint32_t bases, int64_t offset) {
393 // CHECK-LABEL: test_svld1_gather_u32base_offset_u32
394 // CHECK: %[[PG:.*]] = call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg)
395 // CHECK: %[[LOAD:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.ld1.gather.scalar.offset.nxv4i32.nxv4i32(<vscale x 4 x i1> %[[PG]], <vscale x 4 x i32> %bases, i64 %offset)
396 // CHECK: ret <vscale x 4 x i32> %[[LOAD]]
397 return SVE_ACLE_FUNC(svld1_gather, _u32base, _offset_u32, )(pg, bases, offset);
398 }
399
test_svld1_gather_u64base_offset_u64(svbool_t pg,svuint64_t bases,int64_t offset)400 svuint64_t test_svld1_gather_u64base_offset_u64(svbool_t pg, svuint64_t bases, int64_t offset) {
401 // CHECK-LABEL: test_svld1_gather_u64base_offset_u64
402 // CHECK: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg)
403 // CHECK: %[[LOAD:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.ld1.gather.scalar.offset.nxv2i64.nxv2i64(<vscale x 2 x i1> %[[PG]], <vscale x 2 x i64> %bases, i64 %offset)
404 // CHECK: ret <vscale x 2 x i64> %[[LOAD]]
405 return SVE_ACLE_FUNC(svld1_gather, _u64base, _offset_u64, )(pg, bases, offset);
406 }
407
test_svld1_gather_u32base_offset_f32(svbool_t pg,svuint32_t bases,int64_t offset)408 svfloat32_t test_svld1_gather_u32base_offset_f32(svbool_t pg, svuint32_t bases, int64_t offset) {
409 // CHECK-LABEL: test_svld1_gather_u32base_offset_f32
410 // CHECK: %[[PG:.*]] = call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg)
411 // CHECK: %[[LOAD:.*]] = call <vscale x 4 x float> @llvm.aarch64.sve.ld1.gather.scalar.offset.nxv4f32.nxv4i32(<vscale x 4 x i1> %[[PG]], <vscale x 4 x i32> %bases, i64 %offset)
412 // CHECK: ret <vscale x 4 x float> %[[LOAD]]
413 return SVE_ACLE_FUNC(svld1_gather, _u32base, _offset_f32, )(pg, bases, offset);
414 }
415
test_svld1_gather_u64base_offset_f64(svbool_t pg,svuint64_t bases,int64_t offset)416 svfloat64_t test_svld1_gather_u64base_offset_f64(svbool_t pg, svuint64_t bases, int64_t offset) {
417 // CHECK-LABEL: test_svld1_gather_u64base_offset_f64
418 // CHECK: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg)
419 // CHECK: %[[LOAD:.*]] = call <vscale x 2 x double> @llvm.aarch64.sve.ld1.gather.scalar.offset.nxv2f64.nxv2i64(<vscale x 2 x i1> %[[PG]], <vscale x 2 x i64> %bases, i64 %offset)
420 // CHECK: ret <vscale x 2 x double> %[[LOAD]]
421 return SVE_ACLE_FUNC(svld1_gather, _u64base, _offset_f64, )(pg, bases, offset);
422 }
423
test_svld1_gather_s32index_s32(svbool_t pg,const int32_t * base,svint32_t indices)424 svint32_t test_svld1_gather_s32index_s32(svbool_t pg, const int32_t *base, svint32_t indices) {
425 // CHECK-LABEL: test_svld1_gather_s32index_s32
426 // CHECK: %[[PG:.*]] = call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg)
427 // CHECK: %[[LOAD:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.ld1.gather.sxtw.index.nxv4i32(<vscale x 4 x i1> %[[PG]], i32* %base, <vscale x 4 x i32> %indices)
428 // CHECK: ret <vscale x 4 x i32> %[[LOAD]]
429 return SVE_ACLE_FUNC(svld1_gather_, s32, index, _s32)(pg, base, indices);
430 }
431
test_svld1_gather_s64index_s64(svbool_t pg,const int64_t * base,svint64_t indices)432 svint64_t test_svld1_gather_s64index_s64(svbool_t pg, const int64_t *base, svint64_t indices) {
433 // CHECK-LABEL: test_svld1_gather_s64index_s64
434 // CHECK: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg)
435 // CHECK: %[[LOAD:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.ld1.gather.index.nxv2i64(<vscale x 2 x i1> %[[PG]], i64* %base, <vscale x 2 x i64> %indices)
436 // CHECK: ret <vscale x 2 x i64> %[[LOAD]]
437 return SVE_ACLE_FUNC(svld1_gather_, s64, index, _s64)(pg, base, indices);
438 }
439
test_svld1_gather_s32index_u32(svbool_t pg,const uint32_t * base,svint32_t indices)440 svuint32_t test_svld1_gather_s32index_u32(svbool_t pg, const uint32_t *base, svint32_t indices) {
441 // CHECK-LABEL: test_svld1_gather_s32index_u32
442 // CHECK: %[[PG:.*]] = call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg)
443 // CHECK: %[[LOAD:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.ld1.gather.sxtw.index.nxv4i32(<vscale x 4 x i1> %[[PG]], i32* %base, <vscale x 4 x i32> %indices)
444 // CHECK: ret <vscale x 4 x i32> %[[LOAD]]
445 return SVE_ACLE_FUNC(svld1_gather_, s32, index, _u32)(pg, base, indices);
446 }
447
test_svld1_gather_s64index_u64(svbool_t pg,const uint64_t * base,svint64_t indices)448 svuint64_t test_svld1_gather_s64index_u64(svbool_t pg, const uint64_t *base, svint64_t indices) {
449 // CHECK-LABEL: test_svld1_gather_s64index_u64
450 // CHECK: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg)
451 // CHECK: %[[LOAD:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.ld1.gather.index.nxv2i64(<vscale x 2 x i1> %[[PG]], i64* %base, <vscale x 2 x i64> %indices)
452 // CHECK: ret <vscale x 2 x i64> %[[LOAD]]
453 return SVE_ACLE_FUNC(svld1_gather_, s64, index, _u64)(pg, base, indices);
454 }
455
test_svld1_gather_s32index_f32(svbool_t pg,const float32_t * base,svint32_t indices)456 svfloat32_t test_svld1_gather_s32index_f32(svbool_t pg, const float32_t *base, svint32_t indices) {
457 // CHECK-LABEL: test_svld1_gather_s32index_f32
458 // CHECK: %[[PG:.*]] = call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg)
459 // CHECK: %[[LOAD:.*]] = call <vscale x 4 x float> @llvm.aarch64.sve.ld1.gather.sxtw.index.nxv4f32(<vscale x 4 x i1> %[[PG]], float* %base, <vscale x 4 x i32> %indices)
460 // CHECK: ret <vscale x 4 x float> %[[LOAD]]
461 return SVE_ACLE_FUNC(svld1_gather_, s32, index, _f32)(pg, base, indices);
462 }
463
test_svld1_gather_s64index_f64(svbool_t pg,const float64_t * base,svint64_t indices)464 svfloat64_t test_svld1_gather_s64index_f64(svbool_t pg, const float64_t *base, svint64_t indices) {
465 // CHECK-LABEL: test_svld1_gather_s64index_f64
466 // CHECK: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg)
467 // CHECK: %[[LOAD:.*]] = call <vscale x 2 x double> @llvm.aarch64.sve.ld1.gather.index.nxv2f64(<vscale x 2 x i1> %[[PG]], double* %base, <vscale x 2 x i64> %indices)
468 // CHECK: ret <vscale x 2 x double> %[[LOAD]]
469 return SVE_ACLE_FUNC(svld1_gather_, s64, index, _f64)(pg, base, indices);
470 }
471
test_svld1_gather_u32index_s32(svbool_t pg,const int32_t * base,svuint32_t indices)472 svint32_t test_svld1_gather_u32index_s32(svbool_t pg, const int32_t *base, svuint32_t indices) {
473 // CHECK-LABEL: test_svld1_gather_u32index_s32
474 // CHECK: %[[PG:.*]] = call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg)
475 // CHECK: %[[LOAD:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.ld1.gather.uxtw.index.nxv4i32(<vscale x 4 x i1> %[[PG]], i32* %base, <vscale x 4 x i32> %indices)
476 // CHECK: ret <vscale x 4 x i32> %[[LOAD]]
477 return SVE_ACLE_FUNC(svld1_gather_, u32, index, _s32)(pg, base, indices);
478 }
479
test_svld1_gather_u64index_s64(svbool_t pg,const int64_t * base,svuint64_t indices)480 svint64_t test_svld1_gather_u64index_s64(svbool_t pg, const int64_t *base, svuint64_t indices) {
481 // CHECK-LABEL: test_svld1_gather_u64index_s64
482 // CHECK: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg)
483 // CHECK: %[[LOAD:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.ld1.gather.index.nxv2i64(<vscale x 2 x i1> %[[PG]], i64* %base, <vscale x 2 x i64> %indices)
484 // CHECK: ret <vscale x 2 x i64> %[[LOAD]]
485 return SVE_ACLE_FUNC(svld1_gather_, u64, index, _s64)(pg, base, indices);
486 }
487
test_svld1_gather_u32index_u32(svbool_t pg,const uint32_t * base,svuint32_t indices)488 svuint32_t test_svld1_gather_u32index_u32(svbool_t pg, const uint32_t *base, svuint32_t indices) {
489 // CHECK-LABEL: test_svld1_gather_u32index_u32
490 // CHECK: %[[PG:.*]] = call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg)
491 // CHECK: %[[LOAD:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.ld1.gather.uxtw.index.nxv4i32(<vscale x 4 x i1> %[[PG]], i32* %base, <vscale x 4 x i32> %indices)
492 // CHECK: ret <vscale x 4 x i32> %[[LOAD]]
493 return SVE_ACLE_FUNC(svld1_gather_, u32, index, _u32)(pg, base, indices);
494 }
495
test_svld1_gather_u64index_u64(svbool_t pg,const uint64_t * base,svuint64_t indices)496 svuint64_t test_svld1_gather_u64index_u64(svbool_t pg, const uint64_t *base, svuint64_t indices) {
497 // CHECK-LABEL: test_svld1_gather_u64index_u64
498 // CHECK: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg)
499 // CHECK: %[[LOAD:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.ld1.gather.index.nxv2i64(<vscale x 2 x i1> %[[PG]], i64* %base, <vscale x 2 x i64> %indices)
500 // CHECK: ret <vscale x 2 x i64> %[[LOAD]]
501 return SVE_ACLE_FUNC(svld1_gather_, u64, index, _u64)(pg, base, indices);
502 }
503
test_svld1_gather_u32index_f32(svbool_t pg,const float32_t * base,svuint32_t indices)504 svfloat32_t test_svld1_gather_u32index_f32(svbool_t pg, const float32_t *base, svuint32_t indices) {
505 // CHECK-LABEL: test_svld1_gather_u32index_f32
506 // CHECK: %[[PG:.*]] = call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg)
507 // CHECK: %[[LOAD:.*]] = call <vscale x 4 x float> @llvm.aarch64.sve.ld1.gather.uxtw.index.nxv4f32(<vscale x 4 x i1> %[[PG]], float* %base, <vscale x 4 x i32> %indices)
508 // CHECK: ret <vscale x 4 x float> %[[LOAD]]
509 return SVE_ACLE_FUNC(svld1_gather_, u32, index, _f32)(pg, base, indices);
510 }
511
test_svld1_gather_u64index_f64(svbool_t pg,const float64_t * base,svuint64_t indices)512 svfloat64_t test_svld1_gather_u64index_f64(svbool_t pg, const float64_t *base, svuint64_t indices) {
513 // CHECK-LABEL: test_svld1_gather_u64index_f64
514 // CHECK: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg)
515 // CHECK: %[[LOAD:.*]] = call <vscale x 2 x double> @llvm.aarch64.sve.ld1.gather.index.nxv2f64(<vscale x 2 x i1> %[[PG]], double* %base, <vscale x 2 x i64> %indices)
516 // CHECK: ret <vscale x 2 x double> %[[LOAD]]
517 return SVE_ACLE_FUNC(svld1_gather_, u64, index, _f64)(pg, base, indices);
518 }
519
test_svld1_gather_u32base_index_s32(svbool_t pg,svuint32_t bases,int64_t index)520 svint32_t test_svld1_gather_u32base_index_s32(svbool_t pg, svuint32_t bases, int64_t index) {
521 // CHECK-LABEL: test_svld1_gather_u32base_index_s32
522 // CHECK-DAG: %[[PG:.*]] = call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg)
523 // CHECK-DAG: %[[SHL:.*]] = shl i64 %index, 2
524 // CHECK: %[[LOAD:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.ld1.gather.scalar.offset.nxv4i32.nxv4i32(<vscale x 4 x i1> %[[PG]], <vscale x 4 x i32> %bases, i64 %[[SHL]])
525 // CHECK: ret <vscale x 4 x i32> %[[LOAD]]
526 return SVE_ACLE_FUNC(svld1_gather, _u32base, _index_s32, )(pg, bases, index);
527 }
528
test_svld1_gather_u64base_index_s64(svbool_t pg,svuint64_t bases,int64_t index)529 svint64_t test_svld1_gather_u64base_index_s64(svbool_t pg, svuint64_t bases, int64_t index) {
530 // CHECK-LABEL: test_svld1_gather_u64base_index_s64
531 // CHECK-DAG: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg)
532 // CHECK-DAG: %[[SHL:.*]] = shl i64 %index, 3
533 // CHECK: %[[LOAD:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.ld1.gather.scalar.offset.nxv2i64.nxv2i64(<vscale x 2 x i1> %[[PG]], <vscale x 2 x i64> %bases, i64 %[[SHL]])
534 // CHECK: ret <vscale x 2 x i64> %[[LOAD]]
535 return SVE_ACLE_FUNC(svld1_gather, _u64base, _index_s64, )(pg, bases, index);
536 }
537
test_svld1_gather_u32base_index_u32(svbool_t pg,svuint32_t bases,int64_t index)538 svuint32_t test_svld1_gather_u32base_index_u32(svbool_t pg, svuint32_t bases, int64_t index) {
539 // CHECK-LABEL: test_svld1_gather_u32base_index_u32
540 // CHECK-DAG: %[[PG:.*]] = call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg)
541 // CHECK-DAG: %[[SHL:.*]] = shl i64 %index, 2
542 // CHECK: %[[LOAD:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.ld1.gather.scalar.offset.nxv4i32.nxv4i32(<vscale x 4 x i1> %[[PG]], <vscale x 4 x i32> %bases, i64 %[[SHL]])
543 // CHECK: ret <vscale x 4 x i32> %[[LOAD]]
544 return SVE_ACLE_FUNC(svld1_gather, _u32base, _index_u32, )(pg, bases, index);
545 }
546
test_svld1_gather_u64base_index_u64(svbool_t pg,svuint64_t bases,int64_t index)547 svuint64_t test_svld1_gather_u64base_index_u64(svbool_t pg, svuint64_t bases, int64_t index) {
548 // CHECK-LABEL: test_svld1_gather_u64base_index_u64
549 // CHECK-DAG: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg)
550 // CHECK-DAG: %[[SHL:.*]] = shl i64 %index, 3
551 // CHECK: %[[LOAD:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.ld1.gather.scalar.offset.nxv2i64.nxv2i64(<vscale x 2 x i1> %[[PG]], <vscale x 2 x i64> %bases, i64 %[[SHL]])
552 // CHECK: ret <vscale x 2 x i64> %[[LOAD]]
553 return SVE_ACLE_FUNC(svld1_gather, _u64base, _index_u64, )(pg, bases, index);
554 }
555
test_svld1_gather_u32base_index_f32(svbool_t pg,svuint32_t bases,int64_t index)556 svfloat32_t test_svld1_gather_u32base_index_f32(svbool_t pg, svuint32_t bases, int64_t index) {
557 // CHECK-LABEL: test_svld1_gather_u32base_index_f32
558 // CHECK-DAG: %[[PG:.*]] = call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg)
559 // CHECK-DAG: %[[SHL:.*]] = shl i64 %index, 2
560 // CHECK: %[[LOAD:.*]] = call <vscale x 4 x float> @llvm.aarch64.sve.ld1.gather.scalar.offset.nxv4f32.nxv4i32(<vscale x 4 x i1> %[[PG]], <vscale x 4 x i32> %bases, i64 %[[SHL]])
561 // CHECK: ret <vscale x 4 x float> %[[LOAD]]
562 return SVE_ACLE_FUNC(svld1_gather, _u32base, _index_f32, )(pg, bases, index);
563 }
564
test_svld1_gather_u64base_index_f64(svbool_t pg,svuint64_t bases,int64_t index)565 svfloat64_t test_svld1_gather_u64base_index_f64(svbool_t pg, svuint64_t bases, int64_t index) {
566 // CHECK-LABEL: test_svld1_gather_u64base_index_f64
567 // CHECK-DAG: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg)
568 // CHECK-DAG: %[[SHL:.*]] = shl i64 %index, 3
569 // CHECK: %[[LOAD:.*]] = call <vscale x 2 x double> @llvm.aarch64.sve.ld1.gather.scalar.offset.nxv2f64.nxv2i64(<vscale x 2 x i1> %[[PG]], <vscale x 2 x i64> %bases, i64 %[[SHL]])
570 // CHECK: ret <vscale x 2 x double> %[[LOAD]]
571 return SVE_ACLE_FUNC(svld1_gather, _u64base, _index_f64, )(pg, bases, index);
572 }
573