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 -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
4 // RUN: %clang_cc1 -triple aarch64-none-linux-gnu -target-feature +sve -fallow-half-arguments-and-returns -S -O1 -Werror -o - %s >/dev/null 2>%t
5 // RUN: FileCheck --check-prefix=ASM --allow-empty %s <%t
6
7 // If this check fails please read test/CodeGen/aarch64-sve-intrinsics/README for instructions on how to resolve it.
8 // ASM-NOT: warning
9 #include <arm_sve.h>
10
11 #ifdef SVE_OVERLOADED_FORMS
12 // A simple used,unused... macro, long enough to represent any SVE builtin.
13 #define SVE_ACLE_FUNC(A1,A2_UNUSED,A3,A4_UNUSED) A1##A3
14 #else
15 #define SVE_ACLE_FUNC(A1,A2,A3,A4) A1##A2##A3##A4
16 #endif
17
test_svld1_s8(svbool_t pg,const int8_t * base)18 svint8_t test_svld1_s8(svbool_t pg, const int8_t *base)
19 {
20 // CHECK-LABEL: test_svld1_s8
21 // CHECK: %[[LOAD:.*]] = call <vscale x 16 x i8> @llvm.aarch64.sve.ld1.nxv16i8(<vscale x 16 x i1> %pg, i8* %base)
22 // CHECK: ret <vscale x 16 x i8> %[[LOAD]]
23 return SVE_ACLE_FUNC(svld1,_s8,,)(pg, base);
24 }
25
test_svld1_s16(svbool_t pg,const int16_t * base)26 svint16_t test_svld1_s16(svbool_t pg, const int16_t *base)
27 {
28 // CHECK-LABEL: test_svld1_s16
29 // CHECK: %[[PG:.*]] = call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> %pg)
30 // CHECK: %[[LOAD:.*]] = call <vscale x 8 x i16> @llvm.aarch64.sve.ld1.nxv8i16(<vscale x 8 x i1> %[[PG]], i16* %base)
31 // CHECK: ret <vscale x 8 x i16> %[[LOAD]]
32 return SVE_ACLE_FUNC(svld1,_s16,,)(pg, base);
33 }
34
test_svld1_s32(svbool_t pg,const int32_t * base)35 svint32_t test_svld1_s32(svbool_t pg, const int32_t *base)
36 {
37 // CHECK-LABEL: test_svld1_s32
38 // CHECK: %[[PG:.*]] = call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg)
39 // CHECK: %[[LOAD:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.ld1.nxv4i32(<vscale x 4 x i1> %[[PG]], i32* %base)
40 // CHECK: ret <vscale x 4 x i32> %[[LOAD]]
41 return SVE_ACLE_FUNC(svld1,_s32,,)(pg, base);
42 }
43
test_svld1_s64(svbool_t pg,const int64_t * base)44 svint64_t test_svld1_s64(svbool_t pg, const int64_t *base)
45 {
46 // CHECK-LABEL: test_svld1_s64
47 // CHECK: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg)
48 // CHECK: %[[LOAD:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.ld1.nxv2i64(<vscale x 2 x i1> %[[PG]], i64* %base)
49 // CHECK: ret <vscale x 2 x i64> %[[LOAD]]
50 return SVE_ACLE_FUNC(svld1,_s64,,)(pg, base);
51 }
52
test_svld1_u8(svbool_t pg,const uint8_t * base)53 svuint8_t test_svld1_u8(svbool_t pg, const uint8_t *base)
54 {
55 // CHECK-LABEL: test_svld1_u8
56 // CHECK: %[[LOAD:.*]] = call <vscale x 16 x i8> @llvm.aarch64.sve.ld1.nxv16i8(<vscale x 16 x i1> %pg, i8* %base)
57 // CHECK: ret <vscale x 16 x i8> %[[LOAD]]
58 return SVE_ACLE_FUNC(svld1,_u8,,)(pg, base);
59 }
60
test_svld1_u16(svbool_t pg,const uint16_t * base)61 svuint16_t test_svld1_u16(svbool_t pg, const uint16_t *base)
62 {
63 // CHECK-LABEL: test_svld1_u16
64 // CHECK: %[[PG:.*]] = call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> %pg)
65 // CHECK: %[[LOAD:.*]] = call <vscale x 8 x i16> @llvm.aarch64.sve.ld1.nxv8i16(<vscale x 8 x i1> %[[PG]], i16* %base)
66 // CHECK: ret <vscale x 8 x i16> %[[LOAD]]
67 return SVE_ACLE_FUNC(svld1,_u16,,)(pg, base);
68 }
69
test_svld1_u32(svbool_t pg,const uint32_t * base)70 svuint32_t test_svld1_u32(svbool_t pg, const uint32_t *base)
71 {
72 // CHECK-LABEL: test_svld1_u32
73 // CHECK: %[[PG:.*]] = call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg)
74 // CHECK: %[[LOAD:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.ld1.nxv4i32(<vscale x 4 x i1> %[[PG]], i32* %base)
75 // CHECK: ret <vscale x 4 x i32> %[[LOAD]]
76 return SVE_ACLE_FUNC(svld1,_u32,,)(pg, base);
77 }
78
test_svld1_u64(svbool_t pg,const uint64_t * base)79 svuint64_t test_svld1_u64(svbool_t pg, const uint64_t *base)
80 {
81 // CHECK-LABEL: test_svld1_u64
82 // CHECK: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg)
83 // CHECK: %[[LOAD:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.ld1.nxv2i64(<vscale x 2 x i1> %[[PG]], i64* %base)
84 // CHECK: ret <vscale x 2 x i64> %[[LOAD]]
85 return SVE_ACLE_FUNC(svld1,_u64,,)(pg, base);
86 }
87
test_svld1_f16(svbool_t pg,const float16_t * base)88 svfloat16_t test_svld1_f16(svbool_t pg, const float16_t *base)
89 {
90 // CHECK-LABEL: test_svld1_f16
91 // CHECK: %[[PG:.*]] = call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> %pg)
92 // CHECK: %[[LOAD:.*]] = call <vscale x 8 x half> @llvm.aarch64.sve.ld1.nxv8f16(<vscale x 8 x i1> %[[PG]], half* %base)
93 // CHECK: ret <vscale x 8 x half> %[[LOAD]]
94 return SVE_ACLE_FUNC(svld1,_f16,,)(pg, base);
95 }
96
test_svld1_f32(svbool_t pg,const float32_t * base)97 svfloat32_t test_svld1_f32(svbool_t pg, const float32_t *base)
98 {
99 // CHECK-LABEL: test_svld1_f32
100 // CHECK: %[[PG:.*]] = call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg)
101 // CHECK: %[[LOAD:.*]] = call <vscale x 4 x float> @llvm.aarch64.sve.ld1.nxv4f32(<vscale x 4 x i1> %[[PG]], float* %base)
102 // CHECK: ret <vscale x 4 x float> %[[LOAD]]
103 return SVE_ACLE_FUNC(svld1,_f32,,)(pg, base);
104 }
105
test_svld1_f64(svbool_t pg,const float64_t * base)106 svfloat64_t test_svld1_f64(svbool_t pg, const float64_t *base)
107 {
108 // CHECK-LABEL: test_svld1_f64
109 // CHECK: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg)
110 // CHECK: %[[LOAD:.*]] = call <vscale x 2 x double> @llvm.aarch64.sve.ld1.nxv2f64(<vscale x 2 x i1> %[[PG]], double* %base)
111 // CHECK: ret <vscale x 2 x double> %[[LOAD]]
112 return SVE_ACLE_FUNC(svld1,_f64,,)(pg, base);
113 }
114
test_svld1_vnum_s8(svbool_t pg,const int8_t * base,int64_t vnum)115 svint8_t test_svld1_vnum_s8(svbool_t pg, const int8_t *base, int64_t vnum)
116 {
117 // CHECK-LABEL: test_svld1_vnum_s8
118 // CHECK: %[[BITCAST:.*]] = bitcast i8* %base to <vscale x 16 x i8>*
119 // CHECK: %[[GEP:.*]] = getelementptr <vscale x 16 x i8>, <vscale x 16 x i8>* %[[BITCAST]], i64 %vnum, i64 0
120 // CHECK: %[[LOAD:.*]] = call <vscale x 16 x i8> @llvm.aarch64.sve.ld1.nxv16i8(<vscale x 16 x i1> %pg, i8* %[[GEP]])
121 // CHECK: ret <vscale x 16 x i8> %[[LOAD]]
122 return SVE_ACLE_FUNC(svld1_vnum,_s8,,)(pg, base, vnum);
123 }
124
test_svld1_vnum_s16(svbool_t pg,const int16_t * base,int64_t vnum)125 svint16_t test_svld1_vnum_s16(svbool_t pg, const int16_t *base, int64_t vnum)
126 {
127 // CHECK-LABEL: test_svld1_vnum_s16
128 // CHECK-DAG: %[[PG:.*]] = call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> %pg)
129 // CHECK-DAG: %[[BITCAST:.*]] = bitcast i16* %base to <vscale x 8 x i16>*
130 // CHECK-DAG: %[[GEP:.*]] = getelementptr <vscale x 8 x i16>, <vscale x 8 x i16>* %[[BITCAST]], i64 %vnum, i64 0
131 // CHECK: %[[LOAD:.*]] = call <vscale x 8 x i16> @llvm.aarch64.sve.ld1.nxv8i16(<vscale x 8 x i1> %[[PG]], i16* %[[GEP]])
132 // CHECK: ret <vscale x 8 x i16> %[[LOAD]]
133 return SVE_ACLE_FUNC(svld1_vnum,_s16,,)(pg, base, vnum);
134 }
135
test_svld1_vnum_s32(svbool_t pg,const int32_t * base,int64_t vnum)136 svint32_t test_svld1_vnum_s32(svbool_t pg, const int32_t *base, int64_t vnum)
137 {
138 // CHECK-LABEL: test_svld1_vnum_s32
139 // CHECK-DAG: %[[PG:.*]] = call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg)
140 // CHECK-DAG: %[[BITCAST:.*]] = bitcast i32* %base to <vscale x 4 x i32>*
141 // CHECK-DAG: %[[GEP:.*]] = getelementptr <vscale x 4 x i32>, <vscale x 4 x i32>* %[[BITCAST]], i64 %vnum, i64 0
142 // CHECK: %[[LOAD:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.ld1.nxv4i32(<vscale x 4 x i1> %[[PG]], i32* %[[GEP]])
143 // CHECK: ret <vscale x 4 x i32> %[[LOAD]]
144 return SVE_ACLE_FUNC(svld1_vnum,_s32,,)(pg, base, vnum);
145 }
146
test_svld1_vnum_s64(svbool_t pg,const int64_t * base,int64_t vnum)147 svint64_t test_svld1_vnum_s64(svbool_t pg, const int64_t *base, int64_t vnum)
148 {
149 // CHECK-LABEL: test_svld1_vnum_s64
150 // CHECK-DAG: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg)
151 // CHECK-DAG: %[[BITCAST:.*]] = bitcast i64* %base to <vscale x 2 x i64>*
152 // CHECK-DAG: %[[GEP:.*]] = getelementptr <vscale x 2 x i64>, <vscale x 2 x i64>* %[[BITCAST]], i64 %vnum, i64 0
153 // CHECK: %[[LOAD:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.ld1.nxv2i64(<vscale x 2 x i1> %[[PG]], i64* %[[GEP]])
154 // CHECK: ret <vscale x 2 x i64> %[[LOAD]]
155 return SVE_ACLE_FUNC(svld1_vnum,_s64,,)(pg, base, vnum);
156 }
157
test_svld1_vnum_u8(svbool_t pg,const uint8_t * base,int64_t vnum)158 svuint8_t test_svld1_vnum_u8(svbool_t pg, const uint8_t *base, int64_t vnum)
159 {
160 // CHECK-LABEL: test_svld1_vnum_u8
161 // CHECK: %[[BITCAST:.*]] = bitcast i8* %base to <vscale x 16 x i8>*
162 // CHECK: %[[GEP:.*]] = getelementptr <vscale x 16 x i8>, <vscale x 16 x i8>* %[[BITCAST]], i64 %vnum, i64 0
163 // CHECK: %[[LOAD:.*]] = call <vscale x 16 x i8> @llvm.aarch64.sve.ld1.nxv16i8(<vscale x 16 x i1> %pg, i8* %[[GEP]])
164 // CHECK: ret <vscale x 16 x i8> %[[LOAD]]
165 return SVE_ACLE_FUNC(svld1_vnum,_u8,,)(pg, base, vnum);
166 }
167
test_svld1_vnum_u16(svbool_t pg,const uint16_t * base,int64_t vnum)168 svuint16_t test_svld1_vnum_u16(svbool_t pg, const uint16_t *base, int64_t vnum)
169 {
170 // CHECK-LABEL: test_svld1_vnum_u16
171 // CHECK-DAG: %[[PG:.*]] = call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> %pg)
172 // CHECK-DAG: %[[BITCAST:.*]] = bitcast i16* %base to <vscale x 8 x i16>*
173 // CHECK-DAG: %[[GEP:.*]] = getelementptr <vscale x 8 x i16>, <vscale x 8 x i16>* %[[BITCAST]], i64 %vnum, i64 0
174 // CHECK: %[[LOAD:.*]] = call <vscale x 8 x i16> @llvm.aarch64.sve.ld1.nxv8i16(<vscale x 8 x i1> %[[PG]], i16* %[[GEP]])
175 // CHECK: ret <vscale x 8 x i16> %[[LOAD]]
176 return SVE_ACLE_FUNC(svld1_vnum,_u16,,)(pg, base, vnum);
177 }
178
test_svld1_vnum_u32(svbool_t pg,const uint32_t * base,int64_t vnum)179 svuint32_t test_svld1_vnum_u32(svbool_t pg, const uint32_t *base, int64_t vnum)
180 {
181 // CHECK-LABEL: test_svld1_vnum_u32
182 // CHECK-DAG: %[[PG:.*]] = call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg)
183 // CHECK-DAG: %[[BITCAST:.*]] = bitcast i32* %base to <vscale x 4 x i32>*
184 // CHECK-DAG: %[[GEP:.*]] = getelementptr <vscale x 4 x i32>, <vscale x 4 x i32>* %[[BITCAST]], i64 %vnum, i64 0
185 // CHECK: %[[LOAD:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.ld1.nxv4i32(<vscale x 4 x i1> %[[PG]], i32* %[[GEP]])
186 // CHECK: ret <vscale x 4 x i32> %[[LOAD]]
187 return SVE_ACLE_FUNC(svld1_vnum,_u32,,)(pg, base, vnum);
188 }
189
test_svld1_vnum_u64(svbool_t pg,const uint64_t * base,int64_t vnum)190 svuint64_t test_svld1_vnum_u64(svbool_t pg, const uint64_t *base, int64_t vnum)
191 {
192 // CHECK-LABEL: test_svld1_vnum_u64
193 // CHECK-DAG: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg)
194 // CHECK-DAG: %[[BITCAST:.*]] = bitcast i64* %base to <vscale x 2 x i64>*
195 // CHECK-DAG: %[[GEP:.*]] = getelementptr <vscale x 2 x i64>, <vscale x 2 x i64>* %[[BITCAST]], i64 %vnum, i64 0
196 // CHECK: %[[LOAD:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.ld1.nxv2i64(<vscale x 2 x i1> %[[PG]], i64* %[[GEP]])
197 // CHECK: ret <vscale x 2 x i64> %[[LOAD]]
198 return SVE_ACLE_FUNC(svld1_vnum,_u64,,)(pg, base, vnum);
199 }
200
test_svld1_vnum_f16(svbool_t pg,const float16_t * base,int64_t vnum)201 svfloat16_t test_svld1_vnum_f16(svbool_t pg, const float16_t *base, int64_t vnum)
202 {
203 // CHECK-LABEL: test_svld1_vnum_f16
204 // CHECK-DAG: %[[PG:.*]] = call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> %pg)
205 // CHECK-DAG: %[[BITCAST:.*]] = bitcast half* %base to <vscale x 8 x half>*
206 // CHECK-DAG: %[[GEP:.*]] = getelementptr <vscale x 8 x half>, <vscale x 8 x half>* %[[BITCAST]], i64 %vnum, i64 0
207 // CHECK: %[[LOAD:.*]] = call <vscale x 8 x half> @llvm.aarch64.sve.ld1.nxv8f16(<vscale x 8 x i1> %[[PG]], half* %[[GEP]])
208 // CHECK: ret <vscale x 8 x half> %[[LOAD]]
209 return SVE_ACLE_FUNC(svld1_vnum,_f16,,)(pg, base, vnum);
210 }
211
test_svld1_vnum_f32(svbool_t pg,const float32_t * base,int64_t vnum)212 svfloat32_t test_svld1_vnum_f32(svbool_t pg, const float32_t *base, int64_t vnum)
213 {
214 // CHECK-LABEL: test_svld1_vnum_f32
215 // CHECK-DAG: %[[PG:.*]] = call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg)
216 // CHECK-DAG: %[[BITCAST:.*]] = bitcast float* %base to <vscale x 4 x float>*
217 // CHECK-DAG: %[[GEP:.*]] = getelementptr <vscale x 4 x float>, <vscale x 4 x float>* %[[BITCAST]], i64 %vnum, i64 0
218 // CHECK: %[[LOAD:.*]] = call <vscale x 4 x float> @llvm.aarch64.sve.ld1.nxv4f32(<vscale x 4 x i1> %[[PG]], float* %[[GEP]])
219 // CHECK: ret <vscale x 4 x float> %[[LOAD]]
220 return SVE_ACLE_FUNC(svld1_vnum,_f32,,)(pg, base, vnum);
221 }
222
test_svld1_vnum_f64(svbool_t pg,const float64_t * base,int64_t vnum)223 svfloat64_t test_svld1_vnum_f64(svbool_t pg, const float64_t *base, int64_t vnum)
224 {
225 // CHECK-LABEL: test_svld1_vnum_f64
226 // CHECK-DAG: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg)
227 // CHECK-DAG: %[[BITCAST:.*]] = bitcast double* %base to <vscale x 2 x double>*
228 // CHECK-DAG: %[[GEP:.*]] = getelementptr <vscale x 2 x double>, <vscale x 2 x double>* %[[BITCAST]], i64 %vnum, i64 0
229 // CHECK: %[[LOAD:.*]] = call <vscale x 2 x double> @llvm.aarch64.sve.ld1.nxv2f64(<vscale x 2 x i1> %[[PG]], double* %[[GEP]])
230 // CHECK: ret <vscale x 2 x double> %[[LOAD]]
231 return SVE_ACLE_FUNC(svld1_vnum,_f64,,)(pg, base, vnum);
232 }
233
test_svld1_gather_u32base_s32(svbool_t pg,svuint32_t bases)234 svint32_t test_svld1_gather_u32base_s32(svbool_t pg, svuint32_t bases) {
235 // CHECK-LABEL: test_svld1_gather_u32base_s32
236 // CHECK: %[[PG:.*]] = call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg)
237 // 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)
238 // CHECK: ret <vscale x 4 x i32> %[[LOAD]]
239 return SVE_ACLE_FUNC(svld1_gather, _u32base, _s32, )(pg, bases);
240 }
241
test_svld1_gather_u64base_s64(svbool_t pg,svuint64_t bases)242 svint64_t test_svld1_gather_u64base_s64(svbool_t pg, svuint64_t bases) {
243 // CHECK-LABEL: test_svld1_gather_u64base_s64
244 // CHECK: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg)
245 // 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)
246 // CHECK: ret <vscale x 2 x i64> %[[LOAD]]
247 return SVE_ACLE_FUNC(svld1_gather, _u64base, _s64, )(pg, bases);
248 }
249
test_svld1_gather_u32base_u32(svbool_t pg,svuint32_t bases)250 svuint32_t test_svld1_gather_u32base_u32(svbool_t pg, svuint32_t bases) {
251 // CHECK-LABEL: test_svld1_gather_u32base_u32
252 // CHECK: %[[PG:.*]] = call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg)
253 // 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)
254 // CHECK: ret <vscale x 4 x i32> %[[LOAD]]
255 return SVE_ACLE_FUNC(svld1_gather, _u32base, _u32, )(pg, bases);
256 }
257
test_svld1_gather_u64base_u64(svbool_t pg,svuint64_t bases)258 svuint64_t test_svld1_gather_u64base_u64(svbool_t pg, svuint64_t bases) {
259 // CHECK-LABEL: test_svld1_gather_u64base_u64
260 // CHECK: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg)
261 // 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)
262 // CHECK: ret <vscale x 2 x i64> %[[LOAD]]
263 return SVE_ACLE_FUNC(svld1_gather, _u64base, _u64, )(pg, bases);
264 }
265
test_svld1_gather_u32base_f32(svbool_t pg,svuint32_t bases)266 svfloat32_t test_svld1_gather_u32base_f32(svbool_t pg, svuint32_t bases) {
267 // CHECK-LABEL: test_svld1_gather_u32base_f32
268 // CHECK: %[[PG:.*]] = call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg)
269 // 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)
270 // CHECK: ret <vscale x 4 x float> %[[LOAD]]
271 return SVE_ACLE_FUNC(svld1_gather, _u32base, _f32, )(pg, bases);
272 }
273
test_svld1_gather_u64base_f64(svbool_t pg,svuint64_t bases)274 svfloat64_t test_svld1_gather_u64base_f64(svbool_t pg, svuint64_t bases) {
275 // CHECK-LABEL: test_svld1_gather_u64base_f64
276 // CHECK: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg)
277 // 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)
278 // CHECK: ret <vscale x 2 x double> %[[LOAD]]
279 return SVE_ACLE_FUNC(svld1_gather, _u64base, _f64, )(pg, bases);
280 }
281
test_svld1_gather_s32offset_s32(svbool_t pg,const int32_t * base,svint32_t offsets)282 svint32_t test_svld1_gather_s32offset_s32(svbool_t pg, const int32_t *base, svint32_t offsets) {
283 // CHECK-LABEL: test_svld1_gather_s32offset_s32
284 // CHECK: %[[PG:.*]] = call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg)
285 // 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)
286 // CHECK: ret <vscale x 4 x i32> %[[LOAD]]
287 return SVE_ACLE_FUNC(svld1_gather_, s32, offset, _s32)(pg, base, offsets);
288 }
289
test_svld1_gather_s64offset_s64(svbool_t pg,const int64_t * base,svint64_t offsets)290 svint64_t test_svld1_gather_s64offset_s64(svbool_t pg, const int64_t *base, svint64_t offsets) {
291 // CHECK-LABEL: test_svld1_gather_s64offset_s64
292 // CHECK: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg)
293 // 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)
294 // CHECK: ret <vscale x 2 x i64> %[[LOAD]]
295 return SVE_ACLE_FUNC(svld1_gather_, s64, offset, _s64)(pg, base, offsets);
296 }
297
test_svld1_gather_s32offset_u32(svbool_t pg,const uint32_t * base,svint32_t offsets)298 svuint32_t test_svld1_gather_s32offset_u32(svbool_t pg, const uint32_t *base, svint32_t offsets) {
299 // CHECK-LABEL: test_svld1_gather_s32offset_u32
300 // CHECK: %[[PG:.*]] = call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg)
301 // 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)
302 // CHECK: ret <vscale x 4 x i32> %[[LOAD]]
303 return SVE_ACLE_FUNC(svld1_gather_, s32, offset, _u32)(pg, base, offsets);
304 }
305
test_svld1_gather_s64offset_u64(svbool_t pg,const uint64_t * base,svint64_t offsets)306 svuint64_t test_svld1_gather_s64offset_u64(svbool_t pg, const uint64_t *base, svint64_t offsets) {
307 // CHECK-LABEL: test_svld1_gather_s64offset_u64
308 // CHECK: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg)
309 // 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)
310 // CHECK: ret <vscale x 2 x i64> %[[LOAD]]
311 return SVE_ACLE_FUNC(svld1_gather_, s64, offset, _u64)(pg, base, offsets);
312 }
313
test_svld1_gather_s32offset_f32(svbool_t pg,const float32_t * base,svint32_t offsets)314 svfloat32_t test_svld1_gather_s32offset_f32(svbool_t pg, const float32_t *base, svint32_t offsets) {
315 // CHECK-LABEL: test_svld1_gather_s32offset_f32
316 // CHECK: %[[PG:.*]] = call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg)
317 // 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)
318 // CHECK: ret <vscale x 4 x float> %[[LOAD]]
319 return SVE_ACLE_FUNC(svld1_gather_, s32, offset, _f32)(pg, base, offsets);
320 }
321
test_svld1_gather_s64offset_f64(svbool_t pg,const float64_t * base,svint64_t offsets)322 svfloat64_t test_svld1_gather_s64offset_f64(svbool_t pg, const float64_t *base, svint64_t offsets) {
323 // CHECK-LABEL: test_svld1_gather_s64offset_f64
324 // CHECK: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg)
325 // 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)
326 // CHECK: ret <vscale x 2 x double> %[[LOAD]]
327 return SVE_ACLE_FUNC(svld1_gather_, s64, offset, _f64)(pg, base, offsets);
328 }
329
test_svld1_gather_u32offset_s32(svbool_t pg,const int32_t * base,svuint32_t offsets)330 svint32_t test_svld1_gather_u32offset_s32(svbool_t pg, const int32_t *base, svuint32_t offsets) {
331 // CHECK-LABEL: test_svld1_gather_u32offset_s32
332 // CHECK: %[[PG:.*]] = call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg)
333 // 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)
334 // CHECK: ret <vscale x 4 x i32> %[[LOAD]]
335 return SVE_ACLE_FUNC(svld1_gather_, u32, offset, _s32)(pg, base, offsets);
336 }
337
test_svld1_gather_u64offset_s64(svbool_t pg,const int64_t * base,svuint64_t offsets)338 svint64_t test_svld1_gather_u64offset_s64(svbool_t pg, const int64_t *base, svuint64_t offsets) {
339 // CHECK-LABEL: test_svld1_gather_u64offset_s64
340 // CHECK: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg)
341 // 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)
342 // CHECK: ret <vscale x 2 x i64> %[[LOAD]]
343 return SVE_ACLE_FUNC(svld1_gather_, u64, offset, _s64)(pg, base, offsets);
344 }
345
test_svld1_gather_u32offset_u32(svbool_t pg,const uint32_t * base,svuint32_t offsets)346 svuint32_t test_svld1_gather_u32offset_u32(svbool_t pg, const uint32_t *base, svuint32_t offsets) {
347 // CHECK-LABEL: test_svld1_gather_u32offset_u32
348 // CHECK: %[[PG:.*]] = call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg)
349 // 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)
350 // CHECK: ret <vscale x 4 x i32> %[[LOAD]]
351 return SVE_ACLE_FUNC(svld1_gather_, u32, offset, _u32)(pg, base, offsets);
352 }
353
test_svld1_gather_u64offset_u64(svbool_t pg,const uint64_t * base,svuint64_t offsets)354 svuint64_t test_svld1_gather_u64offset_u64(svbool_t pg, const uint64_t *base, svuint64_t offsets) {
355 // CHECK-LABEL: test_svld1_gather_u64offset_u64
356 // CHECK: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg)
357 // 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)
358 // CHECK: ret <vscale x 2 x i64> %[[LOAD]]
359 return SVE_ACLE_FUNC(svld1_gather_, u64, offset, _u64)(pg, base, offsets);
360 }
361
test_svld1_gather_u32offset_f32(svbool_t pg,const float32_t * base,svuint32_t offsets)362 svfloat32_t test_svld1_gather_u32offset_f32(svbool_t pg, const float32_t *base, svuint32_t offsets) {
363 // CHECK-LABEL: test_svld1_gather_u32offset_f32
364 // CHECK: %[[PG:.*]] = call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg)
365 // 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)
366 // CHECK: ret <vscale x 4 x float> %[[LOAD]]
367 return SVE_ACLE_FUNC(svld1_gather_, u32, offset, _f32)(pg, base, offsets);
368 }
369
test_svld1_gather_u64offset_f64(svbool_t pg,const float64_t * base,svuint64_t offsets)370 svfloat64_t test_svld1_gather_u64offset_f64(svbool_t pg, const float64_t *base, svuint64_t offsets) {
371 // CHECK-LABEL: test_svld1_gather_u64offset_f64
372 // CHECK: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg)
373 // 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)
374 // CHECK: ret <vscale x 2 x double> %[[LOAD]]
375 return SVE_ACLE_FUNC(svld1_gather_, u64, offset, _f64)(pg, base, offsets);
376 }
377
test_svld1_gather_u32base_offset_s32(svbool_t pg,svuint32_t bases,int64_t offset)378 svint32_t test_svld1_gather_u32base_offset_s32(svbool_t pg, svuint32_t bases, int64_t offset) {
379 // CHECK-LABEL: test_svld1_gather_u32base_offset_s32
380 // CHECK: %[[PG:.*]] = call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg)
381 // 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)
382 // CHECK: ret <vscale x 4 x i32> %[[LOAD]]
383 return SVE_ACLE_FUNC(svld1_gather, _u32base, _offset_s32, )(pg, bases, offset);
384 }
385
test_svld1_gather_u64base_offset_s64(svbool_t pg,svuint64_t bases,int64_t offset)386 svint64_t test_svld1_gather_u64base_offset_s64(svbool_t pg, svuint64_t bases, int64_t offset) {
387 // CHECK-LABEL: test_svld1_gather_u64base_offset_s64
388 // CHECK: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg)
389 // 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)
390 // CHECK: ret <vscale x 2 x i64> %[[LOAD]]
391 return SVE_ACLE_FUNC(svld1_gather, _u64base, _offset_s64, )(pg, bases, offset);
392 }
393
test_svld1_gather_u32base_offset_u32(svbool_t pg,svuint32_t bases,int64_t offset)394 svuint32_t test_svld1_gather_u32base_offset_u32(svbool_t pg, svuint32_t bases, int64_t offset) {
395 // CHECK-LABEL: test_svld1_gather_u32base_offset_u32
396 // CHECK: %[[PG:.*]] = call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg)
397 // 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)
398 // CHECK: ret <vscale x 4 x i32> %[[LOAD]]
399 return SVE_ACLE_FUNC(svld1_gather, _u32base, _offset_u32, )(pg, bases, offset);
400 }
401
test_svld1_gather_u64base_offset_u64(svbool_t pg,svuint64_t bases,int64_t offset)402 svuint64_t test_svld1_gather_u64base_offset_u64(svbool_t pg, svuint64_t bases, int64_t offset) {
403 // CHECK-LABEL: test_svld1_gather_u64base_offset_u64
404 // CHECK: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg)
405 // 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)
406 // CHECK: ret <vscale x 2 x i64> %[[LOAD]]
407 return SVE_ACLE_FUNC(svld1_gather, _u64base, _offset_u64, )(pg, bases, offset);
408 }
409
test_svld1_gather_u32base_offset_f32(svbool_t pg,svuint32_t bases,int64_t offset)410 svfloat32_t test_svld1_gather_u32base_offset_f32(svbool_t pg, svuint32_t bases, int64_t offset) {
411 // CHECK-LABEL: test_svld1_gather_u32base_offset_f32
412 // CHECK: %[[PG:.*]] = call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg)
413 // 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)
414 // CHECK: ret <vscale x 4 x float> %[[LOAD]]
415 return SVE_ACLE_FUNC(svld1_gather, _u32base, _offset_f32, )(pg, bases, offset);
416 }
417
test_svld1_gather_u64base_offset_f64(svbool_t pg,svuint64_t bases,int64_t offset)418 svfloat64_t test_svld1_gather_u64base_offset_f64(svbool_t pg, svuint64_t bases, int64_t offset) {
419 // CHECK-LABEL: test_svld1_gather_u64base_offset_f64
420 // CHECK: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg)
421 // 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)
422 // CHECK: ret <vscale x 2 x double> %[[LOAD]]
423 return SVE_ACLE_FUNC(svld1_gather, _u64base, _offset_f64, )(pg, bases, offset);
424 }
425
test_svld1_gather_s32index_s32(svbool_t pg,const int32_t * base,svint32_t indices)426 svint32_t test_svld1_gather_s32index_s32(svbool_t pg, const int32_t *base, svint32_t indices) {
427 // CHECK-LABEL: test_svld1_gather_s32index_s32
428 // CHECK: %[[PG:.*]] = call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg)
429 // 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)
430 // CHECK: ret <vscale x 4 x i32> %[[LOAD]]
431 return SVE_ACLE_FUNC(svld1_gather_, s32, index, _s32)(pg, base, indices);
432 }
433
test_svld1_gather_s64index_s64(svbool_t pg,const int64_t * base,svint64_t indices)434 svint64_t test_svld1_gather_s64index_s64(svbool_t pg, const int64_t *base, svint64_t indices) {
435 // CHECK-LABEL: test_svld1_gather_s64index_s64
436 // CHECK: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg)
437 // 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)
438 // CHECK: ret <vscale x 2 x i64> %[[LOAD]]
439 return SVE_ACLE_FUNC(svld1_gather_, s64, index, _s64)(pg, base, indices);
440 }
441
test_svld1_gather_s32index_u32(svbool_t pg,const uint32_t * base,svint32_t indices)442 svuint32_t test_svld1_gather_s32index_u32(svbool_t pg, const uint32_t *base, svint32_t indices) {
443 // CHECK-LABEL: test_svld1_gather_s32index_u32
444 // CHECK: %[[PG:.*]] = call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg)
445 // 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)
446 // CHECK: ret <vscale x 4 x i32> %[[LOAD]]
447 return SVE_ACLE_FUNC(svld1_gather_, s32, index, _u32)(pg, base, indices);
448 }
449
test_svld1_gather_s64index_u64(svbool_t pg,const uint64_t * base,svint64_t indices)450 svuint64_t test_svld1_gather_s64index_u64(svbool_t pg, const uint64_t *base, svint64_t indices) {
451 // CHECK-LABEL: test_svld1_gather_s64index_u64
452 // CHECK: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg)
453 // 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)
454 // CHECK: ret <vscale x 2 x i64> %[[LOAD]]
455 return SVE_ACLE_FUNC(svld1_gather_, s64, index, _u64)(pg, base, indices);
456 }
457
test_svld1_gather_s32index_f32(svbool_t pg,const float32_t * base,svint32_t indices)458 svfloat32_t test_svld1_gather_s32index_f32(svbool_t pg, const float32_t *base, svint32_t indices) {
459 // CHECK-LABEL: test_svld1_gather_s32index_f32
460 // CHECK: %[[PG:.*]] = call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg)
461 // 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)
462 // CHECK: ret <vscale x 4 x float> %[[LOAD]]
463 return SVE_ACLE_FUNC(svld1_gather_, s32, index, _f32)(pg, base, indices);
464 }
465
test_svld1_gather_s64index_f64(svbool_t pg,const float64_t * base,svint64_t indices)466 svfloat64_t test_svld1_gather_s64index_f64(svbool_t pg, const float64_t *base, svint64_t indices) {
467 // CHECK-LABEL: test_svld1_gather_s64index_f64
468 // CHECK: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg)
469 // 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)
470 // CHECK: ret <vscale x 2 x double> %[[LOAD]]
471 return SVE_ACLE_FUNC(svld1_gather_, s64, index, _f64)(pg, base, indices);
472 }
473
test_svld1_gather_u32index_s32(svbool_t pg,const int32_t * base,svuint32_t indices)474 svint32_t test_svld1_gather_u32index_s32(svbool_t pg, const int32_t *base, svuint32_t indices) {
475 // CHECK-LABEL: test_svld1_gather_u32index_s32
476 // CHECK: %[[PG:.*]] = call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg)
477 // 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)
478 // CHECK: ret <vscale x 4 x i32> %[[LOAD]]
479 return SVE_ACLE_FUNC(svld1_gather_, u32, index, _s32)(pg, base, indices);
480 }
481
test_svld1_gather_u64index_s64(svbool_t pg,const int64_t * base,svuint64_t indices)482 svint64_t test_svld1_gather_u64index_s64(svbool_t pg, const int64_t *base, svuint64_t indices) {
483 // CHECK-LABEL: test_svld1_gather_u64index_s64
484 // CHECK: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg)
485 // 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)
486 // CHECK: ret <vscale x 2 x i64> %[[LOAD]]
487 return SVE_ACLE_FUNC(svld1_gather_, u64, index, _s64)(pg, base, indices);
488 }
489
test_svld1_gather_u32index_u32(svbool_t pg,const uint32_t * base,svuint32_t indices)490 svuint32_t test_svld1_gather_u32index_u32(svbool_t pg, const uint32_t *base, svuint32_t indices) {
491 // CHECK-LABEL: test_svld1_gather_u32index_u32
492 // CHECK: %[[PG:.*]] = call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg)
493 // 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)
494 // CHECK: ret <vscale x 4 x i32> %[[LOAD]]
495 return SVE_ACLE_FUNC(svld1_gather_, u32, index, _u32)(pg, base, indices);
496 }
497
test_svld1_gather_u64index_u64(svbool_t pg,const uint64_t * base,svuint64_t indices)498 svuint64_t test_svld1_gather_u64index_u64(svbool_t pg, const uint64_t *base, svuint64_t indices) {
499 // CHECK-LABEL: test_svld1_gather_u64index_u64
500 // CHECK: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg)
501 // 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)
502 // CHECK: ret <vscale x 2 x i64> %[[LOAD]]
503 return SVE_ACLE_FUNC(svld1_gather_, u64, index, _u64)(pg, base, indices);
504 }
505
test_svld1_gather_u32index_f32(svbool_t pg,const float32_t * base,svuint32_t indices)506 svfloat32_t test_svld1_gather_u32index_f32(svbool_t pg, const float32_t *base, svuint32_t indices) {
507 // CHECK-LABEL: test_svld1_gather_u32index_f32
508 // CHECK: %[[PG:.*]] = call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg)
509 // 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)
510 // CHECK: ret <vscale x 4 x float> %[[LOAD]]
511 return SVE_ACLE_FUNC(svld1_gather_, u32, index, _f32)(pg, base, indices);
512 }
513
test_svld1_gather_u64index_f64(svbool_t pg,const float64_t * base,svuint64_t indices)514 svfloat64_t test_svld1_gather_u64index_f64(svbool_t pg, const float64_t *base, svuint64_t indices) {
515 // CHECK-LABEL: test_svld1_gather_u64index_f64
516 // CHECK: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg)
517 // 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)
518 // CHECK: ret <vscale x 2 x double> %[[LOAD]]
519 return SVE_ACLE_FUNC(svld1_gather_, u64, index, _f64)(pg, base, indices);
520 }
521
test_svld1_gather_u32base_index_s32(svbool_t pg,svuint32_t bases,int64_t index)522 svint32_t test_svld1_gather_u32base_index_s32(svbool_t pg, svuint32_t bases, int64_t index) {
523 // CHECK-LABEL: test_svld1_gather_u32base_index_s32
524 // CHECK-DAG: %[[PG:.*]] = call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg)
525 // CHECK-DAG: %[[SHL:.*]] = shl i64 %index, 2
526 // 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]])
527 // CHECK: ret <vscale x 4 x i32> %[[LOAD]]
528 return SVE_ACLE_FUNC(svld1_gather, _u32base, _index_s32, )(pg, bases, index);
529 }
530
test_svld1_gather_u64base_index_s64(svbool_t pg,svuint64_t bases,int64_t index)531 svint64_t test_svld1_gather_u64base_index_s64(svbool_t pg, svuint64_t bases, int64_t index) {
532 // CHECK-LABEL: test_svld1_gather_u64base_index_s64
533 // CHECK-DAG: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg)
534 // CHECK-DAG: %[[SHL:.*]] = shl i64 %index, 3
535 // 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]])
536 // CHECK: ret <vscale x 2 x i64> %[[LOAD]]
537 return SVE_ACLE_FUNC(svld1_gather, _u64base, _index_s64, )(pg, bases, index);
538 }
539
test_svld1_gather_u32base_index_u32(svbool_t pg,svuint32_t bases,int64_t index)540 svuint32_t test_svld1_gather_u32base_index_u32(svbool_t pg, svuint32_t bases, int64_t index) {
541 // CHECK-LABEL: test_svld1_gather_u32base_index_u32
542 // CHECK-DAG: %[[PG:.*]] = call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg)
543 // CHECK-DAG: %[[SHL:.*]] = shl i64 %index, 2
544 // 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]])
545 // CHECK: ret <vscale x 4 x i32> %[[LOAD]]
546 return SVE_ACLE_FUNC(svld1_gather, _u32base, _index_u32, )(pg, bases, index);
547 }
548
test_svld1_gather_u64base_index_u64(svbool_t pg,svuint64_t bases,int64_t index)549 svuint64_t test_svld1_gather_u64base_index_u64(svbool_t pg, svuint64_t bases, int64_t index) {
550 // CHECK-LABEL: test_svld1_gather_u64base_index_u64
551 // CHECK-DAG: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg)
552 // CHECK-DAG: %[[SHL:.*]] = shl i64 %index, 3
553 // 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]])
554 // CHECK: ret <vscale x 2 x i64> %[[LOAD]]
555 return SVE_ACLE_FUNC(svld1_gather, _u64base, _index_u64, )(pg, bases, index);
556 }
557
test_svld1_gather_u32base_index_f32(svbool_t pg,svuint32_t bases,int64_t index)558 svfloat32_t test_svld1_gather_u32base_index_f32(svbool_t pg, svuint32_t bases, int64_t index) {
559 // CHECK-LABEL: test_svld1_gather_u32base_index_f32
560 // CHECK-DAG: %[[PG:.*]] = call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg)
561 // CHECK-DAG: %[[SHL:.*]] = shl i64 %index, 2
562 // 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]])
563 // CHECK: ret <vscale x 4 x float> %[[LOAD]]
564 return SVE_ACLE_FUNC(svld1_gather, _u32base, _index_f32, )(pg, bases, index);
565 }
566
test_svld1_gather_u64base_index_f64(svbool_t pg,svuint64_t bases,int64_t index)567 svfloat64_t test_svld1_gather_u64base_index_f64(svbool_t pg, svuint64_t bases, int64_t index) {
568 // CHECK-LABEL: test_svld1_gather_u64base_index_f64
569 // CHECK-DAG: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg)
570 // CHECK-DAG: %[[SHL:.*]] = shl i64 %index, 3
571 // 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]])
572 // CHECK: ret <vscale x 2 x double> %[[LOAD]]
573 return SVE_ACLE_FUNC(svld1_gather, _u64base, _index_f64, )(pg, bases, index);
574 }
575