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 -Wall -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 -Wall -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 -Wall -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 -Wall -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 -Wall -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_svst3_s8(svbool_t pg,int8_t * base,svint8x3_t data)16 void test_svst3_s8(svbool_t pg, int8_t *base, svint8x3_t data)
17 {
18 // CHECK-LABEL: test_svst3_s8
19 // CHECK-DAG: %[[V0:.*]] = call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv16i8.nxv48i8(<vscale x 48 x i8> %data, i32 0)
20 // CHECK-DAG: %[[V1:.*]] = call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv16i8.nxv48i8(<vscale x 48 x i8> %data, i32 1)
21 // CHECK-DAG: %[[V2:.*]] = call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv16i8.nxv48i8(<vscale x 48 x i8> %data, i32 2)
22 // CHECK: call void @llvm.aarch64.sve.st3.nxv16i8(<vscale x 16 x i8> %[[V0]], <vscale x 16 x i8> %[[V1]], <vscale x 16 x i8> %[[V2]], <vscale x 16 x i1> %pg, i8* %base)
23 // CHECK-NEXT: ret
24 return SVE_ACLE_FUNC(svst3,_s8,,)(pg, base, data);
25 }
26
test_svst3_s16(svbool_t pg,int16_t * base,svint16x3_t data)27 void test_svst3_s16(svbool_t pg, int16_t *base, svint16x3_t data)
28 {
29 // CHECK-LABEL: test_svst3_s16
30 // CHECK-DAG: %[[V0:.*]] = call <vscale x 8 x i16> @llvm.aarch64.sve.tuple.get.nxv8i16.nxv24i16(<vscale x 24 x i16> %data, i32 0)
31 // CHECK-DAG: %[[V1:.*]] = call <vscale x 8 x i16> @llvm.aarch64.sve.tuple.get.nxv8i16.nxv24i16(<vscale x 24 x i16> %data, i32 1)
32 // CHECK-DAG: %[[V2:.*]] = call <vscale x 8 x i16> @llvm.aarch64.sve.tuple.get.nxv8i16.nxv24i16(<vscale x 24 x i16> %data, i32 2)
33 // CHECK-DAG: %[[PG:.*]] = call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> %pg)
34 // CHECK: call void @llvm.aarch64.sve.st3.nxv8i16(<vscale x 8 x i16> %[[V0]], <vscale x 8 x i16> %[[V1]], <vscale x 8 x i16> %[[V2]], <vscale x 8 x i1> %[[PG]], i16* %base)
35 // CHECK-NEXT: ret
36 return SVE_ACLE_FUNC(svst3,_s16,,)(pg, base, data);
37 }
38
test_svst3_s32(svbool_t pg,int32_t * base,svint32x3_t data)39 void test_svst3_s32(svbool_t pg, int32_t *base, svint32x3_t data)
40 {
41 // CHECK-LABEL: test_svst3_s32
42 // CHECK-DAG: %[[V0:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.tuple.get.nxv4i32.nxv12i32(<vscale x 12 x i32> %data, i32 0)
43 // CHECK-DAG: %[[V1:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.tuple.get.nxv4i32.nxv12i32(<vscale x 12 x i32> %data, i32 1)
44 // CHECK-DAG: %[[V2:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.tuple.get.nxv4i32.nxv12i32(<vscale x 12 x i32> %data, i32 2)
45 // CHECK-DAG: %[[PG:.*]] = call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg)
46 // CHECK: call void @llvm.aarch64.sve.st3.nxv4i32(<vscale x 4 x i32> %[[V0]], <vscale x 4 x i32> %[[V1]], <vscale x 4 x i32> %[[V2]], <vscale x 4 x i1> %[[PG]], i32* %base)
47 // CHECK-NEXT: ret
48 return SVE_ACLE_FUNC(svst3,_s32,,)(pg, base, data);
49 }
50
test_svst3_s64(svbool_t pg,int64_t * base,svint64x3_t data)51 void test_svst3_s64(svbool_t pg, int64_t *base, svint64x3_t data)
52 {
53 // CHECK-LABEL: test_svst3_s64
54 // CHECK-DAG: %[[V0:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv2i64.nxv6i64(<vscale x 6 x i64> %data, i32 0)
55 // CHECK-DAG: %[[V1:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv2i64.nxv6i64(<vscale x 6 x i64> %data, i32 1)
56 // CHECK-DAG: %[[V2:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv2i64.nxv6i64(<vscale x 6 x i64> %data, i32 2)
57 // CHECK-DAG: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg)
58 // CHECK: call void @llvm.aarch64.sve.st3.nxv2i64(<vscale x 2 x i64> %[[V0]], <vscale x 2 x i64> %[[V1]], <vscale x 2 x i64> %[[V2]], <vscale x 2 x i1> %[[PG]], i64* %base)
59 // CHECK-NEXT: ret
60 return SVE_ACLE_FUNC(svst3,_s64,,)(pg, base, data);
61 }
62
test_svst3_u8(svbool_t pg,uint8_t * base,svuint8x3_t data)63 void test_svst3_u8(svbool_t pg, uint8_t *base, svuint8x3_t data)
64 {
65 // CHECK-LABEL: test_svst3_u8
66 // CHECK-DAG: %[[V0:.*]] = call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv16i8.nxv48i8(<vscale x 48 x i8> %data, i32 0)
67 // CHECK-DAG: %[[V1:.*]] = call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv16i8.nxv48i8(<vscale x 48 x i8> %data, i32 1)
68 // CHECK-DAG: %[[V2:.*]] = call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv16i8.nxv48i8(<vscale x 48 x i8> %data, i32 2)
69 // CHECK: call void @llvm.aarch64.sve.st3.nxv16i8(<vscale x 16 x i8> %[[V0]], <vscale x 16 x i8> %[[V1]], <vscale x 16 x i8> %[[V2]], <vscale x 16 x i1> %pg, i8* %base)
70 // CHECK-NEXT: ret
71 return SVE_ACLE_FUNC(svst3,_u8,,)(pg, base, data);
72 }
73
test_svst3_u16(svbool_t pg,uint16_t * base,svuint16x3_t data)74 void test_svst3_u16(svbool_t pg, uint16_t *base, svuint16x3_t data)
75 {
76 // CHECK-LABEL: test_svst3_u16
77 // CHECK-DAG: %[[V0:.*]] = call <vscale x 8 x i16> @llvm.aarch64.sve.tuple.get.nxv8i16.nxv24i16(<vscale x 24 x i16> %data, i32 0)
78 // CHECK-DAG: %[[V1:.*]] = call <vscale x 8 x i16> @llvm.aarch64.sve.tuple.get.nxv8i16.nxv24i16(<vscale x 24 x i16> %data, i32 1)
79 // CHECK-DAG: %[[V2:.*]] = call <vscale x 8 x i16> @llvm.aarch64.sve.tuple.get.nxv8i16.nxv24i16(<vscale x 24 x i16> %data, i32 2)
80 // CHECK-DAG: %[[PG:.*]] = call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> %pg)
81 // CHECK: call void @llvm.aarch64.sve.st3.nxv8i16(<vscale x 8 x i16> %[[V0]], <vscale x 8 x i16> %[[V1]], <vscale x 8 x i16> %[[V2]], <vscale x 8 x i1> %[[PG]], i16* %base)
82 // CHECK-NEXT: ret
83 return SVE_ACLE_FUNC(svst3,_u16,,)(pg, base, data);
84 }
85
test_svst3_u32(svbool_t pg,uint32_t * base,svuint32x3_t data)86 void test_svst3_u32(svbool_t pg, uint32_t *base, svuint32x3_t data)
87 {
88 // CHECK-LABEL: test_svst3_u32
89 // CHECK-DAG: %[[V0:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.tuple.get.nxv4i32.nxv12i32(<vscale x 12 x i32> %data, i32 0)
90 // CHECK-DAG: %[[V1:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.tuple.get.nxv4i32.nxv12i32(<vscale x 12 x i32> %data, i32 1)
91 // CHECK-DAG: %[[V2:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.tuple.get.nxv4i32.nxv12i32(<vscale x 12 x i32> %data, i32 2)
92 // CHECK-DAG: %[[PG:.*]] = call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg)
93 // CHECK: call void @llvm.aarch64.sve.st3.nxv4i32(<vscale x 4 x i32> %[[V0]], <vscale x 4 x i32> %[[V1]], <vscale x 4 x i32> %[[V2]], <vscale x 4 x i1> %[[PG]], i32* %base)
94 // CHECK-NEXT: ret
95 return SVE_ACLE_FUNC(svst3,_u32,,)(pg, base, data);
96 }
97
test_svst3_u64(svbool_t pg,uint64_t * base,svuint64x3_t data)98 void test_svst3_u64(svbool_t pg, uint64_t *base, svuint64x3_t data)
99 {
100 // CHECK-LABEL: test_svst3_u64
101 // CHECK-DAG: %[[V0:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv2i64.nxv6i64(<vscale x 6 x i64> %data, i32 0)
102 // CHECK-DAG: %[[V1:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv2i64.nxv6i64(<vscale x 6 x i64> %data, i32 1)
103 // CHECK-DAG: %[[V2:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv2i64.nxv6i64(<vscale x 6 x i64> %data, i32 2)
104 // CHECK-DAG: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg)
105 // CHECK: call void @llvm.aarch64.sve.st3.nxv2i64(<vscale x 2 x i64> %[[V0]], <vscale x 2 x i64> %[[V1]], <vscale x 2 x i64> %[[V2]], <vscale x 2 x i1> %[[PG]], i64* %base)
106 // CHECK-NEXT: ret
107 return SVE_ACLE_FUNC(svst3,_u64,,)(pg, base, data);
108 }
109
test_svst3_f16(svbool_t pg,float16_t * base,svfloat16x3_t data)110 void test_svst3_f16(svbool_t pg, float16_t *base, svfloat16x3_t data)
111 {
112 // CHECK-LABEL: test_svst3_f16
113 // CHECK-DAG: %[[V0:.*]] = call <vscale x 8 x half> @llvm.aarch64.sve.tuple.get.nxv8f16.nxv24f16(<vscale x 24 x half> %data, i32 0)
114 // CHECK-DAG: %[[V1:.*]] = call <vscale x 8 x half> @llvm.aarch64.sve.tuple.get.nxv8f16.nxv24f16(<vscale x 24 x half> %data, i32 1)
115 // CHECK-DAG: %[[V2:.*]] = call <vscale x 8 x half> @llvm.aarch64.sve.tuple.get.nxv8f16.nxv24f16(<vscale x 24 x half> %data, i32 2)
116 // CHECK-DAG: %[[PG:.*]] = call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> %pg)
117 // CHECK: call void @llvm.aarch64.sve.st3.nxv8f16(<vscale x 8 x half> %[[V0]], <vscale x 8 x half> %[[V1]], <vscale x 8 x half> %[[V2]], <vscale x 8 x i1> %[[PG]], half* %base)
118 // CHECK-NEXT: ret
119 return SVE_ACLE_FUNC(svst3,_f16,,)(pg, base, data);
120 }
121
test_svst3_f32(svbool_t pg,float32_t * base,svfloat32x3_t data)122 void test_svst3_f32(svbool_t pg, float32_t *base, svfloat32x3_t data)
123 {
124 // CHECK-LABEL: test_svst3_f32
125 // CHECK-DAG: %[[V0:.*]] = call <vscale x 4 x float> @llvm.aarch64.sve.tuple.get.nxv4f32.nxv12f32(<vscale x 12 x float> %data, i32 0)
126 // CHECK-DAG: %[[V1:.*]] = call <vscale x 4 x float> @llvm.aarch64.sve.tuple.get.nxv4f32.nxv12f32(<vscale x 12 x float> %data, i32 1)
127 // CHECK-DAG: %[[V2:.*]] = call <vscale x 4 x float> @llvm.aarch64.sve.tuple.get.nxv4f32.nxv12f32(<vscale x 12 x float> %data, i32 2)
128 // CHECK-DAG: %[[PG:.*]] = call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg)
129 // CHECK: call void @llvm.aarch64.sve.st3.nxv4f32(<vscale x 4 x float> %[[V0]], <vscale x 4 x float> %[[V1]], <vscale x 4 x float> %[[V2]], <vscale x 4 x i1> %[[PG]], float* %base)
130 // CHECK-NEXT: ret
131 return SVE_ACLE_FUNC(svst3,_f32,,)(pg, base, data);
132 }
133
test_svst3_f64(svbool_t pg,float64_t * base,svfloat64x3_t data)134 void test_svst3_f64(svbool_t pg, float64_t *base, svfloat64x3_t data)
135 {
136 // CHECK-LABEL: test_svst3_f64
137 // CHECK-DAG: %[[V0:.*]] = call <vscale x 2 x double> @llvm.aarch64.sve.tuple.get.nxv2f64.nxv6f64(<vscale x 6 x double> %data, i32 0)
138 // CHECK-DAG: %[[V1:.*]] = call <vscale x 2 x double> @llvm.aarch64.sve.tuple.get.nxv2f64.nxv6f64(<vscale x 6 x double> %data, i32 1)
139 // CHECK-DAG: %[[V2:.*]] = call <vscale x 2 x double> @llvm.aarch64.sve.tuple.get.nxv2f64.nxv6f64(<vscale x 6 x double> %data, i32 2)
140 // CHECK-DAG: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg)
141 // CHECK: call void @llvm.aarch64.sve.st3.nxv2f64(<vscale x 2 x double> %[[V0]], <vscale x 2 x double> %[[V1]], <vscale x 2 x double> %[[V2]], <vscale x 2 x i1> %[[PG]], double* %base)
142 // CHECK-NEXT: ret
143 return SVE_ACLE_FUNC(svst3,_f64,,)(pg, base, data);
144 }
145
test_svst3_vnum_s8(svbool_t pg,int8_t * base,int64_t vnum,svint8x3_t data)146 void test_svst3_vnum_s8(svbool_t pg, int8_t *base, int64_t vnum, svint8x3_t data)
147 {
148 // CHECK-LABEL: test_svst3_vnum_s8
149 // CHECK-DAG: %[[BITCAST:.*]] = bitcast i8* %base to <vscale x 16 x i8>*
150 // CHECK-DAG: %[[GEP:.*]] = getelementptr <vscale x 16 x i8>, <vscale x 16 x i8>* %[[BITCAST]], i64 %vnum, i64 0
151 // CHECK-DAG: %[[V0:.*]] = call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv16i8.nxv48i8(<vscale x 48 x i8> %data, i32 0)
152 // CHECK-DAG: %[[V1:.*]] = call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv16i8.nxv48i8(<vscale x 48 x i8> %data, i32 1)
153 // CHECK-DAG: %[[V2:.*]] = call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv16i8.nxv48i8(<vscale x 48 x i8> %data, i32 2)
154 // CHECK: call void @llvm.aarch64.sve.st3.nxv16i8(<vscale x 16 x i8> %[[V0]], <vscale x 16 x i8> %[[V1]], <vscale x 16 x i8> %[[V2]], <vscale x 16 x i1> %pg, i8* %[[GEP]])
155 // CHECK-NEXT: ret
156 return SVE_ACLE_FUNC(svst3_vnum,_s8,,)(pg, base, vnum, data);
157 }
158
test_svst3_vnum_s16(svbool_t pg,int16_t * base,int64_t vnum,svint16x3_t data)159 void test_svst3_vnum_s16(svbool_t pg, int16_t *base, int64_t vnum, svint16x3_t data)
160 {
161 // CHECK-LABEL: test_svst3_vnum_s16
162 // CHECK-DAG: %[[BITCAST:.*]] = bitcast i16* %base to <vscale x 8 x i16>*
163 // CHECK-DAG: %[[GEP:.*]] = getelementptr <vscale x 8 x i16>, <vscale x 8 x i16>* %[[BITCAST]], i64 %vnum, i64 0
164 // CHECK-DAG: %[[V0:.*]] = call <vscale x 8 x i16> @llvm.aarch64.sve.tuple.get.nxv8i16.nxv24i16(<vscale x 24 x i16> %data, i32 0)
165 // CHECK-DAG: %[[V1:.*]] = call <vscale x 8 x i16> @llvm.aarch64.sve.tuple.get.nxv8i16.nxv24i16(<vscale x 24 x i16> %data, i32 1)
166 // CHECK-DAG: %[[V2:.*]] = call <vscale x 8 x i16> @llvm.aarch64.sve.tuple.get.nxv8i16.nxv24i16(<vscale x 24 x i16> %data, i32 2)
167 // CHECK-DAG: %[[PG:.*]] = call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> %pg)
168 // CHECK: call void @llvm.aarch64.sve.st3.nxv8i16(<vscale x 8 x i16> %[[V0]], <vscale x 8 x i16> %[[V1]], <vscale x 8 x i16> %[[V2]], <vscale x 8 x i1> %[[PG]], i16* %[[GEP]])
169 // CHECK-NEXT: ret
170 return SVE_ACLE_FUNC(svst3_vnum,_s16,,)(pg, base, vnum, data);
171 }
172
test_svst3_vnum_s32(svbool_t pg,int32_t * base,int64_t vnum,svint32x3_t data)173 void test_svst3_vnum_s32(svbool_t pg, int32_t *base, int64_t vnum, svint32x3_t data)
174 {
175 // CHECK-LABEL: test_svst3_vnum_s32
176 // CHECK-DAG: %[[BITCAST:.*]] = bitcast i32* %base to <vscale x 4 x i32>*
177 // CHECK-DAG: %[[GEP:.*]] = getelementptr <vscale x 4 x i32>, <vscale x 4 x i32>* %[[BITCAST]], i64 %vnum, i64 0
178 // CHECK-DAG: %[[V0:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.tuple.get.nxv4i32.nxv12i32(<vscale x 12 x i32> %data, i32 0)
179 // CHECK-DAG: %[[V1:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.tuple.get.nxv4i32.nxv12i32(<vscale x 12 x i32> %data, i32 1)
180 // CHECK-DAG: %[[V2:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.tuple.get.nxv4i32.nxv12i32(<vscale x 12 x i32> %data, i32 2)
181 // CHECK-DAG: %[[PG:.*]] = call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg)
182 // CHECK: call void @llvm.aarch64.sve.st3.nxv4i32(<vscale x 4 x i32> %[[V0]], <vscale x 4 x i32> %[[V1]], <vscale x 4 x i32> %[[V2]], <vscale x 4 x i1> %[[PG]], i32* %[[GEP]])
183 // CHECK-NEXT: ret
184 return SVE_ACLE_FUNC(svst3_vnum,_s32,,)(pg, base, vnum, data);
185 }
186
test_svst3_vnum_s64(svbool_t pg,int64_t * base,int64_t vnum,svint64x3_t data)187 void test_svst3_vnum_s64(svbool_t pg, int64_t *base, int64_t vnum, svint64x3_t data)
188 {
189 // CHECK-LABEL: test_svst3_vnum_s64
190 // CHECK-DAG: %[[BITCAST:.*]] = bitcast i64* %base to <vscale x 2 x i64>*
191 // CHECK-DAG: %[[GEP:.*]] = getelementptr <vscale x 2 x i64>, <vscale x 2 x i64>* %[[BITCAST]], i64 %vnum, i64 0
192 // CHECK-DAG: %[[V0:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv2i64.nxv6i64(<vscale x 6 x i64> %data, i32 0)
193 // CHECK-DAG: %[[V1:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv2i64.nxv6i64(<vscale x 6 x i64> %data, i32 1)
194 // CHECK-DAG: %[[V2:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv2i64.nxv6i64(<vscale x 6 x i64> %data, i32 2)
195 // CHECK-DAG: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg)
196 // CHECK: call void @llvm.aarch64.sve.st3.nxv2i64(<vscale x 2 x i64> %[[V0]], <vscale x 2 x i64> %[[V1]], <vscale x 2 x i64> %[[V2]], <vscale x 2 x i1> %[[PG]], i64* %[[GEP]])
197 // CHECK-NEXT: ret
198 return SVE_ACLE_FUNC(svst3_vnum,_s64,,)(pg, base, vnum, data);
199 }
200
test_svst3_vnum_u8(svbool_t pg,uint8_t * base,int64_t vnum,svuint8x3_t data)201 void test_svst3_vnum_u8(svbool_t pg, uint8_t *base, int64_t vnum, svuint8x3_t data)
202 {
203 // CHECK-LABEL: test_svst3_vnum_u8
204 // CHECK-DAG: %[[BITCAST:.*]] = bitcast i8* %base to <vscale x 16 x i8>*
205 // CHECK-DAG: %[[GEP:.*]] = getelementptr <vscale x 16 x i8>, <vscale x 16 x i8>* %[[BITCAST]], i64 %vnum, i64 0
206 // CHECK-DAG: %[[V0:.*]] = call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv16i8.nxv48i8(<vscale x 48 x i8> %data, i32 0)
207 // CHECK-DAG: %[[V1:.*]] = call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv16i8.nxv48i8(<vscale x 48 x i8> %data, i32 1)
208 // CHECK-DAG: %[[V2:.*]] = call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv16i8.nxv48i8(<vscale x 48 x i8> %data, i32 2)
209 // CHECK: call void @llvm.aarch64.sve.st3.nxv16i8(<vscale x 16 x i8> %[[V0]], <vscale x 16 x i8> %[[V1]], <vscale x 16 x i8> %[[V2]], <vscale x 16 x i1> %pg, i8* %[[GEP]])
210 // CHECK-NEXT: ret
211 return SVE_ACLE_FUNC(svst3_vnum,_u8,,)(pg, base, vnum, data);
212 }
213
test_svst3_vnum_u16(svbool_t pg,uint16_t * base,int64_t vnum,svuint16x3_t data)214 void test_svst3_vnum_u16(svbool_t pg, uint16_t *base, int64_t vnum, svuint16x3_t data)
215 {
216 // CHECK-LABEL: test_svst3_vnum_u16
217 // CHECK-DAG: %[[BITCAST:.*]] = bitcast i16* %base to <vscale x 8 x i16>*
218 // CHECK-DAG: %[[GEP:.*]] = getelementptr <vscale x 8 x i16>, <vscale x 8 x i16>* %[[BITCAST]], i64 %vnum, i64 0
219 // CHECK-DAG: %[[V0:.*]] = call <vscale x 8 x i16> @llvm.aarch64.sve.tuple.get.nxv8i16.nxv24i16(<vscale x 24 x i16> %data, i32 0)
220 // CHECK-DAG: %[[V1:.*]] = call <vscale x 8 x i16> @llvm.aarch64.sve.tuple.get.nxv8i16.nxv24i16(<vscale x 24 x i16> %data, i32 1)
221 // CHECK-DAG: %[[V2:.*]] = call <vscale x 8 x i16> @llvm.aarch64.sve.tuple.get.nxv8i16.nxv24i16(<vscale x 24 x i16> %data, i32 2)
222 // CHECK-DAG: %[[PG:.*]] = call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> %pg)
223 // CHECK: call void @llvm.aarch64.sve.st3.nxv8i16(<vscale x 8 x i16> %[[V0]], <vscale x 8 x i16> %[[V1]], <vscale x 8 x i16> %[[V2]], <vscale x 8 x i1> %[[PG]], i16* %[[GEP]])
224 // CHECK-NEXT: ret
225 return SVE_ACLE_FUNC(svst3_vnum,_u16,,)(pg, base, vnum, data);
226 }
227
test_svst3_vnum_u32(svbool_t pg,uint32_t * base,int64_t vnum,svuint32x3_t data)228 void test_svst3_vnum_u32(svbool_t pg, uint32_t *base, int64_t vnum, svuint32x3_t data)
229 {
230 // CHECK-LABEL: test_svst3_vnum_u32
231 // CHECK-DAG: %[[BITCAST:.*]] = bitcast i32* %base to <vscale x 4 x i32>*
232 // CHECK-DAG: %[[GEP:.*]] = getelementptr <vscale x 4 x i32>, <vscale x 4 x i32>* %[[BITCAST]], i64 %vnum, i64 0
233 // CHECK-DAG: %[[V0:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.tuple.get.nxv4i32.nxv12i32(<vscale x 12 x i32> %data, i32 0)
234 // CHECK-DAG: %[[V1:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.tuple.get.nxv4i32.nxv12i32(<vscale x 12 x i32> %data, i32 1)
235 // CHECK-DAG: %[[V2:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.tuple.get.nxv4i32.nxv12i32(<vscale x 12 x i32> %data, i32 2)
236 // CHECK-DAG: %[[PG:.*]] = call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg)
237 // CHECK: call void @llvm.aarch64.sve.st3.nxv4i32(<vscale x 4 x i32> %[[V0]], <vscale x 4 x i32> %[[V1]], <vscale x 4 x i32> %[[V2]], <vscale x 4 x i1> %[[PG]], i32* %[[GEP]])
238 // CHECK-NEXT: ret
239 return SVE_ACLE_FUNC(svst3_vnum,_u32,,)(pg, base, vnum, data);
240 }
241
test_svst3_vnum_u64(svbool_t pg,uint64_t * base,int64_t vnum,svuint64x3_t data)242 void test_svst3_vnum_u64(svbool_t pg, uint64_t *base, int64_t vnum, svuint64x3_t data)
243 {
244 // CHECK-LABEL: test_svst3_vnum_u64
245 // CHECK-DAG: %[[BITCAST:.*]] = bitcast i64* %base to <vscale x 2 x i64>*
246 // CHECK-DAG: %[[GEP:.*]] = getelementptr <vscale x 2 x i64>, <vscale x 2 x i64>* %[[BITCAST]], i64 %vnum, i64 0
247 // CHECK-DAG: %[[V0:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv2i64.nxv6i64(<vscale x 6 x i64> %data, i32 0)
248 // CHECK-DAG: %[[V1:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv2i64.nxv6i64(<vscale x 6 x i64> %data, i32 1)
249 // CHECK-DAG: %[[V2:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv2i64.nxv6i64(<vscale x 6 x i64> %data, i32 2)
250 // CHECK-DAG: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg)
251 // CHECK: call void @llvm.aarch64.sve.st3.nxv2i64(<vscale x 2 x i64> %[[V0]], <vscale x 2 x i64> %[[V1]], <vscale x 2 x i64> %[[V2]], <vscale x 2 x i1> %[[PG]], i64* %[[GEP]])
252 // CHECK-NEXT: ret
253 return SVE_ACLE_FUNC(svst3_vnum,_u64,,)(pg, base, vnum, data);
254 }
255
test_svst3_vnum_f16(svbool_t pg,float16_t * base,int64_t vnum,svfloat16x3_t data)256 void test_svst3_vnum_f16(svbool_t pg, float16_t *base, int64_t vnum, svfloat16x3_t data)
257 {
258 // CHECK-LABEL: test_svst3_vnum_f16
259 // CHECK-DAG: %[[BITCAST:.*]] = bitcast half* %base to <vscale x 8 x half>*
260 // CHECK-DAG: %[[GEP:.*]] = getelementptr <vscale x 8 x half>, <vscale x 8 x half>* %[[BITCAST]], i64 %vnum, i64 0
261 // CHECK-DAG: %[[V0:.*]] = call <vscale x 8 x half> @llvm.aarch64.sve.tuple.get.nxv8f16.nxv24f16(<vscale x 24 x half> %data, i32 0)
262 // CHECK-DAG: %[[V1:.*]] = call <vscale x 8 x half> @llvm.aarch64.sve.tuple.get.nxv8f16.nxv24f16(<vscale x 24 x half> %data, i32 1)
263 // CHECK-DAG: %[[V2:.*]] = call <vscale x 8 x half> @llvm.aarch64.sve.tuple.get.nxv8f16.nxv24f16(<vscale x 24 x half> %data, i32 2)
264 // CHECK-DAG: %[[PG:.*]] = call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> %pg)
265 // CHECK: call void @llvm.aarch64.sve.st3.nxv8f16(<vscale x 8 x half> %[[V0]], <vscale x 8 x half> %[[V1]], <vscale x 8 x half> %[[V2]], <vscale x 8 x i1> %[[PG]], half* %[[GEP]])
266 // CHECK-NEXT: ret
267 return SVE_ACLE_FUNC(svst3_vnum,_f16,,)(pg, base, vnum, data);
268 }
269
test_svst3_vnum_f32(svbool_t pg,float32_t * base,int64_t vnum,svfloat32x3_t data)270 void test_svst3_vnum_f32(svbool_t pg, float32_t *base, int64_t vnum, svfloat32x3_t data)
271 {
272 // CHECK-LABEL: test_svst3_vnum_f32
273 // CHECK-DAG: %[[BITCAST:.*]] = bitcast float* %base to <vscale x 4 x float>*
274 // CHECK-DAG: %[[GEP:.*]] = getelementptr <vscale x 4 x float>, <vscale x 4 x float>* %[[BITCAST]], i64 %vnum, i64 0
275 // CHECK-DAG: %[[V0:.*]] = call <vscale x 4 x float> @llvm.aarch64.sve.tuple.get.nxv4f32.nxv12f32(<vscale x 12 x float> %data, i32 0)
276 // CHECK-DAG: %[[V1:.*]] = call <vscale x 4 x float> @llvm.aarch64.sve.tuple.get.nxv4f32.nxv12f32(<vscale x 12 x float> %data, i32 1)
277 // CHECK-DAG: %[[V2:.*]] = call <vscale x 4 x float> @llvm.aarch64.sve.tuple.get.nxv4f32.nxv12f32(<vscale x 12 x float> %data, i32 2)
278 // CHECK-DAG: %[[PG:.*]] = call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg)
279 // CHECK: call void @llvm.aarch64.sve.st3.nxv4f32(<vscale x 4 x float> %[[V0]], <vscale x 4 x float> %[[V1]], <vscale x 4 x float> %[[V2]], <vscale x 4 x i1> %[[PG]], float* %[[GEP]])
280 // CHECK-NEXT: ret
281 return SVE_ACLE_FUNC(svst3_vnum,_f32,,)(pg, base, vnum, data);
282 }
283
test_svst3_vnum_f64(svbool_t pg,float64_t * base,int64_t vnum,svfloat64x3_t data)284 void test_svst3_vnum_f64(svbool_t pg, float64_t *base, int64_t vnum, svfloat64x3_t data)
285 {
286 // CHECK-LABEL: test_svst3_vnum_f64
287 // CHECK-DAG: %[[BITCAST:.*]] = bitcast double* %base to <vscale x 2 x double>*
288 // CHECK-DAG: %[[GEP:.*]] = getelementptr <vscale x 2 x double>, <vscale x 2 x double>* %[[BITCAST]], i64 %vnum, i64 0
289 // CHECK-DAG: %[[V0:.*]] = call <vscale x 2 x double> @llvm.aarch64.sve.tuple.get.nxv2f64.nxv6f64(<vscale x 6 x double> %data, i32 0)
290 // CHECK-DAG: %[[V1:.*]] = call <vscale x 2 x double> @llvm.aarch64.sve.tuple.get.nxv2f64.nxv6f64(<vscale x 6 x double> %data, i32 1)
291 // CHECK-DAG: %[[V2:.*]] = call <vscale x 2 x double> @llvm.aarch64.sve.tuple.get.nxv2f64.nxv6f64(<vscale x 6 x double> %data, i32 2)
292 // CHECK-DAG: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg)
293 // CHECK: call void @llvm.aarch64.sve.st3.nxv2f64(<vscale x 2 x double> %[[V0]], <vscale x 2 x double> %[[V1]], <vscale x 2 x double> %[[V2]], <vscale x 2 x i1> %[[PG]], double* %[[GEP]])
294 // CHECK-NEXT: ret
295 return SVE_ACLE_FUNC(svst3_vnum,_f64,,)(pg, base, vnum, data);
296 }
297