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_svst4_s8(svbool_t pg,int8_t * base,svint8x4_t data)16 void test_svst4_s8(svbool_t pg, int8_t *base, svint8x4_t data)
17 {
18 // CHECK-LABEL: test_svst4_s8
19 // CHECK-DAG: %[[V0:.*]] = call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv16i8.nxv64i8(<vscale x 64 x i8> %data, i32 0)
20 // CHECK-DAG: %[[V1:.*]] = call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv16i8.nxv64i8(<vscale x 64 x i8> %data, i32 1)
21 // CHECK-DAG: %[[V2:.*]] = call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv16i8.nxv64i8(<vscale x 64 x i8> %data, i32 2)
22 // CHECK-DAG: %[[V3:.*]] = call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv16i8.nxv64i8(<vscale x 64 x i8> %data, i32 3)
23 // CHECK: call void @llvm.aarch64.sve.st4.nxv16i8(<vscale x 16 x i8> %[[V0]], <vscale x 16 x i8> %[[V1]], <vscale x 16 x i8> %[[V2]], <vscale x 16 x i8> %[[V3]], <vscale x 16 x i1> %pg, i8* %base)
24 // CHECK-NEXT: ret
25 return SVE_ACLE_FUNC(svst4,_s8,,)(pg, base, data);
26 }
27
test_svst4_s16(svbool_t pg,int16_t * base,svint16x4_t data)28 void test_svst4_s16(svbool_t pg, int16_t *base, svint16x4_t data)
29 {
30 // CHECK-LABEL: test_svst4_s16
31 // CHECK-DAG: %[[V0:.*]] = call <vscale x 8 x i16> @llvm.aarch64.sve.tuple.get.nxv8i16.nxv32i16(<vscale x 32 x i16> %data, i32 0)
32 // CHECK-DAG: %[[V1:.*]] = call <vscale x 8 x i16> @llvm.aarch64.sve.tuple.get.nxv8i16.nxv32i16(<vscale x 32 x i16> %data, i32 1)
33 // CHECK-DAG: %[[V2:.*]] = call <vscale x 8 x i16> @llvm.aarch64.sve.tuple.get.nxv8i16.nxv32i16(<vscale x 32 x i16> %data, i32 2)
34 // CHECK-DAG: %[[V3:.*]] = call <vscale x 8 x i16> @llvm.aarch64.sve.tuple.get.nxv8i16.nxv32i16(<vscale x 32 x i16> %data, i32 3)
35 // CHECK-DAG: %[[PG:.*]] = call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> %pg)
36 // CHECK: call void @llvm.aarch64.sve.st4.nxv8i16(<vscale x 8 x i16> %[[V0]], <vscale x 8 x i16> %[[V1]], <vscale x 8 x i16> %[[V2]], <vscale x 8 x i16> %[[V3]], <vscale x 8 x i1> %[[PG]], i16* %base)
37 // CHECK-NEXT: ret
38 return SVE_ACLE_FUNC(svst4,_s16,,)(pg, base, data);
39 }
40
test_svst4_s32(svbool_t pg,int32_t * base,svint32x4_t data)41 void test_svst4_s32(svbool_t pg, int32_t *base, svint32x4_t data)
42 {
43 // CHECK-LABEL: test_svst4_s32
44 // CHECK-DAG: %[[V0:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.tuple.get.nxv4i32.nxv16i32(<vscale x 16 x i32> %data, i32 0)
45 // CHECK-DAG: %[[V1:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.tuple.get.nxv4i32.nxv16i32(<vscale x 16 x i32> %data, i32 1)
46 // CHECK-DAG: %[[V2:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.tuple.get.nxv4i32.nxv16i32(<vscale x 16 x i32> %data, i32 2)
47 // CHECK-DAG: %[[V3:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.tuple.get.nxv4i32.nxv16i32(<vscale x 16 x i32> %data, i32 3)
48 // CHECK-DAG: %[[PG:.*]] = call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg)
49 // CHECK: call void @llvm.aarch64.sve.st4.nxv4i32(<vscale x 4 x i32> %[[V0]], <vscale x 4 x i32> %[[V1]], <vscale x 4 x i32> %[[V2]], <vscale x 4 x i32> %[[V3]], <vscale x 4 x i1> %[[PG]], i32* %base)
50 // CHECK-NEXT: ret
51 return SVE_ACLE_FUNC(svst4,_s32,,)(pg, base, data);
52 }
53
test_svst4_s64(svbool_t pg,int64_t * base,svint64x4_t data)54 void test_svst4_s64(svbool_t pg, int64_t *base, svint64x4_t data)
55 {
56 // CHECK-LABEL: test_svst4_s64
57 // CHECK-DAG: %[[V0:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv2i64.nxv8i64(<vscale x 8 x i64> %data, i32 0)
58 // CHECK-DAG: %[[V1:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv2i64.nxv8i64(<vscale x 8 x i64> %data, i32 1)
59 // CHECK-DAG: %[[V2:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv2i64.nxv8i64(<vscale x 8 x i64> %data, i32 2)
60 // CHECK-DAG: %[[V3:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv2i64.nxv8i64(<vscale x 8 x i64> %data, i32 3)
61 // CHECK-DAG: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg)
62 // CHECK: call void @llvm.aarch64.sve.st4.nxv2i64(<vscale x 2 x i64> %[[V0]], <vscale x 2 x i64> %[[V1]], <vscale x 2 x i64> %[[V2]], <vscale x 2 x i64> %[[V3]], <vscale x 2 x i1> %[[PG]], i64* %base)
63 // CHECK-NEXT: ret
64 return SVE_ACLE_FUNC(svst4,_s64,,)(pg, base, data);
65 }
66
test_svst4_u8(svbool_t pg,uint8_t * base,svuint8x4_t data)67 void test_svst4_u8(svbool_t pg, uint8_t *base, svuint8x4_t data)
68 {
69 // CHECK-LABEL: test_svst4_u8
70 // CHECK-DAG: %[[V0:.*]] = call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv16i8.nxv64i8(<vscale x 64 x i8> %data, i32 0)
71 // CHECK-DAG: %[[V1:.*]] = call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv16i8.nxv64i8(<vscale x 64 x i8> %data, i32 1)
72 // CHECK-DAG: %[[V2:.*]] = call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv16i8.nxv64i8(<vscale x 64 x i8> %data, i32 2)
73 // CHECK-DAG: %[[V3:.*]] = call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv16i8.nxv64i8(<vscale x 64 x i8> %data, i32 3)
74 // CHECK: call void @llvm.aarch64.sve.st4.nxv16i8(<vscale x 16 x i8> %[[V0]], <vscale x 16 x i8> %[[V1]], <vscale x 16 x i8> %[[V2]], <vscale x 16 x i8> %[[V3]], <vscale x 16 x i1> %pg, i8* %base)
75 // CHECK-NEXT: ret
76 return SVE_ACLE_FUNC(svst4,_u8,,)(pg, base, data);
77 }
78
test_svst4_u16(svbool_t pg,uint16_t * base,svuint16x4_t data)79 void test_svst4_u16(svbool_t pg, uint16_t *base, svuint16x4_t data)
80 {
81 // CHECK-LABEL: test_svst4_u16
82 // CHECK-DAG: %[[V0:.*]] = call <vscale x 8 x i16> @llvm.aarch64.sve.tuple.get.nxv8i16.nxv32i16(<vscale x 32 x i16> %data, i32 0)
83 // CHECK-DAG: %[[V1:.*]] = call <vscale x 8 x i16> @llvm.aarch64.sve.tuple.get.nxv8i16.nxv32i16(<vscale x 32 x i16> %data, i32 1)
84 // CHECK-DAG: %[[V2:.*]] = call <vscale x 8 x i16> @llvm.aarch64.sve.tuple.get.nxv8i16.nxv32i16(<vscale x 32 x i16> %data, i32 2)
85 // CHECK-DAG: %[[V3:.*]] = call <vscale x 8 x i16> @llvm.aarch64.sve.tuple.get.nxv8i16.nxv32i16(<vscale x 32 x i16> %data, i32 3)
86 // CHECK-DAG: %[[PG:.*]] = call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> %pg)
87 // CHECK: call void @llvm.aarch64.sve.st4.nxv8i16(<vscale x 8 x i16> %[[V0]], <vscale x 8 x i16> %[[V1]], <vscale x 8 x i16> %[[V2]], <vscale x 8 x i16> %[[V3]], <vscale x 8 x i1> %[[PG]], i16* %base)
88 // CHECK-NEXT: ret
89 return SVE_ACLE_FUNC(svst4,_u16,,)(pg, base, data);
90 }
91
test_svst4_u32(svbool_t pg,uint32_t * base,svuint32x4_t data)92 void test_svst4_u32(svbool_t pg, uint32_t *base, svuint32x4_t data)
93 {
94 // CHECK-LABEL: test_svst4_u32
95 // CHECK-DAG: %[[V0:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.tuple.get.nxv4i32.nxv16i32(<vscale x 16 x i32> %data, i32 0)
96 // CHECK-DAG: %[[V1:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.tuple.get.nxv4i32.nxv16i32(<vscale x 16 x i32> %data, i32 1)
97 // CHECK-DAG: %[[V2:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.tuple.get.nxv4i32.nxv16i32(<vscale x 16 x i32> %data, i32 2)
98 // CHECK-DAG: %[[V3:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.tuple.get.nxv4i32.nxv16i32(<vscale x 16 x i32> %data, i32 3)
99 // CHECK-DAG: %[[PG:.*]] = call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg)
100 // CHECK: call void @llvm.aarch64.sve.st4.nxv4i32(<vscale x 4 x i32> %[[V0]], <vscale x 4 x i32> %[[V1]], <vscale x 4 x i32> %[[V2]], <vscale x 4 x i32> %[[V3]], <vscale x 4 x i1> %[[PG]], i32* %base)
101 // CHECK-NEXT: ret
102 return SVE_ACLE_FUNC(svst4,_u32,,)(pg, base, data);
103 }
104
test_svst4_u64(svbool_t pg,uint64_t * base,svuint64x4_t data)105 void test_svst4_u64(svbool_t pg, uint64_t *base, svuint64x4_t data)
106 {
107 // CHECK-LABEL: test_svst4_u64
108 // CHECK-DAG: %[[V0:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv2i64.nxv8i64(<vscale x 8 x i64> %data, i32 0)
109 // CHECK-DAG: %[[V1:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv2i64.nxv8i64(<vscale x 8 x i64> %data, i32 1)
110 // CHECK-DAG: %[[V2:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv2i64.nxv8i64(<vscale x 8 x i64> %data, i32 2)
111 // CHECK-DAG: %[[V3:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv2i64.nxv8i64(<vscale x 8 x i64> %data, i32 3)
112 // CHECK-DAG: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg)
113 // CHECK: call void @llvm.aarch64.sve.st4.nxv2i64(<vscale x 2 x i64> %[[V0]], <vscale x 2 x i64> %[[V1]], <vscale x 2 x i64> %[[V2]], <vscale x 2 x i64> %[[V3]], <vscale x 2 x i1> %[[PG]], i64* %base)
114 // CHECK-NEXT: ret
115 return SVE_ACLE_FUNC(svst4,_u64,,)(pg, base, data);
116 }
117
test_svst4_f16(svbool_t pg,float16_t * base,svfloat16x4_t data)118 void test_svst4_f16(svbool_t pg, float16_t *base, svfloat16x4_t data)
119 {
120 // CHECK-LABEL: test_svst4_f16
121 // CHECK-DAG: %[[PG:.*]] = call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> %pg)
122 // CHECK-DAG: %[[V0:.*]] = call <vscale x 8 x half> @llvm.aarch64.sve.tuple.get.nxv8f16.nxv32f16(<vscale x 32 x half> %data, i32 0)
123 // CHECK-DAG: %[[V1:.*]] = call <vscale x 8 x half> @llvm.aarch64.sve.tuple.get.nxv8f16.nxv32f16(<vscale x 32 x half> %data, i32 1)
124 // CHECK-DAG: %[[V2:.*]] = call <vscale x 8 x half> @llvm.aarch64.sve.tuple.get.nxv8f16.nxv32f16(<vscale x 32 x half> %data, i32 2)
125 // CHECK-DAG: %[[V3:.*]] = call <vscale x 8 x half> @llvm.aarch64.sve.tuple.get.nxv8f16.nxv32f16(<vscale x 32 x half> %data, i32 3)
126 // CHECK: call void @llvm.aarch64.sve.st4.nxv8f16(<vscale x 8 x half> %[[V0]], <vscale x 8 x half> %[[V1]], <vscale x 8 x half> %[[V2]], <vscale x 8 x half> %[[V3]], <vscale x 8 x i1> %[[PG]], half* %base)
127 // CHECK-NEXT: ret
128 return SVE_ACLE_FUNC(svst4,_f16,,)(pg, base, data);
129 }
130
test_svst4_f32(svbool_t pg,float32_t * base,svfloat32x4_t data)131 void test_svst4_f32(svbool_t pg, float32_t *base, svfloat32x4_t data)
132 {
133 // CHECK-LABEL: test_svst4_f32
134 // CHECK-DAG: %[[V0:.*]] = call <vscale x 4 x float> @llvm.aarch64.sve.tuple.get.nxv4f32.nxv16f32(<vscale x 16 x float> %data, i32 0)
135 // CHECK-DAG: %[[V1:.*]] = call <vscale x 4 x float> @llvm.aarch64.sve.tuple.get.nxv4f32.nxv16f32(<vscale x 16 x float> %data, i32 1)
136 // CHECK-DAG: %[[V2:.*]] = call <vscale x 4 x float> @llvm.aarch64.sve.tuple.get.nxv4f32.nxv16f32(<vscale x 16 x float> %data, i32 2)
137 // CHECK-DAG: %[[V3:.*]] = call <vscale x 4 x float> @llvm.aarch64.sve.tuple.get.nxv4f32.nxv16f32(<vscale x 16 x float> %data, i32 3)
138 // CHECK-DAG: %[[PG:.*]] = call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg)
139 // CHECK: call void @llvm.aarch64.sve.st4.nxv4f32(<vscale x 4 x float> %[[V0]], <vscale x 4 x float> %[[V1]], <vscale x 4 x float> %[[V2]], <vscale x 4 x float> %[[V3]], <vscale x 4 x i1> %[[PG]], float* %base)
140 // CHECK-NEXT: ret
141 return SVE_ACLE_FUNC(svst4,_f32,,)(pg, base, data);
142 }
143
test_svst4_f64(svbool_t pg,float64_t * base,svfloat64x4_t data)144 void test_svst4_f64(svbool_t pg, float64_t *base, svfloat64x4_t data)
145 {
146 // CHECK-LABEL: test_svst4_f64
147 // CHECK-DAG: %[[V0:.*]] = call <vscale x 2 x double> @llvm.aarch64.sve.tuple.get.nxv2f64.nxv8f64(<vscale x 8 x double> %data, i32 0)
148 // CHECK-DAG: %[[V1:.*]] = call <vscale x 2 x double> @llvm.aarch64.sve.tuple.get.nxv2f64.nxv8f64(<vscale x 8 x double> %data, i32 1)
149 // CHECK-DAG: %[[V2:.*]] = call <vscale x 2 x double> @llvm.aarch64.sve.tuple.get.nxv2f64.nxv8f64(<vscale x 8 x double> %data, i32 2)
150 // CHECK-DAG: %[[V3:.*]] = call <vscale x 2 x double> @llvm.aarch64.sve.tuple.get.nxv2f64.nxv8f64(<vscale x 8 x double> %data, i32 3)
151 // CHECK-DAG: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg)
152 // CHECK: call void @llvm.aarch64.sve.st4.nxv2f64(<vscale x 2 x double> %[[V0]], <vscale x 2 x double> %[[V1]], <vscale x 2 x double> %[[V2]], <vscale x 2 x double> %[[V3]], <vscale x 2 x i1> %[[PG]], double* %base)
153 // CHECK-NEXT: ret
154 return SVE_ACLE_FUNC(svst4,_f64,,)(pg, base, data);
155 }
156
test_svst4_vnum_s8(svbool_t pg,int8_t * base,int64_t vnum,svint8x4_t data)157 void test_svst4_vnum_s8(svbool_t pg, int8_t *base, int64_t vnum, svint8x4_t data)
158 {
159 // CHECK-LABEL: test_svst4_vnum_s8
160 // CHECK-DAG: %[[BITCAST:.*]] = bitcast i8* %base to <vscale x 16 x i8>*
161 // CHECK-DAG: %[[GEP:.*]] = getelementptr <vscale x 16 x i8>, <vscale x 16 x i8>* %[[BITCAST]], i64 %vnum, i64 0
162 // CHECK-DAG: %[[V0:.*]] = call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv16i8.nxv64i8(<vscale x 64 x i8> %data, i32 0)
163 // CHECK-DAG: %[[V1:.*]] = call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv16i8.nxv64i8(<vscale x 64 x i8> %data, i32 1)
164 // CHECK-DAG: %[[V2:.*]] = call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv16i8.nxv64i8(<vscale x 64 x i8> %data, i32 2)
165 // CHECK-DAG: %[[V3:.*]] = call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv16i8.nxv64i8(<vscale x 64 x i8> %data, i32 3)
166 // CHECK: call void @llvm.aarch64.sve.st4.nxv16i8(<vscale x 16 x i8> %[[V0]], <vscale x 16 x i8> %[[V1]], <vscale x 16 x i8> %[[V2]], <vscale x 16 x i8> %[[V3]], <vscale x 16 x i1> %pg, i8* %[[GEP]])
167 // CHECK-NEXT: ret
168 return SVE_ACLE_FUNC(svst4_vnum,_s8,,)(pg, base, vnum, data);
169 }
170
test_svst4_vnum_s16(svbool_t pg,int16_t * base,int64_t vnum,svint16x4_t data)171 void test_svst4_vnum_s16(svbool_t pg, int16_t *base, int64_t vnum, svint16x4_t data)
172 {
173 // CHECK-LABEL: test_svst4_vnum_s16
174 // CHECK-DAG: %[[BITCAST:.*]] = bitcast i16* %base to <vscale x 8 x i16>*
175 // CHECK-DAG: %[[GEP:.*]] = getelementptr <vscale x 8 x i16>, <vscale x 8 x i16>* %[[BITCAST]], i64 %vnum, i64 0
176 // CHECK-DAG: %[[V0:.*]] = call <vscale x 8 x i16> @llvm.aarch64.sve.tuple.get.nxv8i16.nxv32i16(<vscale x 32 x i16> %data, i32 0)
177 // CHECK-DAG: %[[V1:.*]] = call <vscale x 8 x i16> @llvm.aarch64.sve.tuple.get.nxv8i16.nxv32i16(<vscale x 32 x i16> %data, i32 1)
178 // CHECK-DAG: %[[V2:.*]] = call <vscale x 8 x i16> @llvm.aarch64.sve.tuple.get.nxv8i16.nxv32i16(<vscale x 32 x i16> %data, i32 2)
179 // CHECK-DAG: %[[V3:.*]] = call <vscale x 8 x i16> @llvm.aarch64.sve.tuple.get.nxv8i16.nxv32i16(<vscale x 32 x i16> %data, i32 3)
180 // CHECK-DAG: %[[PG:.*]] = call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> %pg)
181 // CHECK: call void @llvm.aarch64.sve.st4.nxv8i16(<vscale x 8 x i16> %[[V0]], <vscale x 8 x i16> %[[V1]], <vscale x 8 x i16> %[[V2]], <vscale x 8 x i16> %[[V3]], <vscale x 8 x i1> %[[PG]], i16* %[[GEP]])
182 // CHECK-NEXT: ret
183 return SVE_ACLE_FUNC(svst4_vnum,_s16,,)(pg, base, vnum, data);
184 }
185
test_svst4_vnum_s32(svbool_t pg,int32_t * base,int64_t vnum,svint32x4_t data)186 void test_svst4_vnum_s32(svbool_t pg, int32_t *base, int64_t vnum, svint32x4_t data)
187 {
188 // CHECK-LABEL: test_svst4_vnum_s32
189 // CHECK-DAG: %[[BITCAST:.*]] = bitcast i32* %base to <vscale x 4 x i32>*
190 // CHECK-DAG: %[[GEP:.*]] = getelementptr <vscale x 4 x i32>, <vscale x 4 x i32>* %[[BITCAST]], i64 %vnum, i64 0
191 // CHECK-DAG: %[[V0:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.tuple.get.nxv4i32.nxv16i32(<vscale x 16 x i32> %data, i32 0)
192 // CHECK-DAG: %[[V1:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.tuple.get.nxv4i32.nxv16i32(<vscale x 16 x i32> %data, i32 1)
193 // CHECK-DAG: %[[V2:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.tuple.get.nxv4i32.nxv16i32(<vscale x 16 x i32> %data, i32 2)
194 // CHECK-DAG: %[[V3:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.tuple.get.nxv4i32.nxv16i32(<vscale x 16 x i32> %data, i32 3)
195 // CHECK-DAG: %[[PG:.*]] = call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg)
196 // CHECK: call void @llvm.aarch64.sve.st4.nxv4i32(<vscale x 4 x i32> %[[V0]], <vscale x 4 x i32> %[[V1]], <vscale x 4 x i32> %[[V2]], <vscale x 4 x i32> %[[V3]], <vscale x 4 x i1> %[[PG]], i32* %[[GEP]])
197 // CHECK-NEXT: ret
198 return SVE_ACLE_FUNC(svst4_vnum,_s32,,)(pg, base, vnum, data);
199 }
200
test_svst4_vnum_s64(svbool_t pg,int64_t * base,int64_t vnum,svint64x4_t data)201 void test_svst4_vnum_s64(svbool_t pg, int64_t *base, int64_t vnum, svint64x4_t data)
202 {
203 // CHECK-LABEL: test_svst4_vnum_s64
204 // CHECK-DAG: %[[BITCAST:.*]] = bitcast i64* %base to <vscale x 2 x i64>*
205 // CHECK-DAG: %[[GEP:.*]] = getelementptr <vscale x 2 x i64>, <vscale x 2 x i64>* %[[BITCAST]], i64 %vnum, i64 0
206 // CHECK-DAG: %[[V0:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv2i64.nxv8i64(<vscale x 8 x i64> %data, i32 0)
207 // CHECK-DAG: %[[V1:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv2i64.nxv8i64(<vscale x 8 x i64> %data, i32 1)
208 // CHECK-DAG: %[[V2:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv2i64.nxv8i64(<vscale x 8 x i64> %data, i32 2)
209 // CHECK-DAG: %[[V3:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv2i64.nxv8i64(<vscale x 8 x i64> %data, i32 3)
210 // CHECK-DAG: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg)
211 // CHECK: call void @llvm.aarch64.sve.st4.nxv2i64(<vscale x 2 x i64> %[[V0]], <vscale x 2 x i64> %[[V1]], <vscale x 2 x i64> %[[V2]], <vscale x 2 x i64> %[[V3]], <vscale x 2 x i1> %[[PG]], i64* %[[GEP]])
212 // CHECK-NEXT: ret
213 return SVE_ACLE_FUNC(svst4_vnum,_s64,,)(pg, base, vnum, data);
214 }
215
test_svst4_vnum_u8(svbool_t pg,uint8_t * base,int64_t vnum,svuint8x4_t data)216 void test_svst4_vnum_u8(svbool_t pg, uint8_t *base, int64_t vnum, svuint8x4_t data)
217 {
218 // CHECK-LABEL: test_svst4_vnum_u8
219 // CHECK-DAG: %[[BITCAST:.*]] = bitcast i8* %base to <vscale x 16 x i8>*
220 // CHECK-DAG: %[[GEP:.*]] = getelementptr <vscale x 16 x i8>, <vscale x 16 x i8>* %[[BITCAST]], i64 %vnum, i64 0
221 // CHECK-DAG: %[[V0:.*]] = call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv16i8.nxv64i8(<vscale x 64 x i8> %data, i32 0)
222 // CHECK-DAG: %[[V1:.*]] = call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv16i8.nxv64i8(<vscale x 64 x i8> %data, i32 1)
223 // CHECK-DAG: %[[V2:.*]] = call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv16i8.nxv64i8(<vscale x 64 x i8> %data, i32 2)
224 // CHECK-DAG: %[[V3:.*]] = call <vscale x 16 x i8> @llvm.aarch64.sve.tuple.get.nxv16i8.nxv64i8(<vscale x 64 x i8> %data, i32 3)
225 // CHECK: call void @llvm.aarch64.sve.st4.nxv16i8(<vscale x 16 x i8> %[[V0]], <vscale x 16 x i8> %[[V1]], <vscale x 16 x i8> %[[V2]], <vscale x 16 x i8> %[[V3]], <vscale x 16 x i1> %pg, i8* %[[GEP]])
226 // CHECK-NEXT: ret
227 return SVE_ACLE_FUNC(svst4_vnum,_u8,,)(pg, base, vnum, data);
228 }
229
test_svst4_vnum_u16(svbool_t pg,uint16_t * base,int64_t vnum,svuint16x4_t data)230 void test_svst4_vnum_u16(svbool_t pg, uint16_t *base, int64_t vnum, svuint16x4_t data)
231 {
232 // CHECK-LABEL: test_svst4_vnum_u16
233 // CHECK-DAG: %[[BITCAST:.*]] = bitcast i16* %base to <vscale x 8 x i16>*
234 // CHECK-DAG: %[[GEP:.*]] = getelementptr <vscale x 8 x i16>, <vscale x 8 x i16>* %[[BITCAST]], i64 %vnum, i64 0
235 // CHECK-DAG: %[[V0:.*]] = call <vscale x 8 x i16> @llvm.aarch64.sve.tuple.get.nxv8i16.nxv32i16(<vscale x 32 x i16> %data, i32 0)
236 // CHECK-DAG: %[[V1:.*]] = call <vscale x 8 x i16> @llvm.aarch64.sve.tuple.get.nxv8i16.nxv32i16(<vscale x 32 x i16> %data, i32 1)
237 // CHECK-DAG: %[[V2:.*]] = call <vscale x 8 x i16> @llvm.aarch64.sve.tuple.get.nxv8i16.nxv32i16(<vscale x 32 x i16> %data, i32 2)
238 // CHECK-DAG: %[[V3:.*]] = call <vscale x 8 x i16> @llvm.aarch64.sve.tuple.get.nxv8i16.nxv32i16(<vscale x 32 x i16> %data, i32 3)
239 // CHECK-DAG: %[[PG:.*]] = call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> %pg)
240 // CHECK: call void @llvm.aarch64.sve.st4.nxv8i16(<vscale x 8 x i16> %[[V0]], <vscale x 8 x i16> %[[V1]], <vscale x 8 x i16> %[[V2]], <vscale x 8 x i16> %[[V3]], <vscale x 8 x i1> %[[PG]], i16* %[[GEP]])
241 // CHECK-NEXT: ret
242 return SVE_ACLE_FUNC(svst4_vnum,_u16,,)(pg, base, vnum, data);
243 }
244
test_svst4_vnum_u32(svbool_t pg,uint32_t * base,int64_t vnum,svuint32x4_t data)245 void test_svst4_vnum_u32(svbool_t pg, uint32_t *base, int64_t vnum, svuint32x4_t data)
246 {
247 // CHECK-LABEL: test_svst4_vnum_u32
248 // CHECK-DAG: %[[BITCAST:.*]] = bitcast i32* %base to <vscale x 4 x i32>*
249 // CHECK-DAG: %[[GEP:.*]] = getelementptr <vscale x 4 x i32>, <vscale x 4 x i32>* %[[BITCAST]], i64 %vnum, i64 0
250 // CHECK-DAG: %[[V0:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.tuple.get.nxv4i32.nxv16i32(<vscale x 16 x i32> %data, i32 0)
251 // CHECK-DAG: %[[V1:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.tuple.get.nxv4i32.nxv16i32(<vscale x 16 x i32> %data, i32 1)
252 // CHECK-DAG: %[[V2:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.tuple.get.nxv4i32.nxv16i32(<vscale x 16 x i32> %data, i32 2)
253 // CHECK-DAG: %[[V3:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.tuple.get.nxv4i32.nxv16i32(<vscale x 16 x i32> %data, i32 3)
254 // CHECK-DAG: %[[PG:.*]] = call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg)
255 // CHECK: call void @llvm.aarch64.sve.st4.nxv4i32(<vscale x 4 x i32> %[[V0]], <vscale x 4 x i32> %[[V1]], <vscale x 4 x i32> %[[V2]], <vscale x 4 x i32> %[[V3]], <vscale x 4 x i1> %[[PG]], i32* %[[GEP]])
256 // CHECK-NEXT: ret
257 return SVE_ACLE_FUNC(svst4_vnum,_u32,,)(pg, base, vnum, data);
258 }
259
test_svst4_vnum_u64(svbool_t pg,uint64_t * base,int64_t vnum,svuint64x4_t data)260 void test_svst4_vnum_u64(svbool_t pg, uint64_t *base, int64_t vnum, svuint64x4_t data)
261 {
262 // CHECK-LABEL: test_svst4_vnum_u64
263 // CHECK-DAG: %[[BITCAST:.*]] = bitcast i64* %base to <vscale x 2 x i64>*
264 // CHECK-DAG: %[[GEP:.*]] = getelementptr <vscale x 2 x i64>, <vscale x 2 x i64>* %[[BITCAST]], i64 %vnum, i64 0
265 // CHECK-DAG: %[[V0:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv2i64.nxv8i64(<vscale x 8 x i64> %data, i32 0)
266 // CHECK-DAG: %[[V1:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv2i64.nxv8i64(<vscale x 8 x i64> %data, i32 1)
267 // CHECK-DAG: %[[V2:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv2i64.nxv8i64(<vscale x 8 x i64> %data, i32 2)
268 // CHECK-DAG: %[[V3:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.tuple.get.nxv2i64.nxv8i64(<vscale x 8 x i64> %data, i32 3)
269 // CHECK-DAG: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg)
270 // CHECK: call void @llvm.aarch64.sve.st4.nxv2i64(<vscale x 2 x i64> %[[V0]], <vscale x 2 x i64> %[[V1]], <vscale x 2 x i64> %[[V2]], <vscale x 2 x i64> %[[V3]], <vscale x 2 x i1> %[[PG]], i64* %[[GEP]])
271 // CHECK-NEXT: ret
272 return SVE_ACLE_FUNC(svst4_vnum,_u64,,)(pg, base, vnum, data);
273 }
274
test_svst4_vnum_f16(svbool_t pg,float16_t * base,int64_t vnum,svfloat16x4_t data)275 void test_svst4_vnum_f16(svbool_t pg, float16_t *base, int64_t vnum, svfloat16x4_t data)
276 {
277 // CHECK-LABEL: test_svst4_vnum_f16
278 // CHECK-DAG: %[[BITCAST:.*]] = bitcast half* %base to <vscale x 8 x half>*
279 // CHECK-DAG: %[[GEP:.*]] = getelementptr <vscale x 8 x half>, <vscale x 8 x half>* %[[BITCAST]], i64 %vnum, i64 0
280 // CHECK-DAG: %[[V0:.*]] = call <vscale x 8 x half> @llvm.aarch64.sve.tuple.get.nxv8f16.nxv32f16(<vscale x 32 x half> %data, i32 0)
281 // CHECK-DAG: %[[V1:.*]] = call <vscale x 8 x half> @llvm.aarch64.sve.tuple.get.nxv8f16.nxv32f16(<vscale x 32 x half> %data, i32 1)
282 // CHECK-DAG: %[[V2:.*]] = call <vscale x 8 x half> @llvm.aarch64.sve.tuple.get.nxv8f16.nxv32f16(<vscale x 32 x half> %data, i32 2)
283 // CHECK-DAG: %[[V3:.*]] = call <vscale x 8 x half> @llvm.aarch64.sve.tuple.get.nxv8f16.nxv32f16(<vscale x 32 x half> %data, i32 3)
284 // CHECK-DAG: %[[PG:.*]] = call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> %pg)
285 // CHECK: call void @llvm.aarch64.sve.st4.nxv8f16(<vscale x 8 x half> %[[V0]], <vscale x 8 x half> %[[V1]], <vscale x 8 x half> %[[V2]], <vscale x 8 x half> %[[V3]], <vscale x 8 x i1> %[[PG]], half* %[[GEP]])
286 // CHECK-NEXT: ret
287 return SVE_ACLE_FUNC(svst4_vnum,_f16,,)(pg, base, vnum, data);
288 }
289
test_svst4_vnum_f32(svbool_t pg,float32_t * base,int64_t vnum,svfloat32x4_t data)290 void test_svst4_vnum_f32(svbool_t pg, float32_t *base, int64_t vnum, svfloat32x4_t data)
291 {
292 // CHECK-LABEL: test_svst4_vnum_f32
293 // CHECK-DAG: %[[BITCAST:.*]] = bitcast float* %base to <vscale x 4 x float>*
294 // CHECK-DAG: %[[GEP:.*]] = getelementptr <vscale x 4 x float>, <vscale x 4 x float>* %[[BITCAST]], i64 %vnum, i64 0
295 // CHECK-DAG: %[[V0:.*]] = call <vscale x 4 x float> @llvm.aarch64.sve.tuple.get.nxv4f32.nxv16f32(<vscale x 16 x float> %data, i32 0)
296 // CHECK-DAG: %[[V1:.*]] = call <vscale x 4 x float> @llvm.aarch64.sve.tuple.get.nxv4f32.nxv16f32(<vscale x 16 x float> %data, i32 1)
297 // CHECK-DAG: %[[V2:.*]] = call <vscale x 4 x float> @llvm.aarch64.sve.tuple.get.nxv4f32.nxv16f32(<vscale x 16 x float> %data, i32 2)
298 // CHECK-DAG: %[[V3:.*]] = call <vscale x 4 x float> @llvm.aarch64.sve.tuple.get.nxv4f32.nxv16f32(<vscale x 16 x float> %data, i32 3)
299 // CHECK-DAG: %[[PG:.*]] = call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> %pg)
300 // CHECK: call void @llvm.aarch64.sve.st4.nxv4f32(<vscale x 4 x float> %[[V0]], <vscale x 4 x float> %[[V1]], <vscale x 4 x float> %[[V2]], <vscale x 4 x float> %[[V3]], <vscale x 4 x i1> %[[PG]], float* %[[GEP]])
301 // CHECK-NEXT: ret
302 return SVE_ACLE_FUNC(svst4_vnum,_f32,,)(pg, base, vnum, data);
303 }
304
test_svst4_vnum_f64(svbool_t pg,float64_t * base,int64_t vnum,svfloat64x4_t data)305 void test_svst4_vnum_f64(svbool_t pg, float64_t *base, int64_t vnum, svfloat64x4_t data)
306 {
307 // CHECK-LABEL: test_svst4_vnum_f64
308 // CHECK-DAG: %[[BITCAST:.*]] = bitcast double* %base to <vscale x 2 x double>*
309 // CHECK-DAG: %[[GEP:.*]] = getelementptr <vscale x 2 x double>, <vscale x 2 x double>* %[[BITCAST]], i64 %vnum, i64 0
310 // CHECK-DAG: %[[V0:.*]] = call <vscale x 2 x double> @llvm.aarch64.sve.tuple.get.nxv2f64.nxv8f64(<vscale x 8 x double> %data, i32 0)
311 // CHECK-DAG: %[[V1:.*]] = call <vscale x 2 x double> @llvm.aarch64.sve.tuple.get.nxv2f64.nxv8f64(<vscale x 8 x double> %data, i32 1)
312 // CHECK-DAG: %[[V2:.*]] = call <vscale x 2 x double> @llvm.aarch64.sve.tuple.get.nxv2f64.nxv8f64(<vscale x 8 x double> %data, i32 2)
313 // CHECK-DAG: %[[V3:.*]] = call <vscale x 2 x double> @llvm.aarch64.sve.tuple.get.nxv2f64.nxv8f64(<vscale x 8 x double> %data, i32 3)
314 // CHECK-DAG: %[[PG:.*]] = call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> %pg)
315 // CHECK: call void @llvm.aarch64.sve.st4.nxv2f64(<vscale x 2 x double> %[[V0]], <vscale x 2 x double> %[[V1]], <vscale x 2 x double> %[[V2]], <vscale x 2 x double> %[[V3]], <vscale x 2 x i1> %[[PG]], double* %[[GEP]])
316 // CHECK-NEXT: ret
317 return SVE_ACLE_FUNC(svst4_vnum,_f64,,)(pg, base, vnum, data);
318 }
319