1 // NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py
2 // RUN: %clang_cc1 -triple thumbv8.1m.main-none-none-eabi -target-feature +mve.fp -mfloat-abi hard -O0 -disable-O0-optnone -emit-llvm -o - %s | opt -S -passes=mem2reg | FileCheck %s
3 // RUN: %clang_cc1 -triple thumbv8.1m.main-none-none-eabi -target-feature +mve.fp -mfloat-abi hard -O0 -disable-O0-optnone -DPOLYMORPHIC -emit-llvm -o - %s | opt -S -passes=mem2reg | FileCheck %s
5 // REQUIRES: aarch64-registered-target || arm-registered-target
9 // CHECK-LABEL: @test_vhaddq_u8(
11 // CHECK-NEXT: [[TMP0:%.*]] = call <16 x i8> @llvm.arm.mve.vhadd.v16i8(<16 x i8> [[A:%.*]], <16 x i8> [[B:%.*]], i32 1)
12 // CHECK-NEXT: ret <16 x i8> [[TMP0]]
14 uint8x16_t
test_vhaddq_u8(uint8x16_t a
, uint8x16_t b
)
18 #else /* POLYMORPHIC */
19 return vhaddq_u8(a
, b
);
20 #endif /* POLYMORPHIC */
23 // CHECK-LABEL: @test_vhaddq_s16(
25 // CHECK-NEXT: [[TMP0:%.*]] = call <8 x i16> @llvm.arm.mve.vhadd.v8i16(<8 x i16> [[A:%.*]], <8 x i16> [[B:%.*]], i32 0)
26 // CHECK-NEXT: ret <8 x i16> [[TMP0]]
28 int16x8_t
test_vhaddq_s16(int16x8_t a
, int16x8_t b
)
32 #else /* POLYMORPHIC */
33 return vhaddq_s16(a
, b
);
34 #endif /* POLYMORPHIC */
37 // CHECK-LABEL: @test_vhaddq_u32(
39 // CHECK-NEXT: [[TMP0:%.*]] = call <4 x i32> @llvm.arm.mve.vhadd.v4i32(<4 x i32> [[A:%.*]], <4 x i32> [[B:%.*]], i32 1)
40 // CHECK-NEXT: ret <4 x i32> [[TMP0]]
42 uint32x4_t
test_vhaddq_u32(uint32x4_t a
, uint32x4_t b
)
46 #else /* POLYMORPHIC */
47 return vhaddq_u32(a
, b
);
48 #endif /* POLYMORPHIC */
51 // CHECK-LABEL: @test_vhaddq_m_s8(
53 // CHECK-NEXT: [[TMP0:%.*]] = zext i16 [[P:%.*]] to i32
54 // CHECK-NEXT: [[TMP1:%.*]] = call <16 x i1> @llvm.arm.mve.pred.i2v.v16i1(i32 [[TMP0]])
55 // CHECK-NEXT: [[TMP2:%.*]] = call <16 x i8> @llvm.arm.mve.hadd.predicated.v16i8.v16i1(<16 x i8> [[A:%.*]], <16 x i8> [[B:%.*]], i32 0, <16 x i1> [[TMP1]], <16 x i8> [[INACTIVE:%.*]])
56 // CHECK-NEXT: ret <16 x i8> [[TMP2]]
58 int8x16_t
test_vhaddq_m_s8(int8x16_t inactive
, int8x16_t a
, int8x16_t b
, mve_pred16_t p
)
61 return vhaddq_m(inactive
, a
, b
, p
);
62 #else /* POLYMORPHIC */
63 return vhaddq_m_s8(inactive
, a
, b
, p
);
64 #endif /* POLYMORPHIC */
67 // CHECK-LABEL: @test_vhaddq_m_u16(
69 // CHECK-NEXT: [[TMP0:%.*]] = zext i16 [[P:%.*]] to i32
70 // CHECK-NEXT: [[TMP1:%.*]] = call <8 x i1> @llvm.arm.mve.pred.i2v.v8i1(i32 [[TMP0]])
71 // CHECK-NEXT: [[TMP2:%.*]] = call <8 x i16> @llvm.arm.mve.hadd.predicated.v8i16.v8i1(<8 x i16> [[A:%.*]], <8 x i16> [[B:%.*]], i32 1, <8 x i1> [[TMP1]], <8 x i16> [[INACTIVE:%.*]])
72 // CHECK-NEXT: ret <8 x i16> [[TMP2]]
74 uint16x8_t
test_vhaddq_m_u16(uint16x8_t inactive
, uint16x8_t a
, uint16x8_t b
, mve_pred16_t p
)
77 return vhaddq_m(inactive
, a
, b
, p
);
78 #else /* POLYMORPHIC */
79 return vhaddq_m_u16(inactive
, a
, b
, p
);
80 #endif /* POLYMORPHIC */
83 // CHECK-LABEL: @test_vhaddq_m_s32(
85 // CHECK-NEXT: [[TMP0:%.*]] = zext i16 [[P:%.*]] to i32
86 // CHECK-NEXT: [[TMP1:%.*]] = call <4 x i1> @llvm.arm.mve.pred.i2v.v4i1(i32 [[TMP0]])
87 // CHECK-NEXT: [[TMP2:%.*]] = call <4 x i32> @llvm.arm.mve.hadd.predicated.v4i32.v4i1(<4 x i32> [[A:%.*]], <4 x i32> [[B:%.*]], i32 0, <4 x i1> [[TMP1]], <4 x i32> [[INACTIVE:%.*]])
88 // CHECK-NEXT: ret <4 x i32> [[TMP2]]
90 int32x4_t
test_vhaddq_m_s32(int32x4_t inactive
, int32x4_t a
, int32x4_t b
, mve_pred16_t p
)
93 return vhaddq_m(inactive
, a
, b
, p
);
94 #else /* POLYMORPHIC */
95 return vhaddq_m_s32(inactive
, a
, b
, p
);
96 #endif /* POLYMORPHIC */
99 // CHECK-LABEL: @test_vhaddq_x_u8(
100 // CHECK-NEXT: entry:
101 // CHECK-NEXT: [[TMP0:%.*]] = zext i16 [[P:%.*]] to i32
102 // CHECK-NEXT: [[TMP1:%.*]] = call <16 x i1> @llvm.arm.mve.pred.i2v.v16i1(i32 [[TMP0]])
103 // CHECK-NEXT: [[TMP2:%.*]] = call <16 x i8> @llvm.arm.mve.hadd.predicated.v16i8.v16i1(<16 x i8> [[A:%.*]], <16 x i8> [[B:%.*]], i32 1, <16 x i1> [[TMP1]], <16 x i8> undef)
104 // CHECK-NEXT: ret <16 x i8> [[TMP2]]
106 uint8x16_t
test_vhaddq_x_u8(uint8x16_t a
, uint8x16_t b
, mve_pred16_t p
)
109 return vhaddq_x(a
, b
, p
);
110 #else /* POLYMORPHIC */
111 return vhaddq_x_u8(a
, b
, p
);
112 #endif /* POLYMORPHIC */
115 // CHECK-LABEL: @test_vhaddq_x_s16(
116 // CHECK-NEXT: entry:
117 // CHECK-NEXT: [[TMP0:%.*]] = zext i16 [[P:%.*]] to i32
118 // CHECK-NEXT: [[TMP1:%.*]] = call <8 x i1> @llvm.arm.mve.pred.i2v.v8i1(i32 [[TMP0]])
119 // CHECK-NEXT: [[TMP2:%.*]] = call <8 x i16> @llvm.arm.mve.hadd.predicated.v8i16.v8i1(<8 x i16> [[A:%.*]], <8 x i16> [[B:%.*]], i32 0, <8 x i1> [[TMP1]], <8 x i16> undef)
120 // CHECK-NEXT: ret <8 x i16> [[TMP2]]
122 int16x8_t
test_vhaddq_x_s16(int16x8_t a
, int16x8_t b
, mve_pred16_t p
)
125 return vhaddq_x(a
, b
, p
);
126 #else /* POLYMORPHIC */
127 return vhaddq_x_s16(a
, b
, p
);
128 #endif /* POLYMORPHIC */
131 // CHECK-LABEL: @test_vhaddq_x_u32(
132 // CHECK-NEXT: entry:
133 // CHECK-NEXT: [[TMP0:%.*]] = zext i16 [[P:%.*]] to i32
134 // CHECK-NEXT: [[TMP1:%.*]] = call <4 x i1> @llvm.arm.mve.pred.i2v.v4i1(i32 [[TMP0]])
135 // CHECK-NEXT: [[TMP2:%.*]] = call <4 x i32> @llvm.arm.mve.hadd.predicated.v4i32.v4i1(<4 x i32> [[A:%.*]], <4 x i32> [[B:%.*]], i32 1, <4 x i1> [[TMP1]], <4 x i32> undef)
136 // CHECK-NEXT: ret <4 x i32> [[TMP2]]
138 uint32x4_t
test_vhaddq_x_u32(uint32x4_t a
, uint32x4_t b
, mve_pred16_t p
)
141 return vhaddq_x(a
, b
, p
);
142 #else /* POLYMORPHIC */
143 return vhaddq_x_u32(a
, b
, p
);
144 #endif /* POLYMORPHIC */
147 // CHECK-LABEL: @test_vhaddq_n_u8(
148 // CHECK-NEXT: entry:
149 // CHECK-NEXT: [[DOTSPLATINSERT:%.*]] = insertelement <16 x i8> poison, i8 [[B:%.*]], i64 0
150 // CHECK-NEXT: [[DOTSPLAT:%.*]] = shufflevector <16 x i8> [[DOTSPLATINSERT]], <16 x i8> poison, <16 x i32> zeroinitializer
151 // CHECK-NEXT: [[TMP0:%.*]] = call <16 x i8> @llvm.arm.mve.vhadd.v16i8(<16 x i8> [[A:%.*]], <16 x i8> [[DOTSPLAT]], i32 1)
152 // CHECK-NEXT: ret <16 x i8> [[TMP0]]
154 uint8x16_t
test_vhaddq_n_u8(uint8x16_t a
, uint8_t b
)
158 #else /* POLYMORPHIC */
159 return vhaddq_n_u8(a
, b
);
160 #endif /* POLYMORPHIC */
163 // CHECK-LABEL: @test_vhaddq_n_s16(
164 // CHECK-NEXT: entry:
165 // CHECK-NEXT: [[DOTSPLATINSERT:%.*]] = insertelement <8 x i16> poison, i16 [[B:%.*]], i64 0
166 // CHECK-NEXT: [[DOTSPLAT:%.*]] = shufflevector <8 x i16> [[DOTSPLATINSERT]], <8 x i16> poison, <8 x i32> zeroinitializer
167 // CHECK-NEXT: [[TMP0:%.*]] = call <8 x i16> @llvm.arm.mve.vhadd.v8i16(<8 x i16> [[A:%.*]], <8 x i16> [[DOTSPLAT]], i32 0)
168 // CHECK-NEXT: ret <8 x i16> [[TMP0]]
170 int16x8_t
test_vhaddq_n_s16(int16x8_t a
, int16_t b
)
174 #else /* POLYMORPHIC */
175 return vhaddq_n_s16(a
, b
);
176 #endif /* POLYMORPHIC */
179 // CHECK-LABEL: @test_vhaddq_n_u32(
180 // CHECK-NEXT: entry:
181 // CHECK-NEXT: [[DOTSPLATINSERT:%.*]] = insertelement <4 x i32> poison, i32 [[B:%.*]], i64 0
182 // CHECK-NEXT: [[DOTSPLAT:%.*]] = shufflevector <4 x i32> [[DOTSPLATINSERT]], <4 x i32> poison, <4 x i32> zeroinitializer
183 // CHECK-NEXT: [[TMP0:%.*]] = call <4 x i32> @llvm.arm.mve.vhadd.v4i32(<4 x i32> [[A:%.*]], <4 x i32> [[DOTSPLAT]], i32 1)
184 // CHECK-NEXT: ret <4 x i32> [[TMP0]]
186 uint32x4_t
test_vhaddq_n_u32(uint32x4_t a
, uint32_t b
)
190 #else /* POLYMORPHIC */
191 return vhaddq_n_u32(a
, b
);
192 #endif /* POLYMORPHIC */
195 // CHECK-LABEL: @test_vhaddq_m_n_s8(
196 // CHECK-NEXT: entry:
197 // CHECK-NEXT: [[DOTSPLATINSERT:%.*]] = insertelement <16 x i8> poison, i8 [[B:%.*]], i64 0
198 // CHECK-NEXT: [[DOTSPLAT:%.*]] = shufflevector <16 x i8> [[DOTSPLATINSERT]], <16 x i8> poison, <16 x i32> zeroinitializer
199 // CHECK-NEXT: [[TMP0:%.*]] = zext i16 [[P:%.*]] to i32
200 // CHECK-NEXT: [[TMP1:%.*]] = call <16 x i1> @llvm.arm.mve.pred.i2v.v16i1(i32 [[TMP0]])
201 // CHECK-NEXT: [[TMP2:%.*]] = call <16 x i8> @llvm.arm.mve.hadd.predicated.v16i8.v16i1(<16 x i8> [[A:%.*]], <16 x i8> [[DOTSPLAT]], i32 0, <16 x i1> [[TMP1]], <16 x i8> [[INACTIVE:%.*]])
202 // CHECK-NEXT: ret <16 x i8> [[TMP2]]
204 int8x16_t
test_vhaddq_m_n_s8(int8x16_t inactive
, int8x16_t a
, int8_t b
, mve_pred16_t p
)
207 return vhaddq_m(inactive
, a
, b
, p
);
208 #else /* POLYMORPHIC */
209 return vhaddq_m_n_s8(inactive
, a
, b
, p
);
210 #endif /* POLYMORPHIC */
213 // CHECK-LABEL: @test_vhaddq_m_n_u16(
214 // CHECK-NEXT: entry:
215 // CHECK-NEXT: [[DOTSPLATINSERT:%.*]] = insertelement <8 x i16> poison, i16 [[B:%.*]], i64 0
216 // CHECK-NEXT: [[DOTSPLAT:%.*]] = shufflevector <8 x i16> [[DOTSPLATINSERT]], <8 x i16> poison, <8 x i32> zeroinitializer
217 // CHECK-NEXT: [[TMP0:%.*]] = zext i16 [[P:%.*]] to i32
218 // CHECK-NEXT: [[TMP1:%.*]] = call <8 x i1> @llvm.arm.mve.pred.i2v.v8i1(i32 [[TMP0]])
219 // CHECK-NEXT: [[TMP2:%.*]] = call <8 x i16> @llvm.arm.mve.hadd.predicated.v8i16.v8i1(<8 x i16> [[A:%.*]], <8 x i16> [[DOTSPLAT]], i32 1, <8 x i1> [[TMP1]], <8 x i16> [[INACTIVE:%.*]])
220 // CHECK-NEXT: ret <8 x i16> [[TMP2]]
222 uint16x8_t
test_vhaddq_m_n_u16(uint16x8_t inactive
, uint16x8_t a
, uint16_t b
, mve_pred16_t p
)
225 return vhaddq_m(inactive
, a
, b
, p
);
226 #else /* POLYMORPHIC */
227 return vhaddq_m_n_u16(inactive
, a
, b
, p
);
228 #endif /* POLYMORPHIC */
231 // CHECK-LABEL: @test_vhaddq_m_n_s32(
232 // CHECK-NEXT: entry:
233 // CHECK-NEXT: [[DOTSPLATINSERT:%.*]] = insertelement <4 x i32> poison, i32 [[B:%.*]], i64 0
234 // CHECK-NEXT: [[DOTSPLAT:%.*]] = shufflevector <4 x i32> [[DOTSPLATINSERT]], <4 x i32> poison, <4 x i32> zeroinitializer
235 // CHECK-NEXT: [[TMP0:%.*]] = zext i16 [[P:%.*]] to i32
236 // CHECK-NEXT: [[TMP1:%.*]] = call <4 x i1> @llvm.arm.mve.pred.i2v.v4i1(i32 [[TMP0]])
237 // CHECK-NEXT: [[TMP2:%.*]] = call <4 x i32> @llvm.arm.mve.hadd.predicated.v4i32.v4i1(<4 x i32> [[A:%.*]], <4 x i32> [[DOTSPLAT]], i32 0, <4 x i1> [[TMP1]], <4 x i32> [[INACTIVE:%.*]])
238 // CHECK-NEXT: ret <4 x i32> [[TMP2]]
240 int32x4_t
test_vhaddq_m_n_s32(int32x4_t inactive
, int32x4_t a
, int32_t b
, mve_pred16_t p
)
243 return vhaddq_m(inactive
, a
, b
, p
);
244 #else /* POLYMORPHIC */
245 return vhaddq_m_n_s32(inactive
, a
, b
, p
);
246 #endif /* POLYMORPHIC */
249 // CHECK-LABEL: @test_vhaddq_x_n_u8(
250 // CHECK-NEXT: entry:
251 // CHECK-NEXT: [[DOTSPLATINSERT:%.*]] = insertelement <16 x i8> poison, i8 [[B:%.*]], i64 0
252 // CHECK-NEXT: [[DOTSPLAT:%.*]] = shufflevector <16 x i8> [[DOTSPLATINSERT]], <16 x i8> poison, <16 x i32> zeroinitializer
253 // CHECK-NEXT: [[TMP0:%.*]] = zext i16 [[P:%.*]] to i32
254 // CHECK-NEXT: [[TMP1:%.*]] = call <16 x i1> @llvm.arm.mve.pred.i2v.v16i1(i32 [[TMP0]])
255 // CHECK-NEXT: [[TMP2:%.*]] = call <16 x i8> @llvm.arm.mve.hadd.predicated.v16i8.v16i1(<16 x i8> [[A:%.*]], <16 x i8> [[DOTSPLAT]], i32 1, <16 x i1> [[TMP1]], <16 x i8> undef)
256 // CHECK-NEXT: ret <16 x i8> [[TMP2]]
258 uint8x16_t
test_vhaddq_x_n_u8(uint8x16_t a
, uint8_t b
, mve_pred16_t p
)
261 return vhaddq_x(a
, b
, p
);
262 #else /* POLYMORPHIC */
263 return vhaddq_x_n_u8(a
, b
, p
);
264 #endif /* POLYMORPHIC */
267 // CHECK-LABEL: @test_vhaddq_x_n_s16(
268 // CHECK-NEXT: entry:
269 // CHECK-NEXT: [[DOTSPLATINSERT:%.*]] = insertelement <8 x i16> poison, i16 [[B:%.*]], i64 0
270 // CHECK-NEXT: [[DOTSPLAT:%.*]] = shufflevector <8 x i16> [[DOTSPLATINSERT]], <8 x i16> poison, <8 x i32> zeroinitializer
271 // CHECK-NEXT: [[TMP0:%.*]] = zext i16 [[P:%.*]] to i32
272 // CHECK-NEXT: [[TMP1:%.*]] = call <8 x i1> @llvm.arm.mve.pred.i2v.v8i1(i32 [[TMP0]])
273 // CHECK-NEXT: [[TMP2:%.*]] = call <8 x i16> @llvm.arm.mve.hadd.predicated.v8i16.v8i1(<8 x i16> [[A:%.*]], <8 x i16> [[DOTSPLAT]], i32 0, <8 x i1> [[TMP1]], <8 x i16> undef)
274 // CHECK-NEXT: ret <8 x i16> [[TMP2]]
276 int16x8_t
test_vhaddq_x_n_s16(int16x8_t a
, int16_t b
, mve_pred16_t p
)
279 return vhaddq_x(a
, b
, p
);
280 #else /* POLYMORPHIC */
281 return vhaddq_x_n_s16(a
, b
, p
);
282 #endif /* POLYMORPHIC */
285 // CHECK-LABEL: @test_vhaddq_x_n_u32(
286 // CHECK-NEXT: entry:
287 // CHECK-NEXT: [[DOTSPLATINSERT:%.*]] = insertelement <4 x i32> poison, i32 [[B:%.*]], i64 0
288 // CHECK-NEXT: [[DOTSPLAT:%.*]] = shufflevector <4 x i32> [[DOTSPLATINSERT]], <4 x i32> poison, <4 x i32> zeroinitializer
289 // CHECK-NEXT: [[TMP0:%.*]] = zext i16 [[P:%.*]] to i32
290 // CHECK-NEXT: [[TMP1:%.*]] = call <4 x i1> @llvm.arm.mve.pred.i2v.v4i1(i32 [[TMP0]])
291 // CHECK-NEXT: [[TMP2:%.*]] = call <4 x i32> @llvm.arm.mve.hadd.predicated.v4i32.v4i1(<4 x i32> [[A:%.*]], <4 x i32> [[DOTSPLAT]], i32 1, <4 x i1> [[TMP1]], <4 x i32> undef)
292 // CHECK-NEXT: ret <4 x i32> [[TMP2]]
294 uint32x4_t
test_vhaddq_x_n_u32(uint32x4_t a
, uint32_t b
, mve_pred16_t p
)
297 return vhaddq_x(a
, b
, p
);
298 #else /* POLYMORPHIC */
299 return vhaddq_x_n_u32(a
, b
, p
);
300 #endif /* POLYMORPHIC */