clang/test/CodeGen/arm-mve-intrinsics/vmldav.c

   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 -S -emit-llvm -o - %s | opt -S -passes=mem2reg,sroa | 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 -S -emit-llvm -o - %s | opt -S -passes=mem2reg,sroa | FileCheck %s
   4
   5 // REQUIRES: aarch64-registered-target || arm-registered-target
   6
   7 #include <arm_mve.h>
   8
   9 // CHECK-LABEL: @test_vmladavaq_s8(
  10 // CHECK-NEXT:  entry:
  11 // CHECK-NEXT:    [[TMP0:%.*]] = call i32 @llvm.arm.mve.vmldava.v16i8(i32 0, i32 0, i32 0, i32 [[A:%.*]], <16 x i8> [[B:%.*]], <16 x i8> [[C:%.*]])
  12 // CHECK-NEXT:    ret i32 [[TMP0]]
  13 //
  14 int32_t test_vmladavaq_s8(int32_t a, int8x16_t b, int8x16_t c) {
  15 #ifdef POLYMORPHIC
  16   return vmladavaq(a, b, c);
  17 #else
  18   return vmladavaq_s8(a, b, c);
  19 #endif
  20 }
  21
  22 // CHECK-LABEL: @test_vmladavaq_s16(
  23 // CHECK-NEXT:  entry:
  24 // CHECK-NEXT:    [[TMP0:%.*]] = call i32 @llvm.arm.mve.vmldava.v8i16(i32 0, i32 0, i32 0, i32 [[A:%.*]], <8 x i16> [[B:%.*]], <8 x i16> [[C:%.*]])
  25 // CHECK-NEXT:    ret i32 [[TMP0]]
  26 //
  27 int32_t test_vmladavaq_s16(int32_t a, int16x8_t b, int16x8_t c) {
  28 #ifdef POLYMORPHIC
  29   return vmladavaq(a, b, c);
  30 #else
  31   return vmladavaq_s16(a, b, c);
  32 #endif
  33 }
  34
  35 // CHECK-LABEL: @test_vmladavaq_s32(
  36 // CHECK-NEXT:  entry:
  37 // CHECK-NEXT:    [[TMP0:%.*]] = call i32 @llvm.arm.mve.vmldava.v4i32(i32 0, i32 0, i32 0, i32 [[A:%.*]], <4 x i32> [[B:%.*]], <4 x i32> [[C:%.*]])
  38 // CHECK-NEXT:    ret i32 [[TMP0]]
  39 //
  40 int32_t test_vmladavaq_s32(int32_t a, int32x4_t b, int32x4_t c) {
  41 #ifdef POLYMORPHIC
  42   return vmladavaq(a, b, c);
  43 #else
  44   return vmladavaq_s32(a, b, c);
  45 #endif
  46 }
  47
  48 // CHECK-LABEL: @test_vmladavaq_u8(
  49 // CHECK-NEXT:  entry:
  50 // CHECK-NEXT:    [[TMP0:%.*]] = call i32 @llvm.arm.mve.vmldava.v16i8(i32 1, i32 0, i32 0, i32 [[A:%.*]], <16 x i8> [[B:%.*]], <16 x i8> [[C:%.*]])
  51 // CHECK-NEXT:    ret i32 [[TMP0]]
  52 //
  53 uint32_t test_vmladavaq_u8(uint32_t a, uint8x16_t b, uint8x16_t c) {
  54 #ifdef POLYMORPHIC
  55   return vmladavaq(a, b, c);
  56 #else
  57   return vmladavaq_u8(a, b, c);
  58 #endif
  59 }
  60
  61 // CHECK-LABEL: @test_vmladavaq_u16(
  62 // CHECK-NEXT:  entry:
  63 // CHECK-NEXT:    [[TMP0:%.*]] = call i32 @llvm.arm.mve.vmldava.v8i16(i32 1, i32 0, i32 0, i32 [[A:%.*]], <8 x i16> [[B:%.*]], <8 x i16> [[C:%.*]])
  64 // CHECK-NEXT:    ret i32 [[TMP0]]
  65 //
  66 uint32_t test_vmladavaq_u16(uint32_t a, uint16x8_t b, uint16x8_t c) {
  67 #ifdef POLYMORPHIC
  68   return vmladavaq(a, b, c);
  69 #else
  70   return vmladavaq_u16(a, b, c);
  71 #endif
  72 }
  73
  74 // CHECK-LABEL: @test_vmladavaq_u32(
  75 // CHECK-NEXT:  entry:
  76 // CHECK-NEXT:    [[TMP0:%.*]] = call i32 @llvm.arm.mve.vmldava.v4i32(i32 1, i32 0, i32 0, i32 [[A:%.*]], <4 x i32> [[B:%.*]], <4 x i32> [[C:%.*]])
  77 // CHECK-NEXT:    ret i32 [[TMP0]]
  78 //
  79 uint32_t test_vmladavaq_u32(uint32_t a, uint32x4_t b, uint32x4_t c) {
  80 #ifdef POLYMORPHIC
  81   return vmladavaq(a, b, c);
  82 #else
  83   return vmladavaq_u32(a, b, c);
  84 #endif
  85 }
  86
  87 // CHECK-LABEL: @test_vmladavaxq_s8(
  88 // CHECK-NEXT:  entry:
  89 // CHECK-NEXT:    [[TMP0:%.*]] = call i32 @llvm.arm.mve.vmldava.v16i8(i32 0, i32 0, i32 1, i32 [[A:%.*]], <16 x i8> [[B:%.*]], <16 x i8> [[C:%.*]])
  90 // CHECK-NEXT:    ret i32 [[TMP0]]
  91 //
  92 int32_t test_vmladavaxq_s8(int32_t a, int8x16_t b, int8x16_t c) {
  93 #ifdef POLYMORPHIC
  94   return vmladavaxq(a, b, c);
  95 #else
  96   return vmladavaxq_s8(a, b, c);
  97 #endif
  98 }
  99
 100 // CHECK-LABEL: @test_vmladavaxq_s16(
 101 // CHECK-NEXT:  entry:
 102 // CHECK-NEXT:    [[TMP0:%.*]] = call i32 @llvm.arm.mve.vmldava.v8i16(i32 0, i32 0, i32 1, i32 [[A:%.*]], <8 x i16> [[B:%.*]], <8 x i16> [[C:%.*]])
 103 // CHECK-NEXT:    ret i32 [[TMP0]]
 104 //
 105 int32_t test_vmladavaxq_s16(int32_t a, int16x8_t b, int16x8_t c) {
 106 #ifdef POLYMORPHIC
 107   return vmladavaxq(a, b, c);
 108 #else
 109   return vmladavaxq_s16(a, b, c);
 110 #endif
 111 }
 112
 113 // CHECK-LABEL: @test_vmladavaxq_s32(
 114 // CHECK-NEXT:  entry:
 115 // CHECK-NEXT:    [[TMP0:%.*]] = call i32 @llvm.arm.mve.vmldava.v4i32(i32 0, i32 0, i32 1, i32 [[A:%.*]], <4 x i32> [[B:%.*]], <4 x i32> [[C:%.*]])
 116 // CHECK-NEXT:    ret i32 [[TMP0]]
 117 //
 118 int32_t test_vmladavaxq_s32(int32_t a, int32x4_t b, int32x4_t c) {
 119 #ifdef POLYMORPHIC
 120   return vmladavaxq(a, b, c);
 121 #else
 122   return vmladavaxq_s32(a, b, c);
 123 #endif
 124 }
 125
 126 // CHECK-LABEL: @test_vmlsdavaq_s8(
 127 // CHECK-NEXT:  entry:
 128 // CHECK-NEXT:    [[TMP0:%.*]] = call i32 @llvm.arm.mve.vmldava.v16i8(i32 0, i32 1, i32 0, i32 [[A:%.*]], <16 x i8> [[B:%.*]], <16 x i8> [[C:%.*]])
 129 // CHECK-NEXT:    ret i32 [[TMP0]]
 130 //
 131 int32_t test_vmlsdavaq_s8(int32_t a, int8x16_t b, int8x16_t c) {
 132 #ifdef POLYMORPHIC
 133   return vmlsdavaq(a, b, c);
 134 #else
 135   return vmlsdavaq_s8(a, b, c);
 136 #endif
 137 }
 138
 139 // CHECK-LABEL: @test_vmlsdavaq_s16(
 140 // CHECK-NEXT:  entry:
 141 // CHECK-NEXT:    [[TMP0:%.*]] = call i32 @llvm.arm.mve.vmldava.v8i16(i32 0, i32 1, i32 0, i32 [[A:%.*]], <8 x i16> [[B:%.*]], <8 x i16> [[C:%.*]])
 142 // CHECK-NEXT:    ret i32 [[TMP0]]
 143 //
 144 int32_t test_vmlsdavaq_s16(int32_t a, int16x8_t b, int16x8_t c) {
 145 #ifdef POLYMORPHIC
 146   return vmlsdavaq(a, b, c);
 147 #else
 148   return vmlsdavaq_s16(a, b, c);
 149 #endif
 150 }
 151
 152 // CHECK-LABEL: @test_vmlsdavaq_s32(
 153 // CHECK-NEXT:  entry:
 154 // CHECK-NEXT:    [[TMP0:%.*]] = call i32 @llvm.arm.mve.vmldava.v4i32(i32 0, i32 1, i32 0, i32 [[A:%.*]], <4 x i32> [[B:%.*]], <4 x i32> [[C:%.*]])
 155 // CHECK-NEXT:    ret i32 [[TMP0]]
 156 //
 157 int32_t test_vmlsdavaq_s32(int32_t a, int32x4_t b, int32x4_t c) {
 158 #ifdef POLYMORPHIC
 159   return vmlsdavaq(a, b, c);
 160 #else
 161   return vmlsdavaq_s32(a, b, c);
 162 #endif
 163 }
 164
 165 // CHECK-LABEL: @test_vmlsdavaxq_s8(
 166 // CHECK-NEXT:  entry:
 167 // CHECK-NEXT:    [[TMP0:%.*]] = call i32 @llvm.arm.mve.vmldava.v16i8(i32 0, i32 1, i32 1, i32 [[A:%.*]], <16 x i8> [[B:%.*]], <16 x i8> [[C:%.*]])
 168 // CHECK-NEXT:    ret i32 [[TMP0]]
 169 //
 170 int32_t test_vmlsdavaxq_s8(int32_t a, int8x16_t b, int8x16_t c) {
 171 #ifdef POLYMORPHIC
 172   return vmlsdavaxq(a, b, c);
 173 #else
 174   return vmlsdavaxq_s8(a, b, c);
 175 #endif
 176 }
 177
 178 // CHECK-LABEL: @test_vmlsdavaxq_s16(
 179 // CHECK-NEXT:  entry:
 180 // CHECK-NEXT:    [[TMP0:%.*]] = call i32 @llvm.arm.mve.vmldava.v8i16(i32 0, i32 1, i32 1, i32 [[A:%.*]], <8 x i16> [[B:%.*]], <8 x i16> [[C:%.*]])
 181 // CHECK-NEXT:    ret i32 [[TMP0]]
 182 //
 183 int32_t test_vmlsdavaxq_s16(int32_t a, int16x8_t b, int16x8_t c) {
 184 #ifdef POLYMORPHIC
 185   return vmlsdavaxq(a, b, c);
 186 #else
 187   return vmlsdavaxq_s16(a, b, c);
 188 #endif
 189 }
 190
 191 // CHECK-LABEL: @test_vmlsdavaxq_s32(
 192 // CHECK-NEXT:  entry:
 193 // CHECK-NEXT:    [[TMP0:%.*]] = call i32 @llvm.arm.mve.vmldava.v4i32(i32 0, i32 1, i32 1, i32 [[A:%.*]], <4 x i32> [[B:%.*]], <4 x i32> [[C:%.*]])
 194 // CHECK-NEXT:    ret i32 [[TMP0]]
 195 //
 196 int32_t test_vmlsdavaxq_s32(int32_t a, int32x4_t b, int32x4_t c) {
 197 #ifdef POLYMORPHIC
 198   return vmlsdavaxq(a, b, c);
 199 #else
 200   return vmlsdavaxq_s32(a, b, c);
 201 #endif
 202 }
 203
 204 // CHECK-LABEL: @test_vmladavaq_p_s8(
 205 // CHECK-NEXT:  entry:
 206 // CHECK-NEXT:    [[TMP0:%.*]] = zext i16 [[P:%.*]] to i32
 207 // CHECK-NEXT:    [[TMP1:%.*]] = call <16 x i1> @llvm.arm.mve.pred.i2v.v16i1(i32 [[TMP0]])
 208 // CHECK-NEXT:    [[TMP2:%.*]] = call i32 @llvm.arm.mve.vmldava.predicated.v16i8.v16i1(i32 0, i32 0, i32 0, i32 [[A:%.*]], <16 x i8> [[B:%.*]], <16 x i8> [[C:%.*]], <16 x i1> [[TMP1]])
 209 // CHECK-NEXT:    ret i32 [[TMP2]]
 210 //
 211 int32_t test_vmladavaq_p_s8(int32_t a, int8x16_t b, int8x16_t c, mve_pred16_t p) {
 212 #ifdef POLYMORPHIC
 213   return vmladavaq_p(a, b, c, p);
 214 #else
 215   return vmladavaq_p_s8(a, b, c, p);
 216 #endif
 217 }
 218
 219 // CHECK-LABEL: @test_vmladavaq_p_s16(
 220 // CHECK-NEXT:  entry:
 221 // CHECK-NEXT:    [[TMP0:%.*]] = zext i16 [[P:%.*]] to i32
 222 // CHECK-NEXT:    [[TMP1:%.*]] = call <8 x i1> @llvm.arm.mve.pred.i2v.v8i1(i32 [[TMP0]])
 223 // CHECK-NEXT:    [[TMP2:%.*]] = call i32 @llvm.arm.mve.vmldava.predicated.v8i16.v8i1(i32 0, i32 0, i32 0, i32 [[A:%.*]], <8 x i16> [[B:%.*]], <8 x i16> [[C:%.*]], <8 x i1> [[TMP1]])
 224 // CHECK-NEXT:    ret i32 [[TMP2]]
 225 //
 226 int32_t test_vmladavaq_p_s16(int32_t a, int16x8_t b, int16x8_t c, mve_pred16_t p) {
 227 #ifdef POLYMORPHIC
 228   return vmladavaq_p(a, b, c, p);
 229 #else
 230   return vmladavaq_p_s16(a, b, c, p);
 231 #endif
 232 }
 233
 234 // CHECK-LABEL: @test_vmladavaq_p_s32(
 235 // CHECK-NEXT:  entry:
 236 // CHECK-NEXT:    [[TMP0:%.*]] = zext i16 [[P:%.*]] to i32
 237 // CHECK-NEXT:    [[TMP1:%.*]] = call <4 x i1> @llvm.arm.mve.pred.i2v.v4i1(i32 [[TMP0]])
 238 // CHECK-NEXT:    [[TMP2:%.*]] = call i32 @llvm.arm.mve.vmldava.predicated.v4i32.v4i1(i32 0, i32 0, i32 0, i32 [[A:%.*]], <4 x i32> [[B:%.*]], <4 x i32> [[C:%.*]], <4 x i1> [[TMP1]])
 239 // CHECK-NEXT:    ret i32 [[TMP2]]
 240 //
 241 int32_t test_vmladavaq_p_s32(int32_t a, int32x4_t b, int32x4_t c, mve_pred16_t p) {
 242 #ifdef POLYMORPHIC
 243   return vmladavaq_p(a, b, c, p);
 244 #else
 245   return vmladavaq_p_s32(a, b, c, p);
 246 #endif
 247 }
 248
 249 // CHECK-LABEL: @test_vmladavaq_p_u8(
 250 // CHECK-NEXT:  entry:
 251 // CHECK-NEXT:    [[TMP0:%.*]] = zext i16 [[P:%.*]] to i32
 252 // CHECK-NEXT:    [[TMP1:%.*]] = call <16 x i1> @llvm.arm.mve.pred.i2v.v16i1(i32 [[TMP0]])
 253 // CHECK-NEXT:    [[TMP2:%.*]] = call i32 @llvm.arm.mve.vmldava.predicated.v16i8.v16i1(i32 1, i32 0, i32 0, i32 [[A:%.*]], <16 x i8> [[B:%.*]], <16 x i8> [[C:%.*]], <16 x i1> [[TMP1]])
 254 // CHECK-NEXT:    ret i32 [[TMP2]]
 255 //
 256 uint32_t test_vmladavaq_p_u8(uint32_t a, uint8x16_t b, uint8x16_t c, mve_pred16_t p) {
 257 #ifdef POLYMORPHIC
 258   return vmladavaq_p(a, b, c, p);
 259 #else
 260   return vmladavaq_p_u8(a, b, c, p);
 261 #endif
 262 }
 263
 264 // CHECK-LABEL: @test_vmladavaq_p_u16(
 265 // CHECK-NEXT:  entry:
 266 // CHECK-NEXT:    [[TMP0:%.*]] = zext i16 [[P:%.*]] to i32
 267 // CHECK-NEXT:    [[TMP1:%.*]] = call <8 x i1> @llvm.arm.mve.pred.i2v.v8i1(i32 [[TMP0]])
 268 // CHECK-NEXT:    [[TMP2:%.*]] = call i32 @llvm.arm.mve.vmldava.predicated.v8i16.v8i1(i32 1, i32 0, i32 0, i32 [[A:%.*]], <8 x i16> [[B:%.*]], <8 x i16> [[C:%.*]], <8 x i1> [[TMP1]])
 269 // CHECK-NEXT:    ret i32 [[TMP2]]
 270 //
 271 uint32_t test_vmladavaq_p_u16(uint32_t a, uint16x8_t b, uint16x8_t c, mve_pred16_t p) {
 272 #ifdef POLYMORPHIC
 273   return vmladavaq_p(a, b, c, p);
 274 #else
 275   return vmladavaq_p_u16(a, b, c, p);
 276 #endif
 277 }
 278
 279 // CHECK-LABEL: @test_vmladavaq_p_u32(
 280 // CHECK-NEXT:  entry:
 281 // CHECK-NEXT:    [[TMP0:%.*]] = zext i16 [[P:%.*]] to i32
 282 // CHECK-NEXT:    [[TMP1:%.*]] = call <4 x i1> @llvm.arm.mve.pred.i2v.v4i1(i32 [[TMP0]])
 283 // CHECK-NEXT:    [[TMP2:%.*]] = call i32 @llvm.arm.mve.vmldava.predicated.v4i32.v4i1(i32 1, i32 0, i32 0, i32 [[A:%.*]], <4 x i32> [[B:%.*]], <4 x i32> [[C:%.*]], <4 x i1> [[TMP1]])
 284 // CHECK-NEXT:    ret i32 [[TMP2]]
 285 //
 286 uint32_t test_vmladavaq_p_u32(uint32_t a, uint32x4_t b, uint32x4_t c, mve_pred16_t p) {
 287 #ifdef POLYMORPHIC
 288   return vmladavaq_p(a, b, c, p);
 289 #else
 290   return vmladavaq_p_u32(a, b, c, p);
 291 #endif
 292 }
 293
 294 // CHECK-LABEL: @test_vmladavaxq_p_s8(
 295 // CHECK-NEXT:  entry:
 296 // CHECK-NEXT:    [[TMP0:%.*]] = zext i16 [[P:%.*]] to i32
 297 // CHECK-NEXT:    [[TMP1:%.*]] = call <16 x i1> @llvm.arm.mve.pred.i2v.v16i1(i32 [[TMP0]])
 298 // CHECK-NEXT:    [[TMP2:%.*]] = call i32 @llvm.arm.mve.vmldava.predicated.v16i8.v16i1(i32 0, i32 0, i32 1, i32 [[A:%.*]], <16 x i8> [[B:%.*]], <16 x i8> [[C:%.*]], <16 x i1> [[TMP1]])
 299 // CHECK-NEXT:    ret i32 [[TMP2]]
 300 //
 301 int32_t test_vmladavaxq_p_s8(int32_t a, int8x16_t b, int8x16_t c, mve_pred16_t p) {
 302 #ifdef POLYMORPHIC
 303   return vmladavaxq_p(a, b, c, p);
 304 #else
 305   return vmladavaxq_p_s8(a, b, c, p);
 306 #endif
 307 }
 308
 309 // CHECK-LABEL: @test_vmladavaxq_p_s16(
 310 // CHECK-NEXT:  entry:
 311 // CHECK-NEXT:    [[TMP0:%.*]] = zext i16 [[P:%.*]] to i32
 312 // CHECK-NEXT:    [[TMP1:%.*]] = call <8 x i1> @llvm.arm.mve.pred.i2v.v8i1(i32 [[TMP0]])
 313 // CHECK-NEXT:    [[TMP2:%.*]] = call i32 @llvm.arm.mve.vmldava.predicated.v8i16.v8i1(i32 0, i32 0, i32 1, i32 [[A:%.*]], <8 x i16> [[B:%.*]], <8 x i16> [[C:%.*]], <8 x i1> [[TMP1]])
 314 // CHECK-NEXT:    ret i32 [[TMP2]]
 315 //
 316 int32_t test_vmladavaxq_p_s16(int32_t a, int16x8_t b, int16x8_t c, mve_pred16_t p) {
 317 #ifdef POLYMORPHIC
 318   return vmladavaxq_p(a, b, c, p);
 319 #else
 320   return vmladavaxq_p_s16(a, b, c, p);
 321 #endif
 322 }
 323
 324 // CHECK-LABEL: @test_vmladavaxq_p_s32(
 325 // CHECK-NEXT:  entry:
 326 // CHECK-NEXT:    [[TMP0:%.*]] = zext i16 [[P:%.*]] to i32
 327 // CHECK-NEXT:    [[TMP1:%.*]] = call <4 x i1> @llvm.arm.mve.pred.i2v.v4i1(i32 [[TMP0]])
 328 // CHECK-NEXT:    [[TMP2:%.*]] = call i32 @llvm.arm.mve.vmldava.predicated.v4i32.v4i1(i32 0, i32 0, i32 1, i32 [[A:%.*]], <4 x i32> [[B:%.*]], <4 x i32> [[C:%.*]], <4 x i1> [[TMP1]])
 329 // CHECK-NEXT:    ret i32 [[TMP2]]
 330 //
 331 int32_t test_vmladavaxq_p_s32(int32_t a, int32x4_t b, int32x4_t c, mve_pred16_t p) {
 332 #ifdef POLYMORPHIC
 333   return vmladavaxq_p(a, b, c, p);
 334 #else
 335   return vmladavaxq_p_s32(a, b, c, p);
 336 #endif
 337 }
 338
 339 // CHECK-LABEL: @test_vmlsdavaq_p_s8(
 340 // CHECK-NEXT:  entry:
 341 // CHECK-NEXT:    [[TMP0:%.*]] = zext i16 [[P:%.*]] to i32
 342 // CHECK-NEXT:    [[TMP1:%.*]] = call <16 x i1> @llvm.arm.mve.pred.i2v.v16i1(i32 [[TMP0]])
 343 // CHECK-NEXT:    [[TMP2:%.*]] = call i32 @llvm.arm.mve.vmldava.predicated.v16i8.v16i1(i32 0, i32 1, i32 0, i32 [[A:%.*]], <16 x i8> [[B:%.*]], <16 x i8> [[C:%.*]], <16 x i1> [[TMP1]])
 344 // CHECK-NEXT:    ret i32 [[TMP2]]
 345 //
 346 int32_t test_vmlsdavaq_p_s8(int32_t a, int8x16_t b, int8x16_t c, mve_pred16_t p) {
 347 #ifdef POLYMORPHIC
 348   return vmlsdavaq_p(a, b, c, p);
 349 #else
 350   return vmlsdavaq_p_s8(a, b, c, p);
 351 #endif
 352 }
 353
 354 // CHECK-LABEL: @test_vmlsdavaq_p_s16(
 355 // CHECK-NEXT:  entry:
 356 // CHECK-NEXT:    [[TMP0:%.*]] = zext i16 [[P:%.*]] to i32
 357 // CHECK-NEXT:    [[TMP1:%.*]] = call <8 x i1> @llvm.arm.mve.pred.i2v.v8i1(i32 [[TMP0]])
 358 // CHECK-NEXT:    [[TMP2:%.*]] = call i32 @llvm.arm.mve.vmldava.predicated.v8i16.v8i1(i32 0, i32 1, i32 0, i32 [[A:%.*]], <8 x i16> [[B:%.*]], <8 x i16> [[C:%.*]], <8 x i1> [[TMP1]])
 359 // CHECK-NEXT:    ret i32 [[TMP2]]
 360 //
 361 int32_t test_vmlsdavaq_p_s16(int32_t a, int16x8_t b, int16x8_t c, mve_pred16_t p) {
 362 #ifdef POLYMORPHIC
 363   return vmlsdavaq_p(a, b, c, p);
 364 #else
 365   return vmlsdavaq_p_s16(a, b, c, p);
 366 #endif
 367 }
 368
 369 // CHECK-LABEL: @test_vmlsdavaq_p_s32(
 370 // CHECK-NEXT:  entry:
 371 // CHECK-NEXT:    [[TMP0:%.*]] = zext i16 [[P:%.*]] to i32
 372 // CHECK-NEXT:    [[TMP1:%.*]] = call <4 x i1> @llvm.arm.mve.pred.i2v.v4i1(i32 [[TMP0]])
 373 // CHECK-NEXT:    [[TMP2:%.*]] = call i32 @llvm.arm.mve.vmldava.predicated.v4i32.v4i1(i32 0, i32 1, i32 0, i32 [[A:%.*]], <4 x i32> [[B:%.*]], <4 x i32> [[C:%.*]], <4 x i1> [[TMP1]])
 374 // CHECK-NEXT:    ret i32 [[TMP2]]
 375 //
 376 int32_t test_vmlsdavaq_p_s32(int32_t a, int32x4_t b, int32x4_t c, mve_pred16_t p) {
 377 #ifdef POLYMORPHIC
 378   return vmlsdavaq_p(a, b, c, p);
 379 #else
 380   return vmlsdavaq_p_s32(a, b, c, p);
 381 #endif
 382 }
 383
 384 // CHECK-LABEL: @test_vmlsdavaxq_p_s8(
 385 // CHECK-NEXT:  entry:
 386 // CHECK-NEXT:    [[TMP0:%.*]] = zext i16 [[P:%.*]] to i32
 387 // CHECK-NEXT:    [[TMP1:%.*]] = call <16 x i1> @llvm.arm.mve.pred.i2v.v16i1(i32 [[TMP0]])
 388 // CHECK-NEXT:    [[TMP2:%.*]] = call i32 @llvm.arm.mve.vmldava.predicated.v16i8.v16i1(i32 0, i32 1, i32 1, i32 [[A:%.*]], <16 x i8> [[B:%.*]], <16 x i8> [[C:%.*]], <16 x i1> [[TMP1]])
 389 // CHECK-NEXT:    ret i32 [[TMP2]]
 390 //
 391 int32_t test_vmlsdavaxq_p_s8(int32_t a, int8x16_t b, int8x16_t c, mve_pred16_t p) {
 392 #ifdef POLYMORPHIC
 393   return vmlsdavaxq_p(a, b, c, p);
 394 #else
 395   return vmlsdavaxq_p_s8(a, b, c, p);
 396 #endif
 397 }
 398
 399 // CHECK-LABEL: @test_vmlsdavaxq_p_s16(
 400 // CHECK-NEXT:  entry:
 401 // CHECK-NEXT:    [[TMP0:%.*]] = zext i16 [[P:%.*]] to i32
 402 // CHECK-NEXT:    [[TMP1:%.*]] = call <8 x i1> @llvm.arm.mve.pred.i2v.v8i1(i32 [[TMP0]])
 403 // CHECK-NEXT:    [[TMP2:%.*]] = call i32 @llvm.arm.mve.vmldava.predicated.v8i16.v8i1(i32 0, i32 1, i32 1, i32 [[A:%.*]], <8 x i16> [[B:%.*]], <8 x i16> [[C:%.*]], <8 x i1> [[TMP1]])
 404 // CHECK-NEXT:    ret i32 [[TMP2]]
 405 //
 406 int32_t test_vmlsdavaxq_p_s16(int32_t a, int16x8_t b, int16x8_t c, mve_pred16_t p) {
 407 #ifdef POLYMORPHIC
 408   return vmlsdavaxq_p(a, b, c, p);
 409 #else
 410   return vmlsdavaxq_p_s16(a, b, c, p);
 411 #endif
 412 }
 413
 414 // CHECK-LABEL: @test_vmlsdavaxq_p_s32(
 415 // CHECK-NEXT:  entry:
 416 // CHECK-NEXT:    [[TMP0:%.*]] = zext i16 [[P:%.*]] to i32
 417 // CHECK-NEXT:    [[TMP1:%.*]] = call <4 x i1> @llvm.arm.mve.pred.i2v.v4i1(i32 [[TMP0]])
 418 // CHECK-NEXT:    [[TMP2:%.*]] = call i32 @llvm.arm.mve.vmldava.predicated.v4i32.v4i1(i32 0, i32 1, i32 1, i32 [[A:%.*]], <4 x i32> [[B:%.*]], <4 x i32> [[C:%.*]], <4 x i1> [[TMP1]])
 419 // CHECK-NEXT:    ret i32 [[TMP2]]
 420 //
 421 int32_t test_vmlsdavaxq_p_s32(int32_t a, int32x4_t b, int32x4_t c, mve_pred16_t p) {
 422 #ifdef POLYMORPHIC
 423   return vmlsdavaxq_p(a, b, c, p);
 424 #else
 425   return vmlsdavaxq_p_s32(a, b, c, p);
 426 #endif
 427 }
 428
 429 // CHECK-LABEL: @test_vmladavq_s8(
 430 // CHECK-NEXT:  entry:
 431 // CHECK-NEXT:    [[TMP0:%.*]] = call i32 @llvm.arm.mve.vmldava.v16i8(i32 0, i32 0, i32 0, i32 0, <16 x i8> [[A:%.*]], <16 x i8> [[B:%.*]])
 432 // CHECK-NEXT:    ret i32 [[TMP0]]
 433 //
 434 int32_t test_vmladavq_s8(int8x16_t a, int8x16_t b) {
 435 #ifdef POLYMORPHIC
 436   return vmladavq(a, b);
 437 #else
 438   return vmladavq_s8(a, b);
 439 #endif
 440 }
 441
 442 // CHECK-LABEL: @test_vmladavq_s16(
 443 // CHECK-NEXT:  entry:
 444 // CHECK-NEXT:    [[TMP0:%.*]] = call i32 @llvm.arm.mve.vmldava.v8i16(i32 0, i32 0, i32 0, i32 0, <8 x i16> [[A:%.*]], <8 x i16> [[B:%.*]])
 445 // CHECK-NEXT:    ret i32 [[TMP0]]
 446 //
 447 int32_t test_vmladavq_s16(int16x8_t a, int16x8_t b) {
 448 #ifdef POLYMORPHIC
 449   return vmladavq(a, b);
 450 #else
 451   return vmladavq_s16(a, b);
 452 #endif
 453 }
 454
 455 // CHECK-LABEL: @test_vmladavq_s32(
 456 // CHECK-NEXT:  entry:
 457 // CHECK-NEXT:    [[TMP0:%.*]] = call i32 @llvm.arm.mve.vmldava.v4i32(i32 0, i32 0, i32 0, i32 0, <4 x i32> [[A:%.*]], <4 x i32> [[B:%.*]])
 458 // CHECK-NEXT:    ret i32 [[TMP0]]
 459 //
 460 int32_t test_vmladavq_s32(int32x4_t a, int32x4_t b) {
 461 #ifdef POLYMORPHIC
 462   return vmladavq(a, b);
 463 #else
 464   return vmladavq_s32(a, b);
 465 #endif
 466 }
 467
 468 // CHECK-LABEL: @test_vmladavq_u8(
 469 // CHECK-NEXT:  entry:
 470 // CHECK-NEXT:    [[TMP0:%.*]] = call i32 @llvm.arm.mve.vmldava.v16i8(i32 1, i32 0, i32 0, i32 0, <16 x i8> [[A:%.*]], <16 x i8> [[B:%.*]])
 471 // CHECK-NEXT:    ret i32 [[TMP0]]
 472 //
 473 uint32_t test_vmladavq_u8(uint8x16_t a, uint8x16_t b) {
 474 #ifdef POLYMORPHIC
 475   return vmladavq(a, b);
 476 #else
 477   return vmladavq_u8(a, b);
 478 #endif
 479 }
 480
 481 // CHECK-LABEL: @test_vmladavq_u16(
 482 // CHECK-NEXT:  entry:
 483 // CHECK-NEXT:    [[TMP0:%.*]] = call i32 @llvm.arm.mve.vmldava.v8i16(i32 1, i32 0, i32 0, i32 0, <8 x i16> [[A:%.*]], <8 x i16> [[B:%.*]])
 484 // CHECK-NEXT:    ret i32 [[TMP0]]
 485 //
 486 uint32_t test_vmladavq_u16(uint16x8_t a, uint16x8_t b) {
 487 #ifdef POLYMORPHIC
 488   return vmladavq(a, b);
 489 #else
 490   return vmladavq_u16(a, b);
 491 #endif
 492 }
 493
 494 // CHECK-LABEL: @test_vmladavq_u32(
 495 // CHECK-NEXT:  entry:
 496 // CHECK-NEXT:    [[TMP0:%.*]] = call i32 @llvm.arm.mve.vmldava.v4i32(i32 1, i32 0, i32 0, i32 0, <4 x i32> [[A:%.*]], <4 x i32> [[B:%.*]])
 497 // CHECK-NEXT:    ret i32 [[TMP0]]
 498 //
 499 uint32_t test_vmladavq_u32(uint32x4_t a, uint32x4_t b) {
 500 #ifdef POLYMORPHIC
 501   return vmladavq(a, b);
 502 #else
 503   return vmladavq_u32(a, b);
 504 #endif
 505 }
 506
 507 // CHECK-LABEL: @test_vmladavxq_s8(
 508 // CHECK-NEXT:  entry:
 509 // CHECK-NEXT:    [[TMP0:%.*]] = call i32 @llvm.arm.mve.vmldava.v16i8(i32 0, i32 0, i32 1, i32 0, <16 x i8> [[A:%.*]], <16 x i8> [[B:%.*]])
 510 // CHECK-NEXT:    ret i32 [[TMP0]]
 511 //
 512 int32_t test_vmladavxq_s8(int8x16_t a, int8x16_t b) {
 513 #ifdef POLYMORPHIC
 514   return vmladavxq(a, b);
 515 #else
 516   return vmladavxq_s8(a, b);
 517 #endif
 518 }
 519
 520 // CHECK-LABEL: @test_vmladavxq_s16(
 521 // CHECK-NEXT:  entry:
 522 // CHECK-NEXT:    [[TMP0:%.*]] = call i32 @llvm.arm.mve.vmldava.v8i16(i32 0, i32 0, i32 1, i32 0, <8 x i16> [[A:%.*]], <8 x i16> [[B:%.*]])
 523 // CHECK-NEXT:    ret i32 [[TMP0]]
 524 //
 525 int32_t test_vmladavxq_s16(int16x8_t a, int16x8_t b) {
 526 #ifdef POLYMORPHIC
 527   return vmladavxq(a, b);
 528 #else
 529   return vmladavxq_s16(a, b);
 530 #endif
 531 }
 532
 533 // CHECK-LABEL: @test_vmladavxq_s32(
 534 // CHECK-NEXT:  entry:
 535 // CHECK-NEXT:    [[TMP0:%.*]] = call i32 @llvm.arm.mve.vmldava.v4i32(i32 0, i32 0, i32 1, i32 0, <4 x i32> [[A:%.*]], <4 x i32> [[B:%.*]])
 536 // CHECK-NEXT:    ret i32 [[TMP0]]
 537 //
 538 int32_t test_vmladavxq_s32(int32x4_t a, int32x4_t b) {
 539 #ifdef POLYMORPHIC
 540   return vmladavxq(a, b);
 541 #else
 542   return vmladavxq_s32(a, b);
 543 #endif
 544 }
 545
 546 // CHECK-LABEL: @test_vmlsdavq_s8(
 547 // CHECK-NEXT:  entry:
 548 // CHECK-NEXT:    [[TMP0:%.*]] = call i32 @llvm.arm.mve.vmldava.v16i8(i32 0, i32 1, i32 0, i32 0, <16 x i8> [[A:%.*]], <16 x i8> [[B:%.*]])
 549 // CHECK-NEXT:    ret i32 [[TMP0]]
 550 //
 551 int32_t test_vmlsdavq_s8(int8x16_t a, int8x16_t b) {
 552 #ifdef POLYMORPHIC
 553   return vmlsdavq(a, b);
 554 #else
 555   return vmlsdavq_s8(a, b);
 556 #endif
 557 }
 558
 559 // CHECK-LABEL: @test_vmlsdavq_s16(
 560 // CHECK-NEXT:  entry:
 561 // CHECK-NEXT:    [[TMP0:%.*]] = call i32 @llvm.arm.mve.vmldava.v8i16(i32 0, i32 1, i32 0, i32 0, <8 x i16> [[A:%.*]], <8 x i16> [[B:%.*]])
 562 // CHECK-NEXT:    ret i32 [[TMP0]]
 563 //
 564 int32_t test_vmlsdavq_s16(int16x8_t a, int16x8_t b) {
 565 #ifdef POLYMORPHIC
 566   return vmlsdavq(a, b);
 567 #else
 568   return vmlsdavq_s16(a, b);
 569 #endif
 570 }
 571
 572 // CHECK-LABEL: @test_vmlsdavq_s32(
 573 // CHECK-NEXT:  entry:
 574 // CHECK-NEXT:    [[TMP0:%.*]] = call i32 @llvm.arm.mve.vmldava.v4i32(i32 0, i32 1, i32 0, i32 0, <4 x i32> [[A:%.*]], <4 x i32> [[B:%.*]])
 575 // CHECK-NEXT:    ret i32 [[TMP0]]
 576 //
 577 int32_t test_vmlsdavq_s32(int32x4_t a, int32x4_t b) {
 578 #ifdef POLYMORPHIC
 579   return vmlsdavq(a, b);
 580 #else
 581   return vmlsdavq_s32(a, b);
 582 #endif
 583 }
 584
 585 // CHECK-LABEL: @test_vmlsdavxq_s8(
 586 // CHECK-NEXT:  entry:
 587 // CHECK-NEXT:    [[TMP0:%.*]] = call i32 @llvm.arm.mve.vmldava.v16i8(i32 0, i32 1, i32 1, i32 0, <16 x i8> [[A:%.*]], <16 x i8> [[B:%.*]])
 588 // CHECK-NEXT:    ret i32 [[TMP0]]
 589 //
 590 int32_t test_vmlsdavxq_s8(int8x16_t a, int8x16_t b) {
 591 #ifdef POLYMORPHIC
 592   return vmlsdavxq(a, b);
 593 #else
 594   return vmlsdavxq_s8(a, b);
 595 #endif
 596 }
 597
 598 // CHECK-LABEL: @test_vmlsdavxq_s16(
 599 // CHECK-NEXT:  entry:
 600 // CHECK-NEXT:    [[TMP0:%.*]] = call i32 @llvm.arm.mve.vmldava.v8i16(i32 0, i32 1, i32 1, i32 0, <8 x i16> [[A:%.*]], <8 x i16> [[B:%.*]])
 601 // CHECK-NEXT:    ret i32 [[TMP0]]
 602 //
 603 int32_t test_vmlsdavxq_s16(int16x8_t a, int16x8_t b) {
 604 #ifdef POLYMORPHIC
 605   return vmlsdavxq(a, b);
 606 #else
 607   return vmlsdavxq_s16(a, b);
 608 #endif
 609 }
 610
 611 // CHECK-LABEL: @test_vmlsdavxq_s32(
 612 // CHECK-NEXT:  entry:
 613 // CHECK-NEXT:    [[TMP0:%.*]] = call i32 @llvm.arm.mve.vmldava.v4i32(i32 0, i32 1, i32 1, i32 0, <4 x i32> [[A:%.*]], <4 x i32> [[B:%.*]])
 614 // CHECK-NEXT:    ret i32 [[TMP0]]
 615 //
 616 int32_t test_vmlsdavxq_s32(int32x4_t a, int32x4_t b) {
 617 #ifdef POLYMORPHIC
 618   return vmlsdavxq(a, b);
 619 #else
 620   return vmlsdavxq_s32(a, b);
 621 #endif
 622 }
 623
 624 // CHECK-LABEL: @test_vmladavq_p_s8(
 625 // CHECK-NEXT:  entry:
 626 // CHECK-NEXT:    [[TMP0:%.*]] = zext i16 [[P:%.*]] to i32
 627 // CHECK-NEXT:    [[TMP1:%.*]] = call <16 x i1> @llvm.arm.mve.pred.i2v.v16i1(i32 [[TMP0]])
 628 // CHECK-NEXT:    [[TMP2:%.*]] = call i32 @llvm.arm.mve.vmldava.predicated.v16i8.v16i1(i32 0, i32 0, i32 0, i32 0, <16 x i8> [[A:%.*]], <16 x i8> [[B:%.*]], <16 x i1> [[TMP1]])
 629 // CHECK-NEXT:    ret i32 [[TMP2]]
 630 //
 631 int32_t test_vmladavq_p_s8(int8x16_t a, int8x16_t b, mve_pred16_t p) {
 632 #ifdef POLYMORPHIC
 633   return vmladavq_p(a, b, p);
 634 #else
 635   return vmladavq_p_s8(a, b, p);
 636 #endif
 637 }
 638
 639 // CHECK-LABEL: @test_vmladavq_p_s16(
 640 // CHECK-NEXT:  entry:
 641 // CHECK-NEXT:    [[TMP0:%.*]] = zext i16 [[P:%.*]] to i32
 642 // CHECK-NEXT:    [[TMP1:%.*]] = call <8 x i1> @llvm.arm.mve.pred.i2v.v8i1(i32 [[TMP0]])
 643 // CHECK-NEXT:    [[TMP2:%.*]] = call i32 @llvm.arm.mve.vmldava.predicated.v8i16.v8i1(i32 0, i32 0, i32 0, i32 0, <8 x i16> [[A:%.*]], <8 x i16> [[B:%.*]], <8 x i1> [[TMP1]])
 644 // CHECK-NEXT:    ret i32 [[TMP2]]
 645 //
 646 int32_t test_vmladavq_p_s16(int16x8_t a, int16x8_t b, mve_pred16_t p) {
 647 #ifdef POLYMORPHIC
 648   return vmladavq_p(a, b, p);
 649 #else
 650   return vmladavq_p_s16(a, b, p);
 651 #endif
 652 }
 653
 654 // CHECK-LABEL: @test_vmladavq_p_s32(
 655 // CHECK-NEXT:  entry:
 656 // CHECK-NEXT:    [[TMP0:%.*]] = zext i16 [[P:%.*]] to i32
 657 // CHECK-NEXT:    [[TMP1:%.*]] = call <4 x i1> @llvm.arm.mve.pred.i2v.v4i1(i32 [[TMP0]])
 658 // CHECK-NEXT:    [[TMP2:%.*]] = call i32 @llvm.arm.mve.vmldava.predicated.v4i32.v4i1(i32 0, i32 0, i32 0, i32 0, <4 x i32> [[A:%.*]], <4 x i32> [[B:%.*]], <4 x i1> [[TMP1]])
 659 // CHECK-NEXT:    ret i32 [[TMP2]]
 660 //
 661 int32_t test_vmladavq_p_s32(int32x4_t a, int32x4_t b, mve_pred16_t p) {
 662 #ifdef POLYMORPHIC
 663   return vmladavq_p(a, b, p);
 664 #else
 665   return vmladavq_p_s32(a, b, p);
 666 #endif
 667 }
 668
 669 // CHECK-LABEL: @test_vmladavq_p_u8(
 670 // CHECK-NEXT:  entry:
 671 // CHECK-NEXT:    [[TMP0:%.*]] = zext i16 [[P:%.*]] to i32
 672 // CHECK-NEXT:    [[TMP1:%.*]] = call <16 x i1> @llvm.arm.mve.pred.i2v.v16i1(i32 [[TMP0]])
 673 // CHECK-NEXT:    [[TMP2:%.*]] = call i32 @llvm.arm.mve.vmldava.predicated.v16i8.v16i1(i32 1, i32 0, i32 0, i32 0, <16 x i8> [[A:%.*]], <16 x i8> [[B:%.*]], <16 x i1> [[TMP1]])
 674 // CHECK-NEXT:    ret i32 [[TMP2]]
 675 //
 676 uint32_t test_vmladavq_p_u8(uint8x16_t a, uint8x16_t b, mve_pred16_t p) {
 677 #ifdef POLYMORPHIC
 678   return vmladavq_p(a, b, p);
 679 #else
 680   return vmladavq_p_u8(a, b, p);
 681 #endif
 682 }
 683
 684 // CHECK-LABEL: @test_vmladavq_p_u16(
 685 // CHECK-NEXT:  entry:
 686 // CHECK-NEXT:    [[TMP0:%.*]] = zext i16 [[P:%.*]] to i32
 687 // CHECK-NEXT:    [[TMP1:%.*]] = call <8 x i1> @llvm.arm.mve.pred.i2v.v8i1(i32 [[TMP0]])
 688 // CHECK-NEXT:    [[TMP2:%.*]] = call i32 @llvm.arm.mve.vmldava.predicated.v8i16.v8i1(i32 1, i32 0, i32 0, i32 0, <8 x i16> [[A:%.*]], <8 x i16> [[B:%.*]], <8 x i1> [[TMP1]])
 689 // CHECK-NEXT:    ret i32 [[TMP2]]
 690 //
 691 uint32_t test_vmladavq_p_u16(uint16x8_t a, uint16x8_t b, mve_pred16_t p) {
 692 #ifdef POLYMORPHIC
 693   return vmladavq_p(a, b, p);
 694 #else
 695   return vmladavq_p_u16(a, b, p);
 696 #endif
 697 }
 698
 699 // CHECK-LABEL: @test_vmladavq_p_u32(
 700 // CHECK-NEXT:  entry:
 701 // CHECK-NEXT:    [[TMP0:%.*]] = zext i16 [[P:%.*]] to i32
 702 // CHECK-NEXT:    [[TMP1:%.*]] = call <4 x i1> @llvm.arm.mve.pred.i2v.v4i1(i32 [[TMP0]])
 703 // CHECK-NEXT:    [[TMP2:%.*]] = call i32 @llvm.arm.mve.vmldava.predicated.v4i32.v4i1(i32 1, i32 0, i32 0, i32 0, <4 x i32> [[A:%.*]], <4 x i32> [[B:%.*]], <4 x i1> [[TMP1]])
 704 // CHECK-NEXT:    ret i32 [[TMP2]]
 705 //
 706 uint32_t test_vmladavq_p_u32(uint32x4_t a, uint32x4_t b, mve_pred16_t p) {
 707 #ifdef POLYMORPHIC
 708   return vmladavq_p(a, b, p);
 709 #else
 710   return vmladavq_p_u32(a, b, p);
 711 #endif
 712 }
 713
 714 // CHECK-LABEL: @test_vmladavxq_p_s8(
 715 // CHECK-NEXT:  entry:
 716 // CHECK-NEXT:    [[TMP0:%.*]] = zext i16 [[P:%.*]] to i32
 717 // CHECK-NEXT:    [[TMP1:%.*]] = call <16 x i1> @llvm.arm.mve.pred.i2v.v16i1(i32 [[TMP0]])
 718 // CHECK-NEXT:    [[TMP2:%.*]] = call i32 @llvm.arm.mve.vmldava.predicated.v16i8.v16i1(i32 0, i32 0, i32 1, i32 0, <16 x i8> [[A:%.*]], <16 x i8> [[B:%.*]], <16 x i1> [[TMP1]])
 719 // CHECK-NEXT:    ret i32 [[TMP2]]
 720 //
 721 int32_t test_vmladavxq_p_s8(int8x16_t a, int8x16_t b, mve_pred16_t p) {
 722 #ifdef POLYMORPHIC
 723   return vmladavxq_p(a, b, p);
 724 #else
 725   return vmladavxq_p_s8(a, b, p);
 726 #endif
 727 }
 728
 729 // CHECK-LABEL: @test_vmladavxq_p_s16(
 730 // CHECK-NEXT:  entry:
 731 // CHECK-NEXT:    [[TMP0:%.*]] = zext i16 [[P:%.*]] to i32
 732 // CHECK-NEXT:    [[TMP1:%.*]] = call <8 x i1> @llvm.arm.mve.pred.i2v.v8i1(i32 [[TMP0]])
 733 // CHECK-NEXT:    [[TMP2:%.*]] = call i32 @llvm.arm.mve.vmldava.predicated.v8i16.v8i1(i32 0, i32 0, i32 1, i32 0, <8 x i16> [[A:%.*]], <8 x i16> [[B:%.*]], <8 x i1> [[TMP1]])
 734 // CHECK-NEXT:    ret i32 [[TMP2]]
 735 //
 736 int32_t test_vmladavxq_p_s16(int16x8_t a, int16x8_t b, mve_pred16_t p) {
 737 #ifdef POLYMORPHIC
 738   return vmladavxq_p(a, b, p);
 739 #else
 740   return vmladavxq_p_s16(a, b, p);
 741 #endif
 742 }
 743
 744 // CHECK-LABEL: @test_vmladavxq_p_s32(
 745 // CHECK-NEXT:  entry:
 746 // CHECK-NEXT:    [[TMP0:%.*]] = zext i16 [[P:%.*]] to i32
 747 // CHECK-NEXT:    [[TMP1:%.*]] = call <4 x i1> @llvm.arm.mve.pred.i2v.v4i1(i32 [[TMP0]])
 748 // CHECK-NEXT:    [[TMP2:%.*]] = call i32 @llvm.arm.mve.vmldava.predicated.v4i32.v4i1(i32 0, i32 0, i32 1, i32 0, <4 x i32> [[A:%.*]], <4 x i32> [[B:%.*]], <4 x i1> [[TMP1]])
 749 // CHECK-NEXT:    ret i32 [[TMP2]]
 750 //
 751 int32_t test_vmladavxq_p_s32(int32x4_t a, int32x4_t b, mve_pred16_t p) {
 752 #ifdef POLYMORPHIC
 753   return vmladavxq_p(a, b, p);
 754 #else
 755   return vmladavxq_p_s32(a, b, p);
 756 #endif
 757 }
 758
 759 // CHECK-LABEL: @test_vmlsdavq_p_s8(
 760 // CHECK-NEXT:  entry:
 761 // CHECK-NEXT:    [[TMP0:%.*]] = zext i16 [[P:%.*]] to i32
 762 // CHECK-NEXT:    [[TMP1:%.*]] = call <16 x i1> @llvm.arm.mve.pred.i2v.v16i1(i32 [[TMP0]])
 763 // CHECK-NEXT:    [[TMP2:%.*]] = call i32 @llvm.arm.mve.vmldava.predicated.v16i8.v16i1(i32 0, i32 1, i32 0, i32 0, <16 x i8> [[A:%.*]], <16 x i8> [[B:%.*]], <16 x i1> [[TMP1]])
 764 // CHECK-NEXT:    ret i32 [[TMP2]]
 765 //
 766 int32_t test_vmlsdavq_p_s8(int8x16_t a, int8x16_t b, mve_pred16_t p) {
 767 #ifdef POLYMORPHIC
 768   return vmlsdavq_p(a, b, p);
 769 #else
 770   return vmlsdavq_p_s8(a, b, p);
 771 #endif
 772 }
 773
 774 // CHECK-LABEL: @test_vmlsdavq_p_s16(
 775 // CHECK-NEXT:  entry:
 776 // CHECK-NEXT:    [[TMP0:%.*]] = zext i16 [[P:%.*]] to i32
 777 // CHECK-NEXT:    [[TMP1:%.*]] = call <8 x i1> @llvm.arm.mve.pred.i2v.v8i1(i32 [[TMP0]])
 778 // CHECK-NEXT:    [[TMP2:%.*]] = call i32 @llvm.arm.mve.vmldava.predicated.v8i16.v8i1(i32 0, i32 1, i32 0, i32 0, <8 x i16> [[A:%.*]], <8 x i16> [[B:%.*]], <8 x i1> [[TMP1]])
 779 // CHECK-NEXT:    ret i32 [[TMP2]]
 780 //
 781 int32_t test_vmlsdavq_p_s16(int16x8_t a, int16x8_t b, mve_pred16_t p) {
 782 #ifdef POLYMORPHIC
 783   return vmlsdavq_p(a, b, p);
 784 #else
 785   return vmlsdavq_p_s16(a, b, p);
 786 #endif
 787 }
 788
 789 // CHECK-LABEL: @test_vmlsdavq_p_s32(
 790 // CHECK-NEXT:  entry:
 791 // CHECK-NEXT:    [[TMP0:%.*]] = zext i16 [[P:%.*]] to i32
 792 // CHECK-NEXT:    [[TMP1:%.*]] = call <4 x i1> @llvm.arm.mve.pred.i2v.v4i1(i32 [[TMP0]])
 793 // CHECK-NEXT:    [[TMP2:%.*]] = call i32 @llvm.arm.mve.vmldava.predicated.v4i32.v4i1(i32 0, i32 1, i32 0, i32 0, <4 x i32> [[A:%.*]], <4 x i32> [[B:%.*]], <4 x i1> [[TMP1]])
 794 // CHECK-NEXT:    ret i32 [[TMP2]]
 795 //
 796 int32_t test_vmlsdavq_p_s32(int32x4_t a, int32x4_t b, mve_pred16_t p) {
 797 #ifdef POLYMORPHIC
 798   return vmlsdavq_p(a, b, p);
 799 #else
 800   return vmlsdavq_p_s32(a, b, p);
 801 #endif
 802 }
 803
 804 // CHECK-LABEL: @test_vmlsdavxq_p_s8(
 805 // CHECK-NEXT:  entry:
 806 // CHECK-NEXT:    [[TMP0:%.*]] = zext i16 [[P:%.*]] to i32
 807 // CHECK-NEXT:    [[TMP1:%.*]] = call <16 x i1> @llvm.arm.mve.pred.i2v.v16i1(i32 [[TMP0]])
 808 // CHECK-NEXT:    [[TMP2:%.*]] = call i32 @llvm.arm.mve.vmldava.predicated.v16i8.v16i1(i32 0, i32 1, i32 1, i32 0, <16 x i8> [[A:%.*]], <16 x i8> [[B:%.*]], <16 x i1> [[TMP1]])
 809 // CHECK-NEXT:    ret i32 [[TMP2]]
 810 //
 811 int32_t test_vmlsdavxq_p_s8(int8x16_t a, int8x16_t b, mve_pred16_t p) {
 812 #ifdef POLYMORPHIC
 813   return vmlsdavxq_p(a, b, p);
 814 #else
 815   return vmlsdavxq_p_s8(a, b, p);
 816 #endif
 817 }
 818
 819 // CHECK-LABEL: @test_vmlsdavxq_p_s16(
 820 // CHECK-NEXT:  entry:
 821 // CHECK-NEXT:    [[TMP0:%.*]] = zext i16 [[P:%.*]] to i32
 822 // CHECK-NEXT:    [[TMP1:%.*]] = call <8 x i1> @llvm.arm.mve.pred.i2v.v8i1(i32 [[TMP0]])
 823 // CHECK-NEXT:    [[TMP2:%.*]] = call i32 @llvm.arm.mve.vmldava.predicated.v8i16.v8i1(i32 0, i32 1, i32 1, i32 0, <8 x i16> [[A:%.*]], <8 x i16> [[B:%.*]], <8 x i1> [[TMP1]])
 824 // CHECK-NEXT:    ret i32 [[TMP2]]
 825 //
 826 int32_t test_vmlsdavxq_p_s16(int16x8_t a, int16x8_t b, mve_pred16_t p) {
 827 #ifdef POLYMORPHIC
 828   return vmlsdavxq_p(a, b, p);
 829 #else
 830   return vmlsdavxq_p_s16(a, b, p);
 831 #endif
 832 }
 833
 834 // CHECK-LABEL: @test_vmlsdavxq_p_s32(
 835 // CHECK-NEXT:  entry:
 836 // CHECK-NEXT:    [[TMP0:%.*]] = zext i16 [[P:%.*]] to i32
 837 // CHECK-NEXT:    [[TMP1:%.*]] = call <4 x i1> @llvm.arm.mve.pred.i2v.v4i1(i32 [[TMP0]])
 838 // CHECK-NEXT:    [[TMP2:%.*]] = call i32 @llvm.arm.mve.vmldava.predicated.v4i32.v4i1(i32 0, i32 1, i32 1, i32 0, <4 x i32> [[A:%.*]], <4 x i32> [[B:%.*]], <4 x i1> [[TMP1]])
 839 // CHECK-NEXT:    ret i32 [[TMP2]]
 840 //
 841 int32_t test_vmlsdavxq_p_s32(int32x4_t a, int32x4_t b, mve_pred16_t p) {
 842 #ifdef POLYMORPHIC
 843   return vmlsdavxq_p(a, b, p);
 844 #else
 845   return vmlsdavxq_p_s32(a, b, p);
 846 #endif
 847 }