clang/test/CodeGen/AArch64/neon-vcadd.c

   1 // RUN: %clang_cc1 -triple aarch64-linux-gnu -target-feature +neon \
   2 // RUN:  -target-feature +v8.3a -target-feature +fullfp16 -emit-llvm -o - %s \
   3 // RUN:  | FileCheck %s
   4
   5 // REQUIRES: aarch64-registered-target || arm-registered-target
   6
   7 #include <arm_neon.h>
   8
   9 void foo16x4_rot90(float16x4_t a, float16x4_t b)
  10 {
  11 // CHECK: call <4 x half> @llvm.aarch64.neon.vcadd.rot90.v4f16
  12   float16x4_t result = vcadd_rot90_f16(a, b);
  13 }
  14
  15 void foo32x2_rot90(float32x2_t a, float32x2_t b)
  16 {
  17 // CHECK: call <2 x float> @llvm.aarch64.neon.vcadd.rot90.v2f32
  18   float32x2_t result = vcadd_rot90_f32(a, b);
  19 }
  20
  21 void foo16x8_rot90(float16x8_t a, float16x8_t b)
  22 {
  23 // CHECK: call <8 x half> @llvm.aarch64.neon.vcadd.rot90.v8f16
  24   float16x8_t result = vcaddq_rot90_f16(a, b);
  25 }
  26
  27 void foo32x4_rot90(float32x4_t a, float32x4_t b)
  28 {
  29 // CHECK: call <4 x float> @llvm.aarch64.neon.vcadd.rot90.v4f32
  30   float32x4_t result = vcaddq_rot90_f32(a, b);
  31 }
  32
  33 void foo64x2_rot90(float64x2_t a, float64x2_t b)
  34 {
  35 // CHECK: call <2 x double> @llvm.aarch64.neon.vcadd.rot90.v2f64
  36   float64x2_t result = vcaddq_rot90_f64(a, b);
  37 }
  38
  39 void foo16x4_rot270(float16x4_t a, float16x4_t b)
  40 {
  41 // CHECK: call <4 x half> @llvm.aarch64.neon.vcadd.rot270.v4f16
  42   float16x4_t result = vcadd_rot270_f16(a, b);
  43 }
  44
  45 void foo32x2_rot270(float32x2_t a, float32x2_t b)
  46 {
  47 // CHECK: call <2 x float> @llvm.aarch64.neon.vcadd.rot270.v2f32
  48   float32x2_t result = vcadd_rot270_f32(a, b);
  49 }
  50
  51 void foo16x8_rot270(float16x8_t a, float16x8_t b)
  52 {
  53 // CHECK: call <8 x half> @llvm.aarch64.neon.vcadd.rot270.v8f16
  54   float16x8_t result = vcaddq_rot270_f16(a, b);
  55 }
  56
  57 void foo32x4_rot270(float32x4_t a, float32x4_t b)
  58 {
  59 // CHECK: call <4 x float> @llvm.aarch64.neon.vcadd.rot270.v4f32
  60   float32x4_t result = vcaddq_rot270_f32(a, b);
  61 }
  62
  63 void foo64x2_rot270(float64x2_t a, float64x2_t b)
  64 {
  65 // CHECK: call <2 x double> @llvm.aarch64.neon.vcadd.rot270.v2f64
  66   float64x2_t result = vcaddq_rot270_f64(a, b);
  67 }