clang/test/Sema/arm-neon-target.c

   1 // RUN: %clang_cc1 -triple armv8a-none-linux-gnu -target-feature +neon -verify -emit-llvm -o - %s
   2 // REQUIRES: arm-registered-target
   3
   4 // Test that functions with the correct target attributes can use the correct NEON intrinsics.
   5
   6 #include <arm_neon.h>
   7
   8 __attribute__((target("dotprod")))
   9 void dotprod(uint32x2_t v2i32, uint8x16_t v16i8, uint8x8_t v8i8) {
  10   vdot_u32(v2i32, v8i8, v8i8);
  11 }
  12
  13 __attribute__((target("fullfp16")))
  14 void fp16(uint32x2_t v2i32, uint32x4_t v4i32, uint16x8_t v8i16, uint8x16_t v16i8, uint8x8_t v8i8, float32x2_t v2f32, float32x4_t v4f32, float16x4_t v4f16, bfloat16x4_t v4bf16) {
  15   vceqz_f16(v4f16);
  16   vrnd_f16(v4f16);
  17   vmaxnm_f16(v4f16, v4f16);
  18 }
  19
  20 __attribute__((target("i8mm")))
  21 void i8mm(uint32x2_t v2i32, uint32x4_t v4i32, uint16x8_t v8i16, uint8x16_t v16i8, uint8x8_t v8i8, float32x2_t v2f32, float32x4_t v4f32, float16x4_t v4f16, bfloat16x4_t v4bf16) {
  22   vmmlaq_s32(v4i32, v8i16, v8i16);
  23 }
  24
  25 __attribute__((target("bf16")))
  26 void bf16(uint32x2_t v2i32, uint32x4_t v4i32, uint16x8_t v8i16, uint8x16_t v16i8, uint8x8_t v8i8, float32x2_t v2f32, float32x4_t v4f32, float16x4_t v4f16, bfloat16x4_t v4bf16, __bf16 bf16) {
  27   vbfdot_f32(v2f32, v4bf16, v4bf16);
  28   vcreate_bf16(10);
  29   vdup_lane_bf16(v4bf16, 2);
  30   vdup_n_bf16(bf16);
  31   vld1_bf16(0);
  32   vcvt_f32_bf16(v4bf16);
  33   vcvt_bf16_f32(v4f32);
  34 }
  35
  36 __attribute__((target("v8.1a")))
  37 void test_v81(int32x2_t d, int32x4_t v, int s) {
  38   vqrdmlahq_s32(v, v, v);
  39 }
  40
  41 __attribute__((target("v8.3a,fullfp16")))
  42 void test_v83(float32x4_t v4f32, float16x4_t v4f16) {
  43   vcaddq_rot90_f32(v4f32, v4f32);
  44   vcmla_rot90_f16(v4f16, v4f16, v4f16);
  45 }
  46
  47 void undefined(uint32x2_t v2i32, uint32x4_t v4i32, uint16x8_t v8i16, uint8x16_t v16i8, uint8x8_t v8i8, float32x2_t v2f32, float32x4_t v4f32, float16x4_t v4f16, bfloat16x4_t v4bf16, __bf16 bf16) {
  48   // dotprod
  49   vdot_u32(v2i32, v8i8, v8i8); // expected-error {{always_inline function 'vdot_u32' requires target feature 'dotprod'}}
  50   // fp16
  51   vceqz_f16(v4f16); // expected-error {{always_inline function 'vceqz_f16' requires target feature 'fullfp16'}}
  52   vrnd_f16(v4f16); // expected-error {{always_inline function 'vrnd_f16' requires target feature 'fullfp16'}}
  53   vmaxnm_f16(v4f16, v4f16); // expected-error {{always_inline function 'vmaxnm_f16' requires target feature 'fullfp16'}}
  54   // i8mm
  55   vmmlaq_s32(v4i32, v8i16, v8i16); // expected-error {{always_inline function 'vmmlaq_s32' requires target feature 'i8mm'}}
  56   // bf16
  57   vbfdot_f32(v2f32, v4bf16, v4bf16); // expected-error {{always_inline function 'vbfdot_f32' requires target feature 'bf16'}}
  58   vcreate_bf16(10);
  59   vdup_lane_bf16(v4bf16, 2); // expected-error {{'__builtin_neon_splat_lane_bf16' needs target feature bf16}}
  60   vdup_n_bf16(bf16); // expected-error {{always_inline function 'vdup_n_bf16' requires target feature 'bf16'}}
  61   vld1_bf16(0); // expected-error {{'__builtin_neon_vld1_bf16' needs target feature bf16}}
  62   vcvt_f32_bf16(v4bf16); // expected-error {{always_inline function 'vcvt_f32_bf16' requires target feature 'bf16'}}
  63   vcvt_bf16_f32(v4f32); // expected-error {{always_inline function 'vcvt_bf16_f32' requires target feature 'bf16'}}
  64   // v8.1 - qrdmla
  65   vqrdmlahq_s32(v4i32, v4i32, v4i32); // expected-error {{always_inline function 'vqrdmlahq_s32' requires target feature 'v8.1a'}}
  66   // 8.3 - complex
  67   vcaddq_rot90_f32(v4f32, v4f32); // expected-error {{always_inline function 'vcaddq_rot90_f32' requires target feature 'v8.3a'}}
  68   vcmla_rot90_f16(v4f16, v4f16, v4f16); // expected-error {{always_inline function 'vcmla_rot90_f16' requires target feature 'v8.3a'}}
  69 }