clang/test/CodeGen/aarch64-neon-misc-constrained.c

   1 // RUN: %clang_cc1 -triple arm64-none-linux-gnu -target-feature +neon \
   2 // RUN:  -disable-O0-optnone -emit-llvm -o - %s \
   3 // RUN: | opt -S -passes=mem2reg | FileCheck --check-prefix=COMMON --check-prefix=COMMONIR --check-prefix=UNCONSTRAINED %s
   4 // RUN: %clang_cc1 -triple arm64-none-linux-gnu -target-feature +neon \
   5 // RUN:  -ffp-exception-behavior=strict \
   6 // RUN:  -disable-O0-optnone -emit-llvm -o - %s \
   7 // RUN: | opt -S -passes=mem2reg | FileCheck --check-prefix=COMMON --check-prefix=COMMONIR --check-prefix=CONSTRAINED %s
   8 // RUN: %clang_cc1 -triple arm64-none-linux-gnu -target-feature +neon \
   9 // RUN:  -disable-O0-optnone -S -o - %s \
  10 // RUN: | FileCheck --check-prefix=COMMON --check-prefix=CHECK-ASM %s
  11 // RUN: %clang_cc1 -triple arm64-none-linux-gnu -target-feature +neon \
  12 // RUN:  -ffp-exception-behavior=strict \
  13 // RUN:  -disable-O0-optnone -S -o - %s \
  14 // RUN: | FileCheck --check-prefix=COMMON --check-prefix=CHECK-ASM %s
  15
  16 // REQUIRES: aarch64-registered-target
  17
  18 // Test new aarch64 intrinsics and types but constrained
  19
  20 #include <arm_neon.h>
  21
  22 // COMMON-LABEL: test_vrndaq_f64
  23 // COMMONIR:   [[TMP0:%.*]] = bitcast <2 x double> %a to <16 x i8>
  24 // UNCONSTRAINED: [[VRNDA1_I:%.*]] = call <2 x double> @llvm.round.v2f64(<2 x double> %a)
  25 // CONSTRAINED:   [[VRNDA1_I:%.*]] = call <2 x double> @llvm.experimental.constrained.round.v2f64(<2 x double> %a, metadata !"fpexcept.strict")
  26 // CHECK-ASM:     frinta v{{[0-9]+}}.2d, v{{[0-9]+}}.2d
  27 // COMMONIR:      ret <2 x double> [[VRNDA1_I]]
  28 float64x2_t test_vrndaq_f64(float64x2_t a) {
  29   return vrndaq_f64(a);
  30 }
  31
  32 // COMMON-LABEL: test_vrndpq_f64
  33 // COMMONIR:   [[TMP0:%.*]] = bitcast <2 x double> %a to <16 x i8>
  34 // UNCONSTRAINED: [[VRNDP1_I:%.*]] = call <2 x double> @llvm.ceil.v2f64(<2 x double> %a)
  35 // CONSTRAINED:   [[VRNDP1_I:%.*]] = call <2 x double> @llvm.experimental.constrained.ceil.v2f64(<2 x double> %a, metadata !"fpexcept.strict")
  36 // CHECK-ASM:     frintp v{{[0-9]+}}.2d, v{{[0-9]+}}.2d
  37 // COMMONIR:      ret <2 x double> [[VRNDP1_I]]
  38 float64x2_t test_vrndpq_f64(float64x2_t a) {
  39   return vrndpq_f64(a);
  40 }
  41
  42 // COMMON-LABEL: test_vsqrtq_f32
  43 // COMMONIR:   [[TMP0:%.*]] = bitcast <4 x float> %a to <16 x i8>
  44 // UNCONSTRAINED: [[VSQRT_I:%.*]] = call <4 x float> @llvm.sqrt.v4f32(<4 x float> %a)
  45 // CONSTRAINED:   [[VSQRT_I:%.*]] = call <4 x float> @llvm.experimental.constrained.sqrt.v4f32(<4 x float> %a, metadata !"round.tonearest", metadata !"fpexcept.strict")
  46 // CHECK-ASM:     fsqrt v{{[0-9]+}}.4s, v{{[0-9]+}}.4s
  47 // COMMONIR:      ret <4 x float> [[VSQRT_I]]
  48 float32x4_t test_vsqrtq_f32(float32x4_t a) {
  49   return vsqrtq_f32(a);
  50 }
  51
  52 // COMMON-LABEL: test_vsqrtq_f64
  53 // COMMONIR:   [[TMP0:%.*]] = bitcast <2 x double> %a to <16 x i8>
  54 // UNCONSTRAINED: [[VSQRT_I:%.*]] = call <2 x double> @llvm.sqrt.v2f64(<2 x double> %a)
  55 // CONSTRAINED:   [[VSQRT_I:%.*]] = call <2 x double> @llvm.experimental.constrained.sqrt.v2f64(<2 x double> %a, metadata !"round.tonearest", metadata !"fpexcept.strict")
  56 // CHECK-ASM:     fsqrt v{{[0-9]+}}.2d, v{{[0-9]+}}.2d
  57 // COMMONIR:      ret <2 x double> [[VSQRT_I]]
  58 float64x2_t test_vsqrtq_f64(float64x2_t a) {
  59   return vsqrtq_f64(a);
  60 }