clang/test/CodeGen/PowerPC/builtins-ppc-fastmath.c

   1 // REQUIRES: powerpc-registered-target
   2 // RUN: %clang_cc1 -triple powerpc64-unknown-unknown \
   3 // RUN:   -emit-llvm %s -o - -target-cpu pwr7 | FileCheck %s
   4 // RUN: %clang_cc1 -triple powerpc64le-unknown-unknown \
   5 // RUN:   -emit-llvm %s -o - -target-cpu pwr8 | FileCheck %s
   6 // RUN: %clang_cc1 -triple powerpc64-unknown-aix \
   7 // RUN:   -emit-llvm %s -o - -target-cpu pwr7 | FileCheck %s
   8 // RUN: %clang_cc1 -triple powerpc-unknown-aix \
   9 // RUN:   -emit-llvm %s -o - -target-cpu pwr7 | FileCheck %s
  10
  11 extern vector float a;
  12 extern vector float b;
  13 extern vector float c;
  14 extern vector double d;
  15 extern vector double e;
  16 extern vector double f;
  17
  18 // CHECK-LABEL: @test_flags_recipdivf(
  19 // CHECK:    [[TMP0:%.*]] = load <4 x float>, ptr @a, align 16
  20 // CHECK-NEXT:    [[TMP1:%.*]] = load <4 x float>, ptr @b, align 16
  21 // CHECK-NEXT:    [[RECIPDIV:%.*]] = fdiv fast <4 x float> [[TMP0]], [[TMP1]]
  22 // CHECK-NEXT:    [[TMP2:%.*]] = load <4 x float>, ptr @c, align 16
  23 // CHECK-NEXT:    [[ADD:%.*]] = fadd <4 x float> [[RECIPDIV]], [[TMP2]]
  24 // CHECK-NEXT:    ret <4 x float> [[ADD]]
  25 //
  26 vector float test_flags_recipdivf() {
  27   return __builtin_ppc_recipdivf(a, b) + c;
  28 }
  29
  30 // CHECK-LABEL: @test_flags_recipdivd(
  31 // CHECK:    [[TMP0:%.*]] = load <2 x double>, ptr @d, align 16
  32 // CHECK-NEXT:    [[TMP1:%.*]] = load <2 x double>, ptr @e, align 16
  33 // CHECK-NEXT:    [[RECIPDIV:%.*]] = fdiv fast <2 x double> [[TMP0]], [[TMP1]]
  34 // CHECK-NEXT:    [[TMP2:%.*]] = load <2 x double>, ptr @f, align 16
  35 // CHECK-NEXT:    [[ADD:%.*]] = fadd <2 x double> [[RECIPDIV]], [[TMP2]]
  36 // CHECK-NEXT:    ret <2 x double> [[ADD]]
  37 //
  38 vector double test_flags_recipdivd() {
  39   return __builtin_ppc_recipdivd(d, e) + f;
  40 }
  41
  42 // CHECK-LABEL: @test_flags_rsqrtf(
  43 // CHECK:    [[TMP0:%.*]] = load <4 x float>, ptr @a, align 16
  44 // CHECK-NEXT:    [[TMP1:%.*]] = call fast <4 x float> @llvm.sqrt.v4f32(<4 x float> [[TMP0]])
  45 // CHECK-NEXT:    [[RSQRT:%.*]] = fdiv fast <4 x float> <float 1.000000e+00, float 1.000000e+00, float 1.000000e+00, float 1.000000e+00>, [[TMP1]]
  46 // CHECK-NEXT:    [[TMP2:%.*]] = load <4 x float>, ptr @b, align 16
  47 // CHECK-NEXT:    [[ADD:%.*]] = fadd <4 x float> [[RSQRT]], [[TMP2]]
  48 // CHECK-NEXT:    ret <4 x float> [[ADD]]
  49 //
  50 vector float test_flags_rsqrtf() {
  51   return __builtin_ppc_rsqrtf(a) + b;
  52 }
  53
  54 // CHECK-LABEL: @test_flags_rsqrtd(
  55 // CHECK:    [[TMP0:%.*]] = load <2 x double>, ptr @d, align 16
  56 // CHECK-NEXT:    [[TMP1:%.*]] = call fast <2 x double> @llvm.sqrt.v2f64(<2 x double> [[TMP0]])
  57 // CHECK-NEXT:    [[RSQRT:%.*]] = fdiv fast <2 x double> <double 1.000000e+00, double 1.000000e+00>, [[TMP1]]
  58 // CHECK-NEXT:    [[TMP2:%.*]] = load <2 x double>, ptr @e, align 16
  59 // CHECK-NEXT:    [[ADD:%.*]] = fadd <2 x double> [[RSQRT]], [[TMP2]]
  60 // CHECK-NEXT:    ret <2 x double> [[ADD]]
  61 //
  62 vector double test_flags_rsqrtd() {
  63   return __builtin_ppc_rsqrtd(d) + e;
  64 }