Home | History | Annotate | Download | only in CodeGenOpenCL
      1 // REQUIRES: r600-registered-target
      2 // RUN: %clang_cc1 -triple r600-unknown-unknown -S -emit-llvm -o - %s | FileCheck %s
      3 // RUN: %clang_cc1 -triple amdgcn-unknown-unknown -S -emit-llvm -o - %s | FileCheck %s
      4 
      5 #pragma OPENCL EXTENSION cl_khr_fp64 : enable
      6 
      7 // CHECK-LABEL: @test_div_scale_f64
      8 // CHECK: call { double, i1 } @llvm.AMDGPU.div.scale.f64(double %a, double %b, i1 true)
      9 // CHECK-DAG: [[FLAG:%.+]] = extractvalue { double, i1 } %{{.+}}, 1
     10 // CHECK-DAG: [[VAL:%.+]] = extractvalue { double, i1 } %{{.+}}, 0
     11 // CHECK: [[FLAGEXT:%.+]] = zext i1 [[FLAG]] to i32
     12 // CHECK: store i32 [[FLAGEXT]]
     13 void test_div_scale_f64(global double* out, global int* flagout, double a, double b)
     14 {
     15   bool flag;
     16   *out = __builtin_amdgpu_div_scale(a, b, true, &flag);
     17   *flagout = flag;
     18 }
     19 
     20 // CHECK-LABEL: @test_div_scale_f32
     21 // CHECK: call { float, i1 } @llvm.AMDGPU.div.scale.f32(float %a, float %b, i1 true)
     22 // CHECK-DAG: [[FLAG:%.+]] = extractvalue { float, i1 } %{{.+}}, 1
     23 // CHECK-DAG: [[VAL:%.+]] = extractvalue { float, i1 } %{{.+}}, 0
     24 // CHECK: [[FLAGEXT:%.+]] = zext i1 [[FLAG]] to i32
     25 // CHECK: store i32 [[FLAGEXT]]
     26 void test_div_scale_f32(global float* out, global int* flagout, float a, float b)
     27 {
     28   bool flag;
     29   *out = __builtin_amdgpu_div_scalef(a, b, true, &flag);
     30   *flagout = flag;
     31 }
     32 
     33 // CHECK-LABEL: @test_div_fmas_f32
     34 // CHECK: call float @llvm.AMDGPU.div.fmas.f32
     35 void test_div_fmas_f32(global float* out, float a, float b, float c, int d)
     36 {
     37   *out = __builtin_amdgpu_div_fmasf(a, b, c, d);
     38 }
     39 
     40 // CHECK-LABEL: @test_div_fmas_f64
     41 // CHECK: call double @llvm.AMDGPU.div.fmas.f64
     42 void test_div_fmas_f64(global double* out, double a, double b, double c, int d)
     43 {
     44   *out = __builtin_amdgpu_div_fmas(a, b, c, d);
     45 }
     46 
     47 // CHECK-LABEL: @test_div_fixup_f32
     48 // CHECK: call float @llvm.AMDGPU.div.fixup.f32
     49 void test_div_fixup_f32(global float* out, float a, float b, float c)
     50 {
     51   *out = __builtin_amdgpu_div_fixupf(a, b, c);
     52 }
     53 
     54 // CHECK-LABEL: @test_div_fixup_f64
     55 // CHECK: call double @llvm.AMDGPU.div.fixup.f64
     56 void test_div_fixup_f64(global double* out, double a, double b, double c)
     57 {
     58   *out = __builtin_amdgpu_div_fixup(a, b, c);
     59 }
     60 
     61 // CHECK-LABEL: @test_trig_preop_f32
     62 // CHECK: call float @llvm.AMDGPU.trig.preop.f32
     63 void test_trig_preop_f32(global float* out, float a, int b)
     64 {
     65   *out = __builtin_amdgpu_trig_preopf(a, b);
     66 }
     67 
     68 // CHECK-LABEL: @test_trig_preop_f64
     69 // CHECK: call double @llvm.AMDGPU.trig.preop.f64
     70 void test_trig_preop_f64(global double* out, double a, int b)
     71 {
     72   *out = __builtin_amdgpu_trig_preop(a, b);
     73 }
     74 
     75 // CHECK-LABEL: @test_rcp_f32
     76 // CHECK: call float @llvm.AMDGPU.rcp.f32
     77 void test_rcp_f32(global float* out, float a)
     78 {
     79   *out = __builtin_amdgpu_rcpf(a);
     80 }
     81 
     82 // CHECK-LABEL: @test_rcp_f64
     83 // CHECK: call double @llvm.AMDGPU.rcp.f64
     84 void test_rcp_f64(global double* out, double a)
     85 {
     86   *out = __builtin_amdgpu_rcp(a);
     87 }
     88 
     89 // CHECK-LABEL: @test_rsq_f32
     90 // CHECK: call float @llvm.AMDGPU.rsq.f32
     91 void test_rsq_f32(global float* out, float a)
     92 {
     93   *out = __builtin_amdgpu_rsqf(a);
     94 }
     95 
     96 // CHECK-LABEL: @test_rsq_f64
     97 // CHECK: call double @llvm.AMDGPU.rsq.f64
     98 void test_rsq_f64(global double* out, double a)
     99 {
    100   *out = __builtin_amdgpu_rsq(a);
    101 }
    102 
    103 // CHECK-LABEL: @test_rsq_clamped_f32
    104 // CHECK: call float @llvm.AMDGPU.rsq.clamped.f32
    105 void test_rsq_clamped_f32(global float* out, float a)
    106 {
    107   *out = __builtin_amdgpu_rsq_clampedf(a);
    108 }
    109 
    110 // CHECK-LABEL: @test_rsq_clamped_f64
    111 // CHECK: call double @llvm.AMDGPU.rsq.clamped.f64
    112 void test_rsq_clamped_f64(global double* out, double a)
    113 {
    114   *out = __builtin_amdgpu_rsq_clamped(a);
    115 }
    116 
    117 // CHECK-LABEL: @test_ldexp_f32
    118 // CHECK: call float @llvm.AMDGPU.ldexp.f32
    119 void test_ldexp_f32(global float* out, float a, int b)
    120 {
    121   *out = __builtin_amdgpu_ldexpf(a, b);
    122 }
    123 
    124 // CHECK-LABEL: @test_ldexp_f64
    125 // CHECK: call double @llvm.AMDGPU.ldexp.f64
    126 void test_ldexp_f64(global double* out, double a, int b)
    127 {
    128   *out = __builtin_amdgpu_ldexp(a, b);
    129 }
    130 
    131 // CHECK-LABEL: @test_class_f32
    132 // CHECK: call i1 @llvm.AMDGPU.class.f32
    133 void test_class_f32(global float* out, float a, int b)
    134 {
    135   *out = __builtin_amdgpu_classf(a, b);
    136 }
    137 
    138 // CHECK-LABEL: @test_class_f64
    139 // CHECK: call i1 @llvm.AMDGPU.class.f64
    140 void test_class_f64(global double* out, double a, int b)
    141 {
    142   *out = __builtin_amdgpu_class(a, b);
    143 }
    144