xref: /llvm-project/llvm/test/CodeGen/AMDGPU/input-mods.r600.ll (revision 9e9907f1cfa424366fba58d9520f9305b537cec9)
1; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 2
2; RUN: llc -mtriple=r600 -mcpu=redwood < %s | FileCheck --check-prefix=EG %s
3; RUN: llc -mtriple=r600 -mcpu=cayman < %s | FileCheck --check-prefix=CM %s
4
5define amdgpu_ps void @test(<4 x float> inreg %reg0) {
6; EG-LABEL: test:
7; EG:       ; %bb.0:
8; EG-NEXT:    ALU 8, @4, KC0[], KC1[]
9; EG-NEXT:    EXPORT T0.X___
10; EG-NEXT:    CF_END
11; EG-NEXT:    PAD
12; EG-NEXT:    ALU clause starting at 4:
13; EG-NEXT:     SETGT * T0.W, literal.x, -|T0.X|,
14; EG-NEXT:    -1023672320(-1.260000e+02), 0(0.000000e+00)
15; EG-NEXT:     CNDE * T1.W, PV.W, 0.0, literal.x,
16; EG-NEXT:    1115684864(6.400000e+01), 0(0.000000e+00)
17; EG-NEXT:     ADD T1.W, -|T0.X|, PV.W,
18; EG-NEXT:     CNDE * T0.W, T0.W, 1.0, literal.x,
19; EG-NEXT:    528482304(5.421011e-20), 0(0.000000e+00)
20; EG-NEXT:     EXP_IEEE * T0.X, PV.W,
21; EG-NEXT:     MUL_IEEE * T0.X, PS, T0.W,
22;
23; CM-LABEL: test:
24; CM:       ; %bb.0:
25; CM-NEXT:    ALU 11, @4, KC0[], KC1[]
26; CM-NEXT:    EXPORT T0.X___
27; CM-NEXT:    CF_END
28; CM-NEXT:    PAD
29; CM-NEXT:    ALU clause starting at 4:
30; CM-NEXT:     SETGT * T0.W, literal.x, -|T0.X|,
31; CM-NEXT:    -1023672320(-1.260000e+02), 0(0.000000e+00)
32; CM-NEXT:     CNDE * T1.W, PV.W, 0.0, literal.x,
33; CM-NEXT:    1115684864(6.400000e+01), 0(0.000000e+00)
34; CM-NEXT:     CNDE T0.Z, T0.W, 1.0, literal.x,
35; CM-NEXT:     ADD * T0.W, -|T0.X|, PV.W,
36; CM-NEXT:    528482304(5.421011e-20), 0(0.000000e+00)
37; CM-NEXT:     EXP_IEEE T0.X, T0.W,
38; CM-NEXT:     EXP_IEEE T0.Y (MASKED), T0.W,
39; CM-NEXT:     EXP_IEEE T0.Z (MASKED), T0.W,
40; CM-NEXT:     EXP_IEEE * T0.W (MASKED), T0.W,
41; CM-NEXT:     MUL_IEEE * T0.X, PV.X, T0.Z,
42   %r0 = extractelement <4 x float> %reg0, i32 0
43   %r1 = call float @llvm.fabs.f32(float %r0)
44   %r2 = fsub float -0.000000e+00, %r1
45   %r3 = call afn float @llvm.exp2.f32(float %r2)
46   %vec = insertelement <4 x float> undef, float %r3, i32 0
47   call void @llvm.r600.store.swizzle(<4 x float> %vec, i32 0, i32 0)
48   ret void
49}
50
51declare float @llvm.exp2.f32(float) readnone
52declare float @llvm.fabs.f32(float) readnone
53declare void @llvm.r600.store.swizzle(<4 x float>, i32, i32)
54