brintos

brintos / llvm-project-archived public Read only

0
0
Text · 2.0 KiB · d1fe78d Raw
52 lines · plain
1; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py2; RUN: llc -mtriple=amdgcn -mcpu=verde < %s | FileCheck -check-prefix=SI %s3; RUN: llc -mtriple=amdgcn -mcpu=tonga < %s | FileCheck -check-prefix=VI %s4 5define amdgpu_ps float @main(float %arg0, float %arg1) #0 {6; SI-LABEL: main:7; SI:       ; %bb.0: ; %bb8; SI-NEXT:    image_load v2, v0, s[0:7] dmask:0x1 unorm9; SI-NEXT:    v_cvt_i32_f32_e32 v0, v010; SI-NEXT:    v_and_b32_e32 v0, 7, v011; SI-NEXT:    s_waitcnt vmcnt(0)12; SI-NEXT:    v_lshrrev_b32_e32 v0, v0, v213; SI-NEXT:    v_and_b32_e32 v0, 1, v014; SI-NEXT:    v_cmp_eq_u32_e32 vcc, 1, v015; SI-NEXT:    v_cndmask_b32_e32 v0, 0, v1, vcc16; SI-NEXT:    v_cvt_pkrtz_f16_f32_e32 v0, s0, v017; SI-NEXT:    ; return to shader part epilog18;19; VI-LABEL: main:20; VI:       ; %bb.0: ; %bb21; VI-NEXT:    image_load v2, v0, s[0:7] dmask:0x1 unorm22; VI-NEXT:    v_cvt_i32_f32_e32 v0, v023; VI-NEXT:    v_and_b32_e32 v0, 7, v024; VI-NEXT:    s_waitcnt vmcnt(0)25; VI-NEXT:    v_lshrrev_b32_e32 v0, v0, v226; VI-NEXT:    v_and_b32_e32 v0, 1, v027; VI-NEXT:    v_cmp_eq_u32_e32 vcc, 1, v028; VI-NEXT:    v_cndmask_b32_e32 v0, 0, v1, vcc29; VI-NEXT:    v_cvt_pkrtz_f16_f32 v0, s0, v030; VI-NEXT:    ; return to shader part epilog31bb:32  %tmp = fptosi float %arg0 to i3233  %tmp1 = call <4 x float> @llvm.amdgcn.image.load.1d.v4f32.i32(i32 15, i32 poison, <8 x i32> poison, i32 0, i32 0)34  %tmp2.f = extractelement <4 x float> %tmp1, i32 035  %tmp2 = bitcast float %tmp2.f to i3236  %tmp3 = and i32 %tmp, 737  %tmp4 = shl i32 1, %tmp338  %tmp5 = and i32 %tmp2, %tmp439  %tmp6 = icmp eq i32 %tmp5, 040  %tmp7 = select i1 %tmp6, float 0.000000e+00, float %arg141  %tmp8 = call <2 x half> @llvm.amdgcn.cvt.pkrtz(float poison, float %tmp7)42  %tmp9 = bitcast <2 x half> %tmp8 to float43  ret float %tmp944}45 46declare <2 x half> @llvm.amdgcn.cvt.pkrtz(float, float) #147declare <4 x float> @llvm.amdgcn.image.load.1d.v4f32.i32(i32, i32, <8 x i32>, i32, i32) #248 49attributes #0 = { nounwind }50attributes #1 = { nounwind readnone }51attributes #2 = { nounwind readonly }52