1; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py 2; RUN: llc -march=amdgcn -mcpu=gfx1010 -amdgpu-atomic-optimizations < %s | FileCheck %s -check-prefixes=GFX10 3 4declare void @llvm.amdgcn.exp.f32(i32 immarg, i32 immarg, float, float, float, float, i1 immarg, i1 immarg) 5declare i32 @llvm.amdgcn.raw.buffer.atomic.and.i32(i32, <4 x i32>, i32, i32, i32 immarg) 6declare float @llvm.amdgcn.strict.wqm.f32(float) 7 8define amdgpu_ps void @main(i32 %arg) { 9; GFX10-LABEL: main: 10; GFX10: ; %bb.0: ; %bb 11; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v0 12; GFX10-NEXT: v_mov_b32_e32 v0, 0 13; GFX10-NEXT: s_mov_b32 s1, exec_lo 14; GFX10-NEXT: s_mov_b32 s4, 0 15; GFX10-NEXT: s_mov_b32 s2, 0 16; GFX10-NEXT: s_branch .LBB0_2 17; GFX10-NEXT: .LBB0_1: ; %Flow 18; GFX10-NEXT: ; in Loop: Header=BB0_2 Depth=1 19; GFX10-NEXT: s_waitcnt_depctr 0xffe3 20; GFX10-NEXT: s_or_b32 exec_lo, exec_lo, s3 21; GFX10-NEXT: s_andn2_b32 exec_lo, exec_lo, s2 22; GFX10-NEXT: s_cbranch_execz .LBB0_5 23; GFX10-NEXT: .LBB0_2: ; %bb4 24; GFX10-NEXT: ; =>This Inner Loop Header: Depth=1 25; GFX10-NEXT: s_and_b32 s0, exec_lo, vcc_lo 26; GFX10-NEXT: s_or_b32 s2, s0, s2 27; GFX10-NEXT: s_and_saveexec_b32 s3, s1 28; GFX10-NEXT: s_cbranch_execz .LBB0_1 29; GFX10-NEXT: ; %bb.3: ; in Loop: Header=BB0_2 Depth=1 30; GFX10-NEXT: v_mbcnt_lo_u32_b32 v1, exec_lo, 0 31; GFX10-NEXT: v_cmp_eq_u32_e64 s0, 0, v1 32; GFX10-NEXT: s_and_b32 exec_lo, exec_lo, s0 33; GFX10-NEXT: s_cbranch_execz .LBB0_1 34; GFX10-NEXT: ; %bb.4: ; in Loop: Header=BB0_2 Depth=1 35; GFX10-NEXT: s_mov_b32 s5, s4 36; GFX10-NEXT: s_mov_b32 s6, s4 37; GFX10-NEXT: s_mov_b32 s7, s4 38; GFX10-NEXT: buffer_atomic_and v0, off, s[4:7], 0 39; GFX10-NEXT: s_branch .LBB0_1 40; GFX10-NEXT: .LBB0_5: ; %bb8 41; GFX10-NEXT: s_or_b32 exec_lo, exec_lo, s2 42; GFX10-NEXT: s_mov_b32 s0, 0 43; GFX10-NEXT: v_mov_b32_e32 v1, 0 44; GFX10-NEXT: v_mov_b32_e32 v0, s0 45; GFX10-NEXT: exp mrt0 off, off, off, off 46; GFX10-NEXT: s_endpgm 47bb: 48 br label %bb4 49 50bb4: 51 %i5 = call i32 @llvm.amdgcn.raw.buffer.atomic.and.i32(i32 0, <4 x i32> zeroinitializer, i32 0, i32 0, i32 0) 52 %i7 = icmp eq i32 %arg, 1 53 br i1 %i7, label %bb8, label %bb4 54 55bb8: 56 %i9 = call float @llvm.amdgcn.strict.wqm.f32(float 0.0) 57 call void @llvm.amdgcn.exp.f32(i32 0, i32 0, float %i9, float 0.0, float 0.0, float 0.0, i1 false, i1 false) 58 ret void 59} 60