diff --git a/llvm/lib/Target/AMDGPU/AMDGPUISelDAGToDAG.cpp b/llvm/lib/Target/AMDGPU/AMDGPUISelDAGToDAG.cpp index 2192a72bb27b7..f95d9edfd1ffc 100644 --- a/llvm/lib/Target/AMDGPU/AMDGPUISelDAGToDAG.cpp +++ b/llvm/lib/Target/AMDGPU/AMDGPUISelDAGToDAG.cpp @@ -1111,8 +1111,7 @@ void AMDGPUDAGToDAGISel::SelectUADDO_USUBO(SDNode *N) { {N->getOperand(0), N->getOperand(1), CurDAG->getTargetConstant(0, {}, MVT::i1) /*clamp bit*/}); } else { - unsigned Opc = N->getOpcode() == ISD::UADDO ? AMDGPU::S_UADDO_PSEUDO - : AMDGPU::S_USUBO_PSEUDO; + unsigned Opc = IsAdd ? AMDGPU::S_UADDO_PSEUDO : AMDGPU::S_USUBO_PSEUDO; CurDAG->SelectNodeTo(N, Opc, N->getVTList(), {N->getOperand(0), N->getOperand(1)}); diff --git a/llvm/lib/Target/AMDGPU/SIISelLowering.cpp b/llvm/lib/Target/AMDGPU/SIISelLowering.cpp index 16530087444d2..0ba231aa40c28 100644 --- a/llvm/lib/Target/AMDGPU/SIISelLowering.cpp +++ b/llvm/lib/Target/AMDGPU/SIISelLowering.cpp @@ -6081,9 +6081,6 @@ SITargetLowering::EmitInstrWithCustomInserter(MachineInstr &MI, MachineOperand &Src0 = MI.getOperand(2); MachineOperand &Src1 = MI.getOperand(3); MachineOperand &Src2 = MI.getOperand(4); - unsigned Opc = (MI.getOpcode() == AMDGPU::S_ADD_CO_PSEUDO) - ? AMDGPU::S_ADDC_U32 - : AMDGPU::S_SUBB_U32; if (Src0.isReg() && TRI->isVectorRegister(MRI, Src0.getReg())) { Register RegOp0 = MRI.createVirtualRegister(&AMDGPU::SReg_32_XM0RegClass); BuildMI(*BB, MII, DL, TII->get(AMDGPU::V_READFIRSTLANE_B32), RegOp0) @@ -6132,11 +6129,11 @@ SITargetLowering::EmitInstrWithCustomInserter(MachineInstr &MI, .addImm(0); } - // clang-format off - BuildMI(*BB, MII, DL, TII->get(Opc), Dest.getReg()) - .add(Src0) - .add(Src1); - // clang-format on + unsigned Opc = MI.getOpcode() == AMDGPU::S_ADD_CO_PSEUDO + ? AMDGPU::S_ADDC_U32 + : AMDGPU::S_SUBB_U32; + + BuildMI(*BB, MII, DL, TII->get(Opc), Dest.getReg()).add(Src0).add(Src1); unsigned SelOpc = ST.isWave64() ? AMDGPU::S_CSELECT_B64 : AMDGPU::S_CSELECT_B32; @@ -16579,6 +16576,53 @@ SDValue SITargetLowering::performSetCCCombine(SDNode *N, } } + // Eliminate setcc by using carryout from add/sub instruction + + // LHS = ADD i64 RHS, Z LHSlo = UADDO i32 RHSlo, Zlo + // setcc LHS ult RHS -> LHSHi = UADDO_CARRY i32 RHShi, Zhi + // similarly for subtraction + + // LHS = ADD i64 Y, 1 LHSlo = UADDO i32 Ylo, 1 + // setcc LHS eq 0 -> LHSHi = UADDO_CARRY i32 Yhi, 0 + + if (VT == MVT::i64 && ((CC == ISD::SETULT && + sd_match(LHS, m_Add(m_Specific(RHS), m_Value()))) || + (CC == ISD::SETUGT && + sd_match(LHS, m_Sub(m_Specific(RHS), m_Value()))) || + (CC == ISD::SETEQ && CRHS && CRHS->isZero() && + sd_match(LHS, m_Add(m_Value(), m_One()))))) { + bool IsAdd = LHS.getOpcode() == ISD::ADD; + + SDValue Op0 = LHS.getOperand(0); + SDValue Op1 = LHS.getOperand(1); + + SDValue Op0Lo = DAG.getNode(ISD::TRUNCATE, SL, MVT::i32, Op0); + SDValue Op1Lo = DAG.getNode(ISD::TRUNCATE, SL, MVT::i32, Op1); + + SDValue Op0Hi = getHiHalf64(Op0, DAG); + SDValue Op1Hi = getHiHalf64(Op1, DAG); + + SDValue NodeLo = + DAG.getNode(IsAdd ? ISD::UADDO : ISD::USUBO, SL, + DAG.getVTList(MVT::i32, MVT::i1), {Op0Lo, Op1Lo}); + + SDValue CarryInHi = NodeLo.getValue(1); + SDValue NodeHi = DAG.getNode(IsAdd ? ISD::UADDO_CARRY : ISD::USUBO_CARRY, + SL, DAG.getVTList(MVT::i32, MVT::i1), + {Op0Hi, Op1Hi, CarryInHi}); + + SDValue ResultLo = NodeLo.getValue(0); + SDValue ResultHi = NodeHi.getValue(0); + + SDValue JoinedResult = + DAG.getBuildVector(MVT::v2i32, SL, {ResultLo, ResultHi}); + + SDValue Result = DAG.getNode(ISD::BITCAST, SL, VT, JoinedResult); + SDValue Overflow = NodeHi.getValue(1); + DCI.CombineTo(LHS.getNode(), Result); + return Overflow; + } + if (VT != MVT::f32 && VT != MVT::f64 && (!Subtarget->has16BitInsts() || VT != MVT::f16)) return SDValue(); diff --git a/llvm/test/CodeGen/AMDGPU/a-v-flat-atomicrmw.ll b/llvm/test/CodeGen/AMDGPU/a-v-flat-atomicrmw.ll index d053425afbb6d..3f5c1d41cc7a9 100644 --- a/llvm/test/CodeGen/AMDGPU/a-v-flat-atomicrmw.ll +++ b/llvm/test/CodeGen/AMDGPU/a-v-flat-atomicrmw.ll @@ -8760,9 +8760,8 @@ define void @flat_atomic_usub_sat_i64_ret_a_a(ptr %ptr) #0 { ; GFX90A-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) ; GFX90A-NEXT: v_sub_co_u32_e32 v0, vcc, v2, v6 ; GFX90A-NEXT: v_subb_co_u32_e32 v1, vcc, v3, v7, vcc -; GFX90A-NEXT: v_cmp_gt_u64_e32 vcc, v[0:1], v[2:3] -; GFX90A-NEXT: v_cndmask_b32_e64 v1, v1, 0, vcc ; GFX90A-NEXT: v_cndmask_b32_e64 v0, v0, 0, vcc +; GFX90A-NEXT: v_cndmask_b32_e64 v1, v1, 0, vcc ; GFX90A-NEXT: flat_atomic_cmpswap_x2 v[0:1], v[4:5], v[0:3] glc ; GFX90A-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) ; GFX90A-NEXT: v_cmp_eq_u64_e32 vcc, v[0:1], v[2:3] @@ -8781,20 +8780,19 @@ define void @flat_atomic_usub_sat_i64_ret_a_a(ptr %ptr) #0 { ; GFX90A-NEXT: s_cbranch_execz .LBB113_6 ; GFX90A-NEXT: ; %bb.5: ; %atomicrmw.private ; GFX90A-NEXT: v_cmp_ne_u64_e32 vcc, 0, v[4:5] -; GFX90A-NEXT: v_cndmask_b32_e32 v4, -1, v4, vcc -; GFX90A-NEXT: buffer_load_dword v0, v4, s[0:3], 0 offen -; GFX90A-NEXT: buffer_load_dword v1, v4, s[0:3], 0 offen offset:4 +; GFX90A-NEXT: v_cndmask_b32_e32 v0, -1, v4, vcc +; GFX90A-NEXT: buffer_load_dword v1, v0, s[0:3], 0 offen +; GFX90A-NEXT: buffer_load_dword v2, v0, s[0:3], 0 offen offset:4 ; GFX90A-NEXT: s_waitcnt vmcnt(1) -; GFX90A-NEXT: v_sub_co_u32_e32 v2, vcc, v0, v6 +; GFX90A-NEXT: v_sub_co_u32_e32 v3, vcc, v1, v6 ; GFX90A-NEXT: s_waitcnt vmcnt(0) -; GFX90A-NEXT: v_subb_co_u32_e32 v3, vcc, v1, v7, vcc -; GFX90A-NEXT: v_cmp_gt_u64_e32 vcc, v[2:3], v[0:1] -; GFX90A-NEXT: v_accvgpr_write_b32 a0, v0 -; GFX90A-NEXT: v_cndmask_b32_e64 v0, v3, 0, vcc -; GFX90A-NEXT: v_accvgpr_write_b32 a1, v1 -; GFX90A-NEXT: v_cndmask_b32_e64 v2, v2, 0, vcc -; GFX90A-NEXT: buffer_store_dword v0, v4, s[0:3], 0 offen offset:4 -; GFX90A-NEXT: buffer_store_dword v2, v4, s[0:3], 0 offen +; GFX90A-NEXT: v_subb_co_u32_e32 v4, vcc, v2, v7, vcc +; GFX90A-NEXT: v_accvgpr_write_b32 a0, v1 +; GFX90A-NEXT: v_cndmask_b32_e64 v3, v3, 0, vcc +; GFX90A-NEXT: v_accvgpr_write_b32 a1, v2 +; GFX90A-NEXT: v_cndmask_b32_e64 v1, v4, 0, vcc +; GFX90A-NEXT: buffer_store_dword v3, v0, s[0:3], 0 offen +; GFX90A-NEXT: buffer_store_dword v1, v0, s[0:3], 0 offen offset:4 ; GFX90A-NEXT: .LBB113_6: ; %atomicrmw.phi ; GFX90A-NEXT: s_or_b64 exec, exec, s[4:5] ; GFX90A-NEXT: ;;#ASMSTART @@ -8828,10 +8826,9 @@ define void @flat_atomic_usub_sat_i64_ret_a_a(ptr %ptr) #0 { ; GFX950-NEXT: v_sub_co_u32_e32 v0, vcc, v2, v6 ; GFX950-NEXT: s_nop 1 ; GFX950-NEXT: v_subb_co_u32_e32 v1, vcc, v3, v7, vcc -; GFX950-NEXT: v_cmp_gt_u64_e32 vcc, v[0:1], v[2:3] ; GFX950-NEXT: s_nop 1 -; GFX950-NEXT: v_cndmask_b32_e64 v1, v1, 0, vcc ; GFX950-NEXT: v_cndmask_b32_e64 v0, v0, 0, vcc +; GFX950-NEXT: v_cndmask_b32_e64 v1, v1, 0, vcc ; GFX950-NEXT: flat_atomic_cmpswap_x2 v[0:1], v[4:5], v[0:3] sc0 ; GFX950-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) ; GFX950-NEXT: v_cmp_eq_u64_e32 vcc, v[0:1], v[2:3] @@ -8857,11 +8854,11 @@ define void @flat_atomic_usub_sat_i64_ret_a_a(ptr %ptr) #0 { ; GFX950-NEXT: v_sub_co_u32_e32 v2, vcc, v0, v6 ; GFX950-NEXT: s_nop 1 ; GFX950-NEXT: v_subb_co_u32_e32 v3, vcc, v1, v7, vcc -; GFX950-NEXT: v_cmp_gt_u64_e32 vcc, v[2:3], v[0:1] ; GFX950-NEXT: v_accvgpr_write_b32 a0, v0 -; GFX950-NEXT: v_accvgpr_write_b32 a1, v1 +; GFX950-NEXT: s_nop 0 ; GFX950-NEXT: v_cndmask_b32_e64 v3, v3, 0, vcc ; GFX950-NEXT: v_cndmask_b32_e64 v2, v2, 0, vcc +; GFX950-NEXT: v_accvgpr_write_b32 a1, v1 ; GFX950-NEXT: scratch_store_dwordx2 v4, v[2:3], off ; GFX950-NEXT: .LBB113_6: ; %atomicrmw.phi ; GFX950-NEXT: s_or_b64 exec, exec, s[0:1] @@ -8901,9 +8898,8 @@ define void @flat_atomic_usub_sat_i64_ret_av_av(ptr %ptr) #0 { ; GFX90A-NEXT: v_pk_mov_b32 v[6:7], v[4:5], v[4:5] op_sel:[0,1] ; GFX90A-NEXT: v_sub_co_u32_e32 v4, vcc, v6, v2 ; GFX90A-NEXT: v_subb_co_u32_e32 v5, vcc, v7, v3, vcc -; GFX90A-NEXT: v_cmp_gt_u64_e32 vcc, v[4:5], v[6:7] -; GFX90A-NEXT: v_cndmask_b32_e64 v5, v5, 0, vcc ; GFX90A-NEXT: v_cndmask_b32_e64 v4, v4, 0, vcc +; GFX90A-NEXT: v_cndmask_b32_e64 v5, v5, 0, vcc ; GFX90A-NEXT: flat_atomic_cmpswap_x2 v[4:5], v[0:1], v[4:7] glc ; GFX90A-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) ; GFX90A-NEXT: v_cmp_eq_u64_e32 vcc, v[4:5], v[6:7] @@ -8919,18 +8915,17 @@ define void @flat_atomic_usub_sat_i64_ret_av_av(ptr %ptr) #0 { ; GFX90A-NEXT: s_cbranch_execz .LBB114_6 ; GFX90A-NEXT: ; %bb.5: ; %atomicrmw.private ; GFX90A-NEXT: v_cmp_ne_u64_e32 vcc, 0, v[0:1] -; GFX90A-NEXT: v_cndmask_b32_e32 v6, -1, v0, vcc -; GFX90A-NEXT: buffer_load_dword v4, v6, s[0:3], 0 offen -; GFX90A-NEXT: buffer_load_dword v5, v6, s[0:3], 0 offen offset:4 +; GFX90A-NEXT: v_cndmask_b32_e32 v0, -1, v0, vcc +; GFX90A-NEXT: buffer_load_dword v4, v0, s[0:3], 0 offen +; GFX90A-NEXT: buffer_load_dword v5, v0, s[0:3], 0 offen offset:4 ; GFX90A-NEXT: s_waitcnt vmcnt(1) -; GFX90A-NEXT: v_sub_co_u32_e32 v0, vcc, v4, v2 +; GFX90A-NEXT: v_sub_co_u32_e32 v1, vcc, v4, v2 ; GFX90A-NEXT: s_waitcnt vmcnt(0) -; GFX90A-NEXT: v_subb_co_u32_e32 v1, vcc, v5, v3, vcc -; GFX90A-NEXT: v_cmp_gt_u64_e32 vcc, v[0:1], v[4:5] -; GFX90A-NEXT: v_cndmask_b32_e64 v0, v0, 0, vcc +; GFX90A-NEXT: v_subb_co_u32_e32 v2, vcc, v5, v3, vcc ; GFX90A-NEXT: v_cndmask_b32_e64 v1, v1, 0, vcc -; GFX90A-NEXT: buffer_store_dword v0, v6, s[0:3], 0 offen -; GFX90A-NEXT: buffer_store_dword v1, v6, s[0:3], 0 offen offset:4 +; GFX90A-NEXT: v_cndmask_b32_e64 v2, v2, 0, vcc +; GFX90A-NEXT: buffer_store_dword v1, v0, s[0:3], 0 offen +; GFX90A-NEXT: buffer_store_dword v2, v0, s[0:3], 0 offen offset:4 ; GFX90A-NEXT: .LBB114_6: ; %atomicrmw.phi ; GFX90A-NEXT: s_or_b64 exec, exec, s[4:5] ; GFX90A-NEXT: ;;#ASMSTART @@ -8963,10 +8958,9 @@ define void @flat_atomic_usub_sat_i64_ret_av_av(ptr %ptr) #0 { ; GFX950-NEXT: v_sub_co_u32_e32 v2, vcc, v8, v0 ; GFX950-NEXT: s_nop 1 ; GFX950-NEXT: v_subb_co_u32_e32 v3, vcc, v9, v1, vcc -; GFX950-NEXT: v_cmp_gt_u64_e32 vcc, v[2:3], v[8:9] ; GFX950-NEXT: s_nop 1 -; GFX950-NEXT: v_cndmask_b32_e64 v7, v3, 0, vcc ; GFX950-NEXT: v_cndmask_b32_e64 v6, v2, 0, vcc +; GFX950-NEXT: v_cndmask_b32_e64 v7, v3, 0, vcc ; GFX950-NEXT: flat_atomic_cmpswap_x2 v[2:3], v[4:5], v[6:9] sc0 ; GFX950-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) ; GFX950-NEXT: v_cmp_eq_u64_e32 vcc, v[2:3], v[8:9] @@ -8989,7 +8983,6 @@ define void @flat_atomic_usub_sat_i64_ret_av_av(ptr %ptr) #0 { ; GFX950-NEXT: v_sub_co_u32_e32 v0, vcc, v2, v0 ; GFX950-NEXT: s_nop 1 ; GFX950-NEXT: v_subb_co_u32_e32 v1, vcc, v3, v1, vcc -; GFX950-NEXT: v_cmp_gt_u64_e32 vcc, v[0:1], v[2:3] ; GFX950-NEXT: s_nop 1 ; GFX950-NEXT: v_cndmask_b32_e64 v1, v1, 0, vcc ; GFX950-NEXT: v_cndmask_b32_e64 v0, v0, 0, vcc @@ -17065,9 +17058,8 @@ define void @flat_atomic_usub_sat_i64_saddr_ret_a_a(ptr inreg %ptr) #0 { ; GFX90A-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) ; GFX90A-NEXT: v_sub_co_u32_e32 v0, vcc, v2, v4 ; GFX90A-NEXT: v_subb_co_u32_e32 v1, vcc, v3, v5, vcc -; GFX90A-NEXT: v_cmp_gt_u64_e32 vcc, v[0:1], v[2:3] -; GFX90A-NEXT: v_cndmask_b32_e64 v1, v1, 0, vcc ; GFX90A-NEXT: v_cndmask_b32_e64 v0, v0, 0, vcc +; GFX90A-NEXT: v_cndmask_b32_e64 v1, v1, 0, vcc ; GFX90A-NEXT: flat_atomic_cmpswap_x2 v[0:1], v[6:7], v[0:3] glc ; GFX90A-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) ; GFX90A-NEXT: v_cmp_eq_u64_e32 vcc, v[0:1], v[2:3] @@ -17086,20 +17078,19 @@ define void @flat_atomic_usub_sat_i64_saddr_ret_a_a(ptr inreg %ptr) #0 { ; GFX90A-NEXT: ; %bb.5: ; %atomicrmw.private ; GFX90A-NEXT: s_cmp_lg_u64 s[4:5], 0 ; GFX90A-NEXT: s_cselect_b32 s4, s4, -1 -; GFX90A-NEXT: v_mov_b32_e32 v6, s4 -; GFX90A-NEXT: buffer_load_dword v0, v6, s[0:3], 0 offen -; GFX90A-NEXT: buffer_load_dword v1, v6, s[0:3], 0 offen offset:4 +; GFX90A-NEXT: v_mov_b32_e32 v0, s4 +; GFX90A-NEXT: buffer_load_dword v1, v0, s[0:3], 0 offen +; GFX90A-NEXT: buffer_load_dword v2, v0, s[0:3], 0 offen offset:4 ; GFX90A-NEXT: s_waitcnt vmcnt(1) -; GFX90A-NEXT: v_sub_co_u32_e32 v2, vcc, v0, v4 +; GFX90A-NEXT: v_sub_co_u32_e32 v3, vcc, v1, v4 ; GFX90A-NEXT: s_waitcnt vmcnt(0) -; GFX90A-NEXT: v_subb_co_u32_e32 v3, vcc, v1, v5, vcc -; GFX90A-NEXT: v_cmp_gt_u64_e32 vcc, v[2:3], v[0:1] -; GFX90A-NEXT: v_accvgpr_write_b32 a0, v0 -; GFX90A-NEXT: v_cndmask_b32_e64 v0, v3, 0, vcc -; GFX90A-NEXT: v_accvgpr_write_b32 a1, v1 -; GFX90A-NEXT: v_cndmask_b32_e64 v2, v2, 0, vcc -; GFX90A-NEXT: buffer_store_dword v0, v6, s[0:3], 0 offen offset:4 -; GFX90A-NEXT: buffer_store_dword v2, v6, s[0:3], 0 offen +; GFX90A-NEXT: v_subb_co_u32_e32 v4, vcc, v2, v5, vcc +; GFX90A-NEXT: v_accvgpr_write_b32 a0, v1 +; GFX90A-NEXT: v_cndmask_b32_e64 v3, v3, 0, vcc +; GFX90A-NEXT: v_accvgpr_write_b32 a1, v2 +; GFX90A-NEXT: v_cndmask_b32_e64 v1, v4, 0, vcc +; GFX90A-NEXT: buffer_store_dword v3, v0, s[0:3], 0 offen +; GFX90A-NEXT: buffer_store_dword v1, v0, s[0:3], 0 offen offset:4 ; GFX90A-NEXT: .LBB221_6: ; %atomicrmw.phi ; GFX90A-NEXT: ;;#ASMSTART ; GFX90A-NEXT: ; use a[0:1] @@ -17132,10 +17123,9 @@ define void @flat_atomic_usub_sat_i64_saddr_ret_a_a(ptr inreg %ptr) #0 { ; GFX950-NEXT: v_sub_co_u32_e32 v0, vcc, v2, v4 ; GFX950-NEXT: s_nop 1 ; GFX950-NEXT: v_subb_co_u32_e32 v1, vcc, v3, v5, vcc -; GFX950-NEXT: v_cmp_gt_u64_e32 vcc, v[0:1], v[2:3] ; GFX950-NEXT: s_nop 1 -; GFX950-NEXT: v_cndmask_b32_e64 v1, v1, 0, vcc ; GFX950-NEXT: v_cndmask_b32_e64 v0, v0, 0, vcc +; GFX950-NEXT: v_cndmask_b32_e64 v1, v1, 0, vcc ; GFX950-NEXT: flat_atomic_cmpswap_x2 v[0:1], v[6:7], v[0:3] sc0 ; GFX950-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) ; GFX950-NEXT: v_cmp_eq_u64_e32 vcc, v[0:1], v[2:3] @@ -17159,11 +17149,11 @@ define void @flat_atomic_usub_sat_i64_saddr_ret_a_a(ptr inreg %ptr) #0 { ; GFX950-NEXT: v_sub_co_u32_e32 v2, vcc, v0, v4 ; GFX950-NEXT: s_nop 1 ; GFX950-NEXT: v_subb_co_u32_e32 v3, vcc, v1, v5, vcc -; GFX950-NEXT: v_cmp_gt_u64_e32 vcc, v[2:3], v[0:1] ; GFX950-NEXT: v_accvgpr_write_b32 a0, v0 -; GFX950-NEXT: v_accvgpr_write_b32 a1, v1 +; GFX950-NEXT: s_nop 0 ; GFX950-NEXT: v_cndmask_b32_e64 v3, v3, 0, vcc ; GFX950-NEXT: v_cndmask_b32_e64 v2, v2, 0, vcc +; GFX950-NEXT: v_accvgpr_write_b32 a1, v1 ; GFX950-NEXT: scratch_store_dwordx2 off, v[2:3], s0 ; GFX950-NEXT: .LBB221_6: ; %atomicrmw.phi ; GFX950-NEXT: ;;#ASMSTART @@ -17202,9 +17192,8 @@ define void @flat_atomic_usub_sat_i64_saddr_ret_av_av(ptr inreg %ptr) #0 { ; GFX90A-NEXT: v_pk_mov_b32 v[8:9], v[2:3], v[2:3] op_sel:[0,1] ; GFX90A-NEXT: v_sub_co_u32_e32 v2, vcc, v8, v0 ; GFX90A-NEXT: v_subb_co_u32_e32 v3, vcc, v9, v1, vcc -; GFX90A-NEXT: v_cmp_gt_u64_e32 vcc, v[2:3], v[8:9] -; GFX90A-NEXT: v_cndmask_b32_e64 v7, v3, 0, vcc ; GFX90A-NEXT: v_cndmask_b32_e64 v6, v2, 0, vcc +; GFX90A-NEXT: v_cndmask_b32_e64 v7, v3, 0, vcc ; GFX90A-NEXT: flat_atomic_cmpswap_x2 v[2:3], v[4:5], v[6:9] glc ; GFX90A-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) ; GFX90A-NEXT: v_cmp_eq_u64_e32 vcc, v[2:3], v[8:9] @@ -17227,7 +17216,6 @@ define void @flat_atomic_usub_sat_i64_saddr_ret_av_av(ptr inreg %ptr) #0 { ; GFX90A-NEXT: v_sub_co_u32_e32 v0, vcc, v2, v0 ; GFX90A-NEXT: s_waitcnt vmcnt(0) ; GFX90A-NEXT: v_subb_co_u32_e32 v1, vcc, v3, v1, vcc -; GFX90A-NEXT: v_cmp_gt_u64_e32 vcc, v[0:1], v[2:3] ; GFX90A-NEXT: v_cndmask_b32_e64 v0, v0, 0, vcc ; GFX90A-NEXT: v_cndmask_b32_e64 v1, v1, 0, vcc ; GFX90A-NEXT: buffer_store_dword v0, v4, s[0:3], 0 offen @@ -17263,10 +17251,9 @@ define void @flat_atomic_usub_sat_i64_saddr_ret_av_av(ptr inreg %ptr) #0 { ; GFX950-NEXT: v_sub_co_u32_e32 v2, vcc, v8, v0 ; GFX950-NEXT: s_nop 1 ; GFX950-NEXT: v_subb_co_u32_e32 v3, vcc, v9, v1, vcc -; GFX950-NEXT: v_cmp_gt_u64_e32 vcc, v[2:3], v[8:9] ; GFX950-NEXT: s_nop 1 -; GFX950-NEXT: v_cndmask_b32_e64 v7, v3, 0, vcc ; GFX950-NEXT: v_cndmask_b32_e64 v6, v2, 0, vcc +; GFX950-NEXT: v_cndmask_b32_e64 v7, v3, 0, vcc ; GFX950-NEXT: flat_atomic_cmpswap_x2 v[2:3], v[4:5], v[6:9] sc0 ; GFX950-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) ; GFX950-NEXT: v_cmp_eq_u64_e32 vcc, v[2:3], v[8:9] @@ -17287,7 +17274,6 @@ define void @flat_atomic_usub_sat_i64_saddr_ret_av_av(ptr inreg %ptr) #0 { ; GFX950-NEXT: v_sub_co_u32_e32 v0, vcc, v2, v0 ; GFX950-NEXT: s_nop 1 ; GFX950-NEXT: v_subb_co_u32_e32 v1, vcc, v3, v1, vcc -; GFX950-NEXT: v_cmp_gt_u64_e32 vcc, v[0:1], v[2:3] ; GFX950-NEXT: s_nop 1 ; GFX950-NEXT: v_cndmask_b32_e64 v1, v1, 0, vcc ; GFX950-NEXT: v_cndmask_b32_e64 v0, v0, 0, vcc diff --git a/llvm/test/CodeGen/AMDGPU/a-v-global-atomicrmw.ll b/llvm/test/CodeGen/AMDGPU/a-v-global-atomicrmw.ll index c98fff96d7b8a..34a4899123749 100644 --- a/llvm/test/CodeGen/AMDGPU/a-v-global-atomicrmw.ll +++ b/llvm/test/CodeGen/AMDGPU/a-v-global-atomicrmw.ll @@ -5804,9 +5804,8 @@ define void @global_atomic_usub_sat_i64_ret_a_a(ptr addrspace(1) %ptr) #0 { ; GFX90A-NEXT: s_waitcnt vmcnt(0) ; GFX90A-NEXT: v_sub_co_u32_e32 v2, vcc, v4, v6 ; GFX90A-NEXT: v_subb_co_u32_e32 v3, vcc, v5, v7, vcc -; GFX90A-NEXT: v_cmp_gt_u64_e32 vcc, v[2:3], v[4:5] -; GFX90A-NEXT: v_cndmask_b32_e64 v3, v3, 0, vcc ; GFX90A-NEXT: v_cndmask_b32_e64 v2, v2, 0, vcc +; GFX90A-NEXT: v_cndmask_b32_e64 v3, v3, 0, vcc ; GFX90A-NEXT: global_atomic_cmpswap_x2 v[2:3], v[0:1], v[2:5], off offset:80 glc ; GFX90A-NEXT: s_waitcnt vmcnt(0) ; GFX90A-NEXT: v_cmp_eq_u64_e32 vcc, v[2:3], v[4:5] @@ -5839,10 +5838,9 @@ define void @global_atomic_usub_sat_i64_ret_a_a(ptr addrspace(1) %ptr) #0 { ; GFX950-NEXT: v_sub_co_u32_e32 v2, vcc, v4, v6 ; GFX950-NEXT: s_nop 1 ; GFX950-NEXT: v_subb_co_u32_e32 v3, vcc, v5, v7, vcc -; GFX950-NEXT: v_cmp_gt_u64_e32 vcc, v[2:3], v[4:5] ; GFX950-NEXT: s_nop 1 -; GFX950-NEXT: v_cndmask_b32_e64 v3, v3, 0, vcc ; GFX950-NEXT: v_cndmask_b32_e64 v2, v2, 0, vcc +; GFX950-NEXT: v_cndmask_b32_e64 v3, v3, 0, vcc ; GFX950-NEXT: global_atomic_cmpswap_x2 v[2:3], v[0:1], v[2:5], off offset:80 sc0 ; GFX950-NEXT: s_waitcnt vmcnt(0) ; GFX950-NEXT: v_cmp_eq_u64_e32 vcc, v[2:3], v[4:5] @@ -5880,9 +5878,8 @@ define void @global_atomic_usub_sat_i64_ret_av_av(ptr addrspace(1) %ptr) #0 { ; GFX90A-NEXT: v_pk_mov_b32 v[6:7], v[4:5], v[4:5] op_sel:[0,1] ; GFX90A-NEXT: v_sub_co_u32_e32 v4, vcc, v6, v2 ; GFX90A-NEXT: v_subb_co_u32_e32 v5, vcc, v7, v3, vcc -; GFX90A-NEXT: v_cmp_gt_u64_e32 vcc, v[4:5], v[6:7] -; GFX90A-NEXT: v_cndmask_b32_e64 v5, v5, 0, vcc ; GFX90A-NEXT: v_cndmask_b32_e64 v4, v4, 0, vcc +; GFX90A-NEXT: v_cndmask_b32_e64 v5, v5, 0, vcc ; GFX90A-NEXT: global_atomic_cmpswap_x2 v[4:5], v[0:1], v[4:7], off offset:80 glc ; GFX90A-NEXT: s_waitcnt vmcnt(0) ; GFX90A-NEXT: v_cmp_eq_u64_e32 vcc, v[4:5], v[6:7] @@ -5911,10 +5908,9 @@ define void @global_atomic_usub_sat_i64_ret_av_av(ptr addrspace(1) %ptr) #0 { ; GFX950-NEXT: v_sub_co_u32_e32 v4, vcc, v6, v2 ; GFX950-NEXT: s_nop 1 ; GFX950-NEXT: v_subb_co_u32_e32 v5, vcc, v7, v3, vcc -; GFX950-NEXT: v_cmp_gt_u64_e32 vcc, v[4:5], v[6:7] ; GFX950-NEXT: s_nop 1 -; GFX950-NEXT: v_cndmask_b32_e64 v5, v5, 0, vcc ; GFX950-NEXT: v_cndmask_b32_e64 v4, v4, 0, vcc +; GFX950-NEXT: v_cndmask_b32_e64 v5, v5, 0, vcc ; GFX950-NEXT: global_atomic_cmpswap_x2 v[4:5], v[0:1], v[4:7], off offset:80 sc0 ; GFX950-NEXT: s_waitcnt vmcnt(0) ; GFX950-NEXT: v_cmp_eq_u64_e32 vcc, v[4:5], v[6:7] @@ -11573,9 +11569,8 @@ define void @global_atomic_usub_sat_i64_saddr_ret_a_a(ptr addrspace(1) inreg %pt ; GFX90A-NEXT: s_waitcnt vmcnt(0) ; GFX90A-NEXT: v_sub_co_u32_e32 v0, vcc, v2, v4 ; GFX90A-NEXT: v_subb_co_u32_e32 v1, vcc, v3, v5, vcc -; GFX90A-NEXT: v_cmp_gt_u64_e32 vcc, v[0:1], v[2:3] -; GFX90A-NEXT: v_cndmask_b32_e64 v1, v1, 0, vcc ; GFX90A-NEXT: v_cndmask_b32_e64 v0, v0, 0, vcc +; GFX90A-NEXT: v_cndmask_b32_e64 v1, v1, 0, vcc ; GFX90A-NEXT: global_atomic_cmpswap_x2 v[0:1], v6, v[0:3], s[16:17] offset:80 glc ; GFX90A-NEXT: s_waitcnt vmcnt(0) ; GFX90A-NEXT: v_cmp_eq_u64_e32 vcc, v[0:1], v[2:3] @@ -11609,10 +11604,9 @@ define void @global_atomic_usub_sat_i64_saddr_ret_a_a(ptr addrspace(1) inreg %pt ; GFX950-NEXT: v_sub_co_u32_e32 v0, vcc, v2, v4 ; GFX950-NEXT: s_nop 1 ; GFX950-NEXT: v_subb_co_u32_e32 v1, vcc, v3, v5, vcc -; GFX950-NEXT: v_cmp_gt_u64_e32 vcc, v[0:1], v[2:3] ; GFX950-NEXT: s_nop 1 -; GFX950-NEXT: v_cndmask_b32_e64 v1, v1, 0, vcc ; GFX950-NEXT: v_cndmask_b32_e64 v0, v0, 0, vcc +; GFX950-NEXT: v_cndmask_b32_e64 v1, v1, 0, vcc ; GFX950-NEXT: global_atomic_cmpswap_x2 v[0:1], v6, v[0:3], s[0:1] offset:80 sc0 ; GFX950-NEXT: s_waitcnt vmcnt(0) ; GFX950-NEXT: v_cmp_eq_u64_e32 vcc, v[0:1], v[2:3] @@ -11651,9 +11645,8 @@ define void @global_atomic_usub_sat_i64_saddr_ret_av_av(ptr addrspace(1) inreg % ; GFX90A-NEXT: v_pk_mov_b32 v[8:9], v[2:3], v[2:3] op_sel:[0,1] ; GFX90A-NEXT: v_sub_co_u32_e32 v2, vcc, v8, v0 ; GFX90A-NEXT: v_subb_co_u32_e32 v3, vcc, v9, v1, vcc -; GFX90A-NEXT: v_cmp_gt_u64_e32 vcc, v[2:3], v[8:9] -; GFX90A-NEXT: v_cndmask_b32_e64 v7, v3, 0, vcc ; GFX90A-NEXT: v_cndmask_b32_e64 v6, v2, 0, vcc +; GFX90A-NEXT: v_cndmask_b32_e64 v7, v3, 0, vcc ; GFX90A-NEXT: global_atomic_cmpswap_x2 v[2:3], v4, v[6:9], s[16:17] offset:80 glc ; GFX90A-NEXT: s_waitcnt vmcnt(0) ; GFX90A-NEXT: v_cmp_eq_u64_e32 vcc, v[2:3], v[8:9] @@ -11683,10 +11676,9 @@ define void @global_atomic_usub_sat_i64_saddr_ret_av_av(ptr addrspace(1) inreg % ; GFX950-NEXT: v_sub_co_u32_e32 v2, vcc, v8, v0 ; GFX950-NEXT: s_nop 1 ; GFX950-NEXT: v_subb_co_u32_e32 v3, vcc, v9, v1, vcc -; GFX950-NEXT: v_cmp_gt_u64_e32 vcc, v[2:3], v[8:9] ; GFX950-NEXT: s_nop 1 -; GFX950-NEXT: v_cndmask_b32_e64 v7, v3, 0, vcc ; GFX950-NEXT: v_cndmask_b32_e64 v6, v2, 0, vcc +; GFX950-NEXT: v_cndmask_b32_e64 v7, v3, 0, vcc ; GFX950-NEXT: global_atomic_cmpswap_x2 v[2:3], v4, v[6:9], s[0:1] offset:80 sc0 ; GFX950-NEXT: s_waitcnt vmcnt(0) ; GFX950-NEXT: v_cmp_eq_u64_e32 vcc, v[2:3], v[8:9] diff --git a/llvm/test/CodeGen/AMDGPU/addsub64_carry.ll b/llvm/test/CodeGen/AMDGPU/addsub64_carry.ll index d326966e849e2..b72eba8cdb519 100644 --- a/llvm/test/CodeGen/AMDGPU/addsub64_carry.ll +++ b/llvm/test/CodeGen/AMDGPU/addsub64_carry.ll @@ -17,12 +17,9 @@ define %struct.uint96 @v_add64_32(i64 %val64A, i64 %val64B, i32 %val32) { ; CHECK-LABEL: v_add64_32: ; CHECK: ; %bb.0: ; CHECK-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) -; CHECK-NEXT: v_add_co_u32_e32 v5, vcc, v0, v2 -; CHECK-NEXT: v_addc_co_u32_e32 v6, vcc, v1, v3, vcc -; CHECK-NEXT: v_cmp_lt_u64_e32 vcc, v[5:6], v[0:1] -; CHECK-NEXT: v_mov_b32_e32 v0, v5 +; CHECK-NEXT: v_add_co_u32_e32 v0, vcc, v0, v2 +; CHECK-NEXT: v_addc_co_u32_e32 v1, vcc, v1, v3, vcc ; CHECK-NEXT: v_addc_co_u32_e32 v2, vcc, 0, v4, vcc -; CHECK-NEXT: v_mov_b32_e32 v1, v6 ; CHECK-NEXT: s_setpc_b64 s[30:31] %sum64 = add i64 %val64A, %val64B %obit = icmp ult i64 %sum64, %val64A @@ -38,16 +35,14 @@ define <2 x i64> @v_uadd_v2i64(<2 x i64> %val0, <2 x i64> %val1, ptr %ptrval) { ; CHECK: ; %bb.0: ; CHECK-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; CHECK-NEXT: v_add_co_u32_e32 v6, vcc, v2, v6 +; CHECK-NEXT: v_add_co_u32_e64 v4, s[4:5], v0, v4 ; CHECK-NEXT: v_addc_co_u32_e32 v7, vcc, v3, v7, vcc -; CHECK-NEXT: v_add_co_u32_e32 v4, vcc, v0, v4 -; CHECK-NEXT: v_addc_co_u32_e32 v5, vcc, v1, v5, vcc -; CHECK-NEXT: v_cmp_lt_u64_e32 vcc, v[4:5], v[0:1] -; CHECK-NEXT: flat_store_dwordx4 v[8:9], v[4:7] -; CHECK-NEXT: v_cndmask_b32_e64 v0, 0, -1, vcc -; CHECK-NEXT: v_cmp_lt_u64_e32 vcc, v[6:7], v[2:3] -; CHECK-NEXT: v_mov_b32_e32 v1, v0 +; CHECK-NEXT: v_addc_co_u32_e64 v5, s[4:5], v1, v5, s[4:5] +; CHECK-NEXT: v_cndmask_b32_e64 v0, 0, -1, s[4:5] ; CHECK-NEXT: v_cndmask_b32_e64 v2, 0, -1, vcc +; CHECK-NEXT: v_mov_b32_e32 v1, v0 ; CHECK-NEXT: v_mov_b32_e32 v3, v2 +; CHECK-NEXT: flat_store_dwordx4 v[8:9], v[4:7] ; CHECK-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) ; CHECK-NEXT: s_setpc_b64 s[30:31] %pair = call {<2 x i64>, <2 x i1>} @llvm.uadd.with.overflow.v2i64(<2 x i64> %val0, <2 x i64> %val1) @@ -63,16 +58,14 @@ define <2 x i64> @v_usub_v2i64(<2 x i64> %val0, <2 x i64> %val1, ptr %ptrval) { ; CHECK: ; %bb.0: ; CHECK-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; CHECK-NEXT: v_sub_co_u32_e32 v6, vcc, v2, v6 +; CHECK-NEXT: v_sub_co_u32_e64 v4, s[4:5], v0, v4 ; CHECK-NEXT: v_subb_co_u32_e32 v7, vcc, v3, v7, vcc -; CHECK-NEXT: v_sub_co_u32_e32 v4, vcc, v0, v4 -; CHECK-NEXT: v_subb_co_u32_e32 v5, vcc, v1, v5, vcc -; CHECK-NEXT: v_cmp_gt_u64_e32 vcc, v[4:5], v[0:1] -; CHECK-NEXT: flat_store_dwordx4 v[8:9], v[4:7] -; CHECK-NEXT: v_cndmask_b32_e64 v0, 0, -1, vcc -; CHECK-NEXT: v_cmp_gt_u64_e32 vcc, v[6:7], v[2:3] -; CHECK-NEXT: v_mov_b32_e32 v1, v0 +; CHECK-NEXT: v_subb_co_u32_e64 v5, s[4:5], v1, v5, s[4:5] +; CHECK-NEXT: v_cndmask_b32_e64 v0, 0, -1, s[4:5] ; CHECK-NEXT: v_cndmask_b32_e64 v2, 0, -1, vcc +; CHECK-NEXT: v_mov_b32_e32 v1, v0 ; CHECK-NEXT: v_mov_b32_e32 v3, v2 +; CHECK-NEXT: flat_store_dwordx4 v[8:9], v[4:7] ; CHECK-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) ; CHECK-NEXT: s_setpc_b64 s[30:31] %pair = call {<2 x i64>, <2 x i1>} @llvm.usub.with.overflow.v2i64(<2 x i64> %val0, <2 x i64> %val1) @@ -87,10 +80,9 @@ define i64 @v_uadd_i64(i64 %val0, i64 %val1, ptr %ptrval) { ; CHECK-LABEL: v_uadd_i64: ; CHECK: ; %bb.0: ; CHECK-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) -; CHECK-NEXT: v_add_co_u32_e32 v2, vcc, v0, v2 -; CHECK-NEXT: v_addc_co_u32_e32 v3, vcc, v1, v3, vcc -; CHECK-NEXT: v_cmp_lt_u64_e32 vcc, v[2:3], v[0:1] -; CHECK-NEXT: flat_store_dwordx2 v[4:5], v[2:3] +; CHECK-NEXT: v_add_co_u32_e32 v0, vcc, v0, v2 +; CHECK-NEXT: v_addc_co_u32_e32 v1, vcc, v1, v3, vcc +; CHECK-NEXT: flat_store_dwordx2 v[4:5], v[0:1] ; CHECK-NEXT: v_cndmask_b32_e64 v0, 0, -1, vcc ; CHECK-NEXT: v_mov_b32_e32 v1, v0 ; CHECK-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) @@ -109,7 +101,6 @@ define i64 @v_uadd_p1(i64 %val0, i64 %val1, ptr %ptrval) { ; CHECK-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; CHECK-NEXT: v_add_co_u32_e32 v0, vcc, 1, v0 ; CHECK-NEXT: v_addc_co_u32_e32 v1, vcc, 0, v1, vcc -; CHECK-NEXT: v_cmp_eq_u64_e32 vcc, 0, v[0:1] ; CHECK-NEXT: flat_store_dwordx2 v[4:5], v[0:1] ; CHECK-NEXT: v_cndmask_b32_e64 v0, 0, -1, vcc ; CHECK-NEXT: v_mov_b32_e32 v1, v0 @@ -147,10 +138,9 @@ define i64 @v_usub_p1(i64 %val0, i64 %val1, ptr %ptrval) { ; CHECK-LABEL: v_usub_p1: ; CHECK: ; %bb.0: ; CHECK-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) -; CHECK-NEXT: v_add_co_u32_e32 v2, vcc, -1, v0 -; CHECK-NEXT: v_addc_co_u32_e32 v3, vcc, -1, v1, vcc -; CHECK-NEXT: v_cmp_gt_u64_e32 vcc, v[2:3], v[0:1] -; CHECK-NEXT: flat_store_dwordx2 v[4:5], v[2:3] +; CHECK-NEXT: v_subrev_co_u32_e32 v0, vcc, 1, v0 +; CHECK-NEXT: v_subbrev_co_u32_e32 v1, vcc, 0, v1, vcc +; CHECK-NEXT: flat_store_dwordx2 v[4:5], v[0:1] ; CHECK-NEXT: v_cndmask_b32_e64 v0, 0, -1, vcc ; CHECK-NEXT: v_mov_b32_e32 v1, v0 ; CHECK-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) @@ -167,10 +157,9 @@ define i64 @v_usub_n1(i64 %val0, i64 %val1, ptr %ptrval) { ; CHECK-LABEL: v_usub_n1: ; CHECK: ; %bb.0: ; CHECK-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) -; CHECK-NEXT: v_add_co_u32_e32 v2, vcc, 1, v0 -; CHECK-NEXT: v_addc_co_u32_e32 v3, vcc, 0, v1, vcc -; CHECK-NEXT: v_cmp_gt_u64_e32 vcc, v[2:3], v[0:1] -; CHECK-NEXT: flat_store_dwordx2 v[4:5], v[2:3] +; CHECK-NEXT: v_subrev_co_u32_e32 v0, vcc, -1, v0 +; CHECK-NEXT: v_subbrev_co_u32_e32 v1, vcc, -1, v1, vcc +; CHECK-NEXT: flat_store_dwordx2 v[4:5], v[0:1] ; CHECK-NEXT: v_cndmask_b32_e64 v0, 0, -1, vcc ; CHECK-NEXT: v_mov_b32_e32 v1, v0 ; CHECK-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) @@ -190,15 +179,13 @@ define i64 @v_usub_n1(i64 %val0, i64 %val1, ptr %ptrval) { define amdgpu_ps %struct.uint96 @s_add64_32(i64 inreg %val64A, i64 inreg %val64B, i32 inreg %val32) { ; CHECK-LABEL: s_add64_32: ; CHECK: ; %bb.0: -; CHECK-NEXT: s_add_u32 s6, s0, s2 -; CHECK-NEXT: v_mov_b32_e32 v0, s0 -; CHECK-NEXT: s_addc_u32 s7, s1, s3 -; CHECK-NEXT: v_mov_b32_e32 v1, s1 -; CHECK-NEXT: v_cmp_lt_u64_e32 vcc, s[6:7], v[0:1] -; CHECK-NEXT: s_mov_b32 s0, s6 -; CHECK-NEXT: s_cmp_lg_u64 vcc, 0 +; CHECK-NEXT: s_add_u32 s0, s0, s2 +; CHECK-NEXT: s_cselect_b64 s[6:7], -1, 0 +; CHECK-NEXT: s_cmp_lg_u64 s[6:7], 0 +; CHECK-NEXT: s_addc_u32 s1, s1, s3 +; CHECK-NEXT: s_cselect_b64 s[2:3], -1, 0 +; CHECK-NEXT: s_cmp_lg_u64 s[2:3], 0 ; CHECK-NEXT: s_addc_u32 s2, s4, 0 -; CHECK-NEXT: s_mov_b32 s1, s7 ; CHECK-NEXT: ; return to shader part epilog %sum64 = add i64 %val64A, %val64B %obit = icmp ult i64 %sum64, %val64A @@ -212,24 +199,24 @@ define amdgpu_ps %struct.uint96 @s_add64_32(i64 inreg %val64A, i64 inreg %val64B define amdgpu_ps <2 x i64> @s_uadd_v2i64(<2 x i64> inreg %val0, <2 x i64> inreg %val1, ptr %ptrval) { ; CHECK-LABEL: s_uadd_v2i64: ; CHECK: ; %bb.0: -; CHECK-NEXT: s_add_u32 s6, s2, s6 -; CHECK-NEXT: v_mov_b32_e32 v9, s3 -; CHECK-NEXT: s_addc_u32 s7, s3, s7 -; CHECK-NEXT: v_mov_b32_e32 v8, s2 -; CHECK-NEXT: s_add_u32 s4, s0, s4 -; CHECK-NEXT: v_mov_b32_e32 v7, s1 -; CHECK-NEXT: v_cmp_lt_u64_e32 vcc, s[6:7], v[8:9] -; CHECK-NEXT: s_addc_u32 s5, s1, s5 -; CHECK-NEXT: v_mov_b32_e32 v6, s0 -; CHECK-NEXT: v_cndmask_b32_e64 v8, 0, -1, vcc -; CHECK-NEXT: v_cmp_lt_u64_e32 vcc, s[4:5], v[6:7] -; CHECK-NEXT: v_readfirstlane_b32 s2, v8 -; CHECK-NEXT: v_cndmask_b32_e64 v6, 0, -1, vcc -; CHECK-NEXT: v_readfirstlane_b32 s0, v6 -; CHECK-NEXT: v_mov_b32_e32 v2, s4 -; CHECK-NEXT: v_mov_b32_e32 v3, s5 -; CHECK-NEXT: v_mov_b32_e32 v4, s6 -; CHECK-NEXT: v_mov_b32_e32 v5, s7 +; CHECK-NEXT: s_add_u32 s10, s2, s6 +; CHECK-NEXT: s_cselect_b64 s[8:9], -1, 0 +; CHECK-NEXT: s_cmp_lg_u64 s[8:9], 0 +; CHECK-NEXT: s_addc_u32 s8, s3, s7 +; CHECK-NEXT: s_cselect_b64 s[2:3], -1, 0 +; CHECK-NEXT: s_add_u32 s0, s0, s4 +; CHECK-NEXT: s_cselect_b64 s[6:7], -1, 0 +; CHECK-NEXT: s_cmp_lg_u64 s[6:7], 0 +; CHECK-NEXT: s_addc_u32 s1, s1, s5 +; CHECK-NEXT: v_mov_b32_e32 v2, s0 +; CHECK-NEXT: v_mov_b32_e32 v3, s1 +; CHECK-NEXT: s_cselect_b64 s[0:1], -1, 0 +; CHECK-NEXT: v_cndmask_b32_e64 v6, 0, -1, s[2:3] +; CHECK-NEXT: v_cndmask_b32_e64 v7, 0, -1, s[0:1] +; CHECK-NEXT: v_readfirstlane_b32 s0, v7 +; CHECK-NEXT: v_readfirstlane_b32 s2, v6 +; CHECK-NEXT: v_mov_b32_e32 v4, s10 +; CHECK-NEXT: v_mov_b32_e32 v5, s8 ; CHECK-NEXT: s_mov_b32 s1, s0 ; CHECK-NEXT: s_mov_b32 s3, s2 ; CHECK-NEXT: flat_store_dwordx4 v[0:1], v[2:5] @@ -246,24 +233,24 @@ define amdgpu_ps <2 x i64> @s_uadd_v2i64(<2 x i64> inreg %val0, <2 x i64> inreg define amdgpu_ps <2 x i64> @s_usub_v2i64(<2 x i64> inreg %val0, <2 x i64> inreg %val1, ptr %ptrval) { ; CHECK-LABEL: s_usub_v2i64: ; CHECK: ; %bb.0: -; CHECK-NEXT: s_sub_u32 s6, s2, s6 -; CHECK-NEXT: v_mov_b32_e32 v9, s3 -; CHECK-NEXT: s_subb_u32 s7, s3, s7 -; CHECK-NEXT: v_mov_b32_e32 v8, s2 -; CHECK-NEXT: s_sub_u32 s4, s0, s4 -; CHECK-NEXT: v_mov_b32_e32 v7, s1 -; CHECK-NEXT: v_cmp_gt_u64_e32 vcc, s[6:7], v[8:9] -; CHECK-NEXT: s_subb_u32 s5, s1, s5 -; CHECK-NEXT: v_mov_b32_e32 v6, s0 -; CHECK-NEXT: v_cndmask_b32_e64 v8, 0, -1, vcc -; CHECK-NEXT: v_cmp_gt_u64_e32 vcc, s[4:5], v[6:7] -; CHECK-NEXT: v_readfirstlane_b32 s2, v8 -; CHECK-NEXT: v_cndmask_b32_e64 v6, 0, -1, vcc -; CHECK-NEXT: v_readfirstlane_b32 s0, v6 -; CHECK-NEXT: v_mov_b32_e32 v2, s4 -; CHECK-NEXT: v_mov_b32_e32 v3, s5 -; CHECK-NEXT: v_mov_b32_e32 v4, s6 -; CHECK-NEXT: v_mov_b32_e32 v5, s7 +; CHECK-NEXT: s_sub_u32 s10, s2, s6 +; CHECK-NEXT: s_cselect_b64 s[8:9], -1, 0 +; CHECK-NEXT: s_cmp_lg_u64 s[8:9], 0 +; CHECK-NEXT: s_subb_u32 s8, s3, s7 +; CHECK-NEXT: s_cselect_b64 s[2:3], -1, 0 +; CHECK-NEXT: s_sub_u32 s0, s0, s4 +; CHECK-NEXT: s_cselect_b64 s[6:7], -1, 0 +; CHECK-NEXT: s_cmp_lg_u64 s[6:7], 0 +; CHECK-NEXT: s_subb_u32 s1, s1, s5 +; CHECK-NEXT: v_mov_b32_e32 v2, s0 +; CHECK-NEXT: v_mov_b32_e32 v3, s1 +; CHECK-NEXT: s_cselect_b64 s[0:1], -1, 0 +; CHECK-NEXT: v_cndmask_b32_e64 v6, 0, -1, s[2:3] +; CHECK-NEXT: v_cndmask_b32_e64 v7, 0, -1, s[0:1] +; CHECK-NEXT: v_readfirstlane_b32 s0, v7 +; CHECK-NEXT: v_readfirstlane_b32 s2, v6 +; CHECK-NEXT: v_mov_b32_e32 v4, s10 +; CHECK-NEXT: v_mov_b32_e32 v5, s8 ; CHECK-NEXT: s_mov_b32 s1, s0 ; CHECK-NEXT: s_mov_b32 s3, s2 ; CHECK-NEXT: flat_store_dwordx4 v[0:1], v[2:5] @@ -280,15 +267,15 @@ define amdgpu_ps <2 x i64> @s_usub_v2i64(<2 x i64> inreg %val0, <2 x i64> inreg define amdgpu_ps i64 @s_uadd_i64(i64 inreg %val0, i64 inreg %val1, ptr %ptrval) { ; CHECK-LABEL: s_uadd_i64: ; CHECK: ; %bb.0: -; CHECK-NEXT: s_add_u32 s2, s0, s2 -; CHECK-NEXT: v_mov_b32_e32 v3, s1 -; CHECK-NEXT: s_addc_u32 s3, s1, s3 +; CHECK-NEXT: s_add_u32 s0, s0, s2 +; CHECK-NEXT: s_cselect_b64 s[4:5], -1, 0 +; CHECK-NEXT: s_cmp_lg_u64 s[4:5], 0 +; CHECK-NEXT: s_addc_u32 s1, s1, s3 ; CHECK-NEXT: v_mov_b32_e32 v2, s0 -; CHECK-NEXT: v_mov_b32_e32 v5, s3 -; CHECK-NEXT: v_cmp_lt_u64_e32 vcc, s[2:3], v[2:3] -; CHECK-NEXT: v_mov_b32_e32 v4, s2 -; CHECK-NEXT: flat_store_dwordx2 v[0:1], v[4:5] -; CHECK-NEXT: v_cndmask_b32_e64 v0, 0, -1, vcc +; CHECK-NEXT: v_mov_b32_e32 v3, s1 +; CHECK-NEXT: s_cselect_b64 s[0:1], -1, 0 +; CHECK-NEXT: flat_store_dwordx2 v[0:1], v[2:3] +; CHECK-NEXT: v_cndmask_b32_e64 v0, 0, -1, s[0:1] ; CHECK-NEXT: v_readfirstlane_b32 s0, v0 ; CHECK-NEXT: s_mov_b32 s1, s0 ; CHECK-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) @@ -305,10 +292,11 @@ define amdgpu_ps i64 @s_uadd_p1(i64 inreg %val0, i64 inreg %val1, ptr %ptrval) { ; CHECK-LABEL: s_uadd_p1: ; CHECK: ; %bb.0: ; CHECK-NEXT: s_add_u32 s0, s0, 1 +; CHECK-NEXT: s_cselect_b64 s[2:3], -1, 0 +; CHECK-NEXT: s_cmp_lg_u64 s[2:3], 0 ; CHECK-NEXT: s_addc_u32 s1, s1, 0 -; CHECK-NEXT: s_cmp_eq_u64 s[0:1], 0 -; CHECK-NEXT: v_mov_b32_e32 v3, s1 ; CHECK-NEXT: v_mov_b32_e32 v2, s0 +; CHECK-NEXT: v_mov_b32_e32 v3, s1 ; CHECK-NEXT: s_cselect_b64 s[0:1], -1, 0 ; CHECK-NEXT: flat_store_dwordx2 v[0:1], v[2:3] ; CHECK-NEXT: v_cndmask_b32_e64 v0, 0, -1, s[0:1] @@ -350,15 +338,15 @@ define amdgpu_ps i64 @s_uadd_n1(i64 inreg %val0, i64 inreg %val1, ptr %ptrval) { define amdgpu_ps i64 @s_usub_p1(i64 inreg %val0, i64 inreg %val1, ptr %ptrval) { ; CHECK-LABEL: s_usub_p1: ; CHECK: ; %bb.0: -; CHECK-NEXT: s_add_u32 s2, s0, -1 -; CHECK-NEXT: v_mov_b32_e32 v3, s1 -; CHECK-NEXT: s_addc_u32 s3, s1, -1 +; CHECK-NEXT: s_sub_u32 s0, s0, 1 +; CHECK-NEXT: s_cselect_b64 s[2:3], -1, 0 +; CHECK-NEXT: s_cmp_lg_u64 s[2:3], 0 +; CHECK-NEXT: s_subb_u32 s1, s1, 0 ; CHECK-NEXT: v_mov_b32_e32 v2, s0 -; CHECK-NEXT: v_mov_b32_e32 v5, s3 -; CHECK-NEXT: v_cmp_gt_u64_e32 vcc, s[2:3], v[2:3] -; CHECK-NEXT: v_mov_b32_e32 v4, s2 -; CHECK-NEXT: flat_store_dwordx2 v[0:1], v[4:5] -; CHECK-NEXT: v_cndmask_b32_e64 v0, 0, -1, vcc +; CHECK-NEXT: v_mov_b32_e32 v3, s1 +; CHECK-NEXT: s_cselect_b64 s[0:1], -1, 0 +; CHECK-NEXT: flat_store_dwordx2 v[0:1], v[2:3] +; CHECK-NEXT: v_cndmask_b32_e64 v0, 0, -1, s[0:1] ; CHECK-NEXT: v_readfirstlane_b32 s0, v0 ; CHECK-NEXT: s_mov_b32 s1, s0 ; CHECK-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) @@ -374,15 +362,15 @@ define amdgpu_ps i64 @s_usub_p1(i64 inreg %val0, i64 inreg %val1, ptr %ptrval) { define amdgpu_ps i64 @s_usub_n1(i64 inreg %val0, i64 inreg %val1, ptr %ptrval) { ; CHECK-LABEL: s_usub_n1: ; CHECK: ; %bb.0: -; CHECK-NEXT: s_add_u32 s2, s0, 1 -; CHECK-NEXT: v_mov_b32_e32 v3, s1 -; CHECK-NEXT: s_addc_u32 s3, s1, 0 +; CHECK-NEXT: s_sub_u32 s0, s0, -1 +; CHECK-NEXT: s_cselect_b64 s[2:3], -1, 0 +; CHECK-NEXT: s_cmp_lg_u64 s[2:3], 0 +; CHECK-NEXT: s_subb_u32 s1, s1, -1 ; CHECK-NEXT: v_mov_b32_e32 v2, s0 -; CHECK-NEXT: v_mov_b32_e32 v5, s3 -; CHECK-NEXT: v_cmp_gt_u64_e32 vcc, s[2:3], v[2:3] -; CHECK-NEXT: v_mov_b32_e32 v4, s2 -; CHECK-NEXT: flat_store_dwordx2 v[0:1], v[4:5] -; CHECK-NEXT: v_cndmask_b32_e64 v0, 0, -1, vcc +; CHECK-NEXT: v_mov_b32_e32 v3, s1 +; CHECK-NEXT: s_cselect_b64 s[0:1], -1, 0 +; CHECK-NEXT: flat_store_dwordx2 v[0:1], v[2:3] +; CHECK-NEXT: v_cndmask_b32_e64 v0, 0, -1, s[0:1] ; CHECK-NEXT: v_readfirstlane_b32 s0, v0 ; CHECK-NEXT: s_mov_b32 s1, s0 ; CHECK-NEXT: s_waitcnt vmcnt(0) lgkmcnt(0) diff --git a/llvm/test/CodeGen/AMDGPU/carryout-selection.ll b/llvm/test/CodeGen/AMDGPU/carryout-selection.ll index 51652a09863e0..d484111596e8b 100644 --- a/llvm/test/CodeGen/AMDGPU/carryout-selection.ll +++ b/llvm/test/CodeGen/AMDGPU/carryout-selection.ll @@ -691,7 +691,8 @@ define amdgpu_kernel void @uaddo32_vcc_user(ptr addrspace(1) %out, ptr addrspace ; GCN-ISEL-LABEL: name: suaddo64 ; GCN-ISEL-LABEL: body: ; GCN-ISEL-LABEL: bb.0 -; GCN-ISEL: S_ADD_U64_PSEUDO +; GCN-ISEL: S_UADDO_PSEUDO +; GCN-ISEL: S_ADD_CO_PSEUDO define amdgpu_kernel void @suaddo64(ptr addrspace(1) %out, ptr addrspace(1) %carryout, i64 %a, i64 %b) #0 { ; CISI-LABEL: suaddo64: @@ -700,21 +701,23 @@ define amdgpu_kernel void @suaddo64(ptr addrspace(1) %out, ptr addrspace(1) %car ; CISI-NEXT: s_mov_b32 s11, 0xf000 ; CISI-NEXT: s_mov_b32 s10, -1 ; CISI-NEXT: s_waitcnt lgkmcnt(0) -; CISI-NEXT: s_add_u32 s6, s4, s6 -; CISI-NEXT: v_mov_b32_e32 v0, s4 -; CISI-NEXT: s_addc_u32 s7, s5, s7 -; CISI-NEXT: v_mov_b32_e32 v1, s5 -; CISI-NEXT: v_cmp_lt_u64_e32 vcc, s[6:7], v[0:1] -; CISI-NEXT: v_mov_b32_e32 v2, s6 +; CISI-NEXT: s_add_u32 s4, s4, s6 +; CISI-NEXT: s_cselect_b64 s[12:13], -1, 0 +; CISI-NEXT: s_or_b32 s6, s12, s13 +; CISI-NEXT: s_cmp_lg_u32 s6, 0 +; CISI-NEXT: s_addc_u32 s5, s5, s7 ; CISI-NEXT: s_mov_b32 s8, s0 ; CISI-NEXT: s_mov_b32 s9, s1 +; CISI-NEXT: v_mov_b32_e32 v0, s4 +; CISI-NEXT: v_mov_b32_e32 v1, s5 +; CISI-NEXT: s_cselect_b64 s[4:5], -1, 0 ; CISI-NEXT: s_mov_b32 s0, s2 ; CISI-NEXT: s_mov_b32 s1, s3 ; CISI-NEXT: s_mov_b32 s2, s10 ; CISI-NEXT: s_mov_b32 s3, s11 -; CISI-NEXT: v_mov_b32_e32 v3, s7 -; CISI-NEXT: v_cndmask_b32_e64 v0, 0, 1, vcc -; CISI-NEXT: buffer_store_dwordx2 v[2:3], off, s[8:11], 0 +; CISI-NEXT: buffer_store_dwordx2 v[0:1], off, s[8:11], 0 +; CISI-NEXT: s_waitcnt expcnt(0) +; CISI-NEXT: v_cndmask_b32_e64 v0, 0, 1, s[4:5] ; CISI-NEXT: buffer_store_byte v0, off, s[0:3], 0 ; CISI-NEXT: s_endpgm ; @@ -722,37 +725,37 @@ define amdgpu_kernel void @suaddo64(ptr addrspace(1) %out, ptr addrspace(1) %car ; VI: ; %bb.0: ; VI-NEXT: s_load_dwordx8 s[0:7], s[4:5], 0x24 ; VI-NEXT: s_waitcnt lgkmcnt(0) +; VI-NEXT: v_mov_b32_e32 v2, s2 +; VI-NEXT: s_add_u32 s2, s4, s6 ; VI-NEXT: v_mov_b32_e32 v0, s0 -; VI-NEXT: s_add_u32 s0, s4, s6 -; VI-NEXT: v_mov_b32_e32 v4, s4 ; VI-NEXT: v_mov_b32_e32 v1, s1 -; VI-NEXT: s_addc_u32 s1, s5, s7 -; VI-NEXT: v_mov_b32_e32 v5, s5 -; VI-NEXT: v_mov_b32_e32 v7, s1 -; VI-NEXT: v_cmp_lt_u64_e32 vcc, s[0:1], v[4:5] -; VI-NEXT: v_mov_b32_e32 v6, s0 -; VI-NEXT: v_mov_b32_e32 v2, s2 +; VI-NEXT: s_cselect_b64 s[0:1], -1, 0 +; VI-NEXT: s_cmp_lg_u64 s[0:1], 0 +; VI-NEXT: s_addc_u32 s0, s5, s7 +; VI-NEXT: v_mov_b32_e32 v4, s2 +; VI-NEXT: v_mov_b32_e32 v5, s0 +; VI-NEXT: s_cselect_b64 s[0:1], -1, 0 ; VI-NEXT: v_mov_b32_e32 v3, s3 -; VI-NEXT: flat_store_dwordx2 v[0:1], v[6:7] -; VI-NEXT: v_cndmask_b32_e64 v0, 0, 1, vcc +; VI-NEXT: flat_store_dwordx2 v[0:1], v[4:5] +; VI-NEXT: v_cndmask_b32_e64 v0, 0, 1, s[0:1] ; VI-NEXT: flat_store_byte v[2:3], v0 ; VI-NEXT: s_endpgm ; ; GFX9-LABEL: suaddo64: ; GFX9: ; %bb.0: ; GFX9-NEXT: s_load_dwordx8 s[8:15], s[4:5], 0x24 -; GFX9-NEXT: v_mov_b32_e32 v4, 0 +; GFX9-NEXT: v_mov_b32_e32 v2, 0 ; GFX9-NEXT: s_waitcnt lgkmcnt(0) -; GFX9-NEXT: s_add_u32 s0, s12, s14 -; GFX9-NEXT: v_mov_b32_e32 v0, s12 -; GFX9-NEXT: v_mov_b32_e32 v1, s13 -; GFX9-NEXT: s_addc_u32 s1, s13, s15 -; GFX9-NEXT: v_mov_b32_e32 v3, s1 -; GFX9-NEXT: v_cmp_lt_u64_e32 vcc, s[0:1], v[0:1] -; GFX9-NEXT: v_mov_b32_e32 v2, s0 -; GFX9-NEXT: v_cndmask_b32_e64 v0, 0, 1, vcc -; GFX9-NEXT: global_store_dwordx2 v4, v[2:3], s[8:9] -; GFX9-NEXT: global_store_byte v4, v0, s[10:11] +; GFX9-NEXT: s_add_u32 s2, s12, s14 +; GFX9-NEXT: s_cselect_b64 s[0:1], -1, 0 +; GFX9-NEXT: s_cmp_lg_u64 s[0:1], 0 +; GFX9-NEXT: s_addc_u32 s0, s13, s15 +; GFX9-NEXT: v_mov_b32_e32 v0, s2 +; GFX9-NEXT: v_mov_b32_e32 v1, s0 +; GFX9-NEXT: s_cselect_b64 s[0:1], -1, 0 +; GFX9-NEXT: v_cndmask_b32_e64 v3, 0, 1, s[0:1] +; GFX9-NEXT: global_store_dwordx2 v2, v[0:1], s[8:9] +; GFX9-NEXT: global_store_byte v2, v3, s[10:11] ; GFX9-NEXT: s_endpgm ; ; GFX1010-LABEL: suaddo64: @@ -761,10 +764,12 @@ define amdgpu_kernel void @suaddo64(ptr addrspace(1) %out, ptr addrspace(1) %car ; GFX1010-NEXT: v_mov_b32_e32 v2, 0 ; GFX1010-NEXT: s_waitcnt lgkmcnt(0) ; GFX1010-NEXT: s_add_u32 s0, s12, s14 -; GFX1010-NEXT: s_addc_u32 s1, s13, s15 +; GFX1010-NEXT: s_cselect_b32 s1, -1, 0 ; GFX1010-NEXT: v_mov_b32_e32 v0, s0 +; GFX1010-NEXT: s_cmp_lg_u32 s1, 0 +; GFX1010-NEXT: s_addc_u32 s1, s13, s15 +; GFX1010-NEXT: s_cselect_b32 s0, -1, 0 ; GFX1010-NEXT: v_mov_b32_e32 v1, s1 -; GFX1010-NEXT: v_cmp_lt_u64_e64 s0, s[0:1], s[12:13] ; GFX1010-NEXT: v_cndmask_b32_e64 v3, 0, 1, s0 ; GFX1010-NEXT: global_store_dwordx2 v2, v[0:1], s[8:9] ; GFX1010-NEXT: global_store_byte v2, v3, s[10:11] @@ -775,11 +780,13 @@ define amdgpu_kernel void @suaddo64(ptr addrspace(1) %out, ptr addrspace(1) %car ; GFX1030W32-NEXT: s_load_dwordx8 s[0:7], s[4:5], 0x24 ; GFX1030W32-NEXT: v_mov_b32_e32 v2, 0 ; GFX1030W32-NEXT: s_waitcnt lgkmcnt(0) -; GFX1030W32-NEXT: s_add_u32 s6, s4, s6 -; GFX1030W32-NEXT: s_addc_u32 s7, s5, s7 -; GFX1030W32-NEXT: v_mov_b32_e32 v0, s6 -; GFX1030W32-NEXT: v_cmp_lt_u64_e64 s4, s[6:7], s[4:5] -; GFX1030W32-NEXT: v_mov_b32_e32 v1, s7 +; GFX1030W32-NEXT: s_add_u32 s4, s4, s6 +; GFX1030W32-NEXT: s_cselect_b32 s6, -1, 0 +; GFX1030W32-NEXT: v_mov_b32_e32 v0, s4 +; GFX1030W32-NEXT: s_cmp_lg_u32 s6, 0 +; GFX1030W32-NEXT: s_addc_u32 s5, s5, s7 +; GFX1030W32-NEXT: s_cselect_b32 s4, -1, 0 +; GFX1030W32-NEXT: v_mov_b32_e32 v1, s5 ; GFX1030W32-NEXT: v_cndmask_b32_e64 v3, 0, 1, s4 ; GFX1030W32-NEXT: global_store_dwordx2 v2, v[0:1], s[0:1] ; GFX1030W32-NEXT: global_store_byte v2, v3, s[2:3] @@ -790,11 +797,13 @@ define amdgpu_kernel void @suaddo64(ptr addrspace(1) %out, ptr addrspace(1) %car ; GFX1030W64-NEXT: s_load_dwordx8 s[0:7], s[4:5], 0x24 ; GFX1030W64-NEXT: v_mov_b32_e32 v2, 0 ; GFX1030W64-NEXT: s_waitcnt lgkmcnt(0) -; GFX1030W64-NEXT: s_add_u32 s6, s4, s6 -; GFX1030W64-NEXT: s_addc_u32 s7, s5, s7 -; GFX1030W64-NEXT: v_mov_b32_e32 v0, s6 -; GFX1030W64-NEXT: v_cmp_lt_u64_e64 s[4:5], s[6:7], s[4:5] -; GFX1030W64-NEXT: v_mov_b32_e32 v1, s7 +; GFX1030W64-NEXT: s_add_u32 s4, s4, s6 +; GFX1030W64-NEXT: s_cselect_b64 s[8:9], -1, 0 +; GFX1030W64-NEXT: v_mov_b32_e32 v0, s4 +; GFX1030W64-NEXT: s_cmp_lg_u64 s[8:9], 0 +; GFX1030W64-NEXT: s_addc_u32 s5, s5, s7 +; GFX1030W64-NEXT: v_mov_b32_e32 v1, s5 +; GFX1030W64-NEXT: s_cselect_b64 s[4:5], -1, 0 ; GFX1030W64-NEXT: v_cndmask_b32_e64 v3, 0, 1, s[4:5] ; GFX1030W64-NEXT: global_store_dwordx2 v2, v[0:1], s[0:1] ; GFX1030W64-NEXT: global_store_byte v2, v3, s[2:3] @@ -804,12 +813,13 @@ define amdgpu_kernel void @suaddo64(ptr addrspace(1) %out, ptr addrspace(1) %car ; GFX11: ; %bb.0: ; GFX11-NEXT: s_load_b256 s[0:7], s[4:5], 0x24 ; GFX11-NEXT: s_waitcnt lgkmcnt(0) -; GFX11-NEXT: s_add_u32 s6, s4, s6 -; GFX11-NEXT: s_addc_u32 s7, s5, s7 -; GFX11-NEXT: v_mov_b32_e32 v0, s6 -; GFX11-NEXT: v_cmp_lt_u64_e64 s4, s[6:7], s[4:5] -; GFX11-NEXT: v_dual_mov_b32 v2, 0 :: v_dual_mov_b32 v1, s7 -; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_2) +; GFX11-NEXT: s_add_u32 s4, s4, s6 +; GFX11-NEXT: s_cselect_b32 s6, -1, 0 +; GFX11-NEXT: v_mov_b32_e32 v0, s4 +; GFX11-NEXT: s_cmp_lg_u32 s6, 0 +; GFX11-NEXT: s_addc_u32 s5, s5, s7 +; GFX11-NEXT: s_cselect_b32 s4, -1, 0 +; GFX11-NEXT: v_dual_mov_b32 v2, 0 :: v_dual_mov_b32 v1, s5 ; GFX11-NEXT: v_cndmask_b32_e64 v3, 0, 1, s4 ; GFX11-NEXT: s_clause 0x1 ; GFX11-NEXT: global_store_b64 v2, v[0:1], s[0:1] @@ -819,12 +829,14 @@ define amdgpu_kernel void @suaddo64(ptr addrspace(1) %out, ptr addrspace(1) %car ; GFX1250-LABEL: suaddo64: ; GFX1250: ; %bb.0: ; GFX1250-NEXT: s_load_b256 s[0:7], s[4:5], 0x24 -; GFX1250-NEXT: v_mov_b32_e32 v2, 0 ; GFX1250-NEXT: s_wait_kmcnt 0x0 -; GFX1250-NEXT: s_add_nc_u64 s[6:7], s[4:5], s[6:7] -; GFX1250-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_1) | instid1(VALU_DEP_2) -; GFX1250-NEXT: v_cmp_lt_u64_e64 s4, s[6:7], s[4:5] -; GFX1250-NEXT: v_mov_b64_e32 v[0:1], s[6:7] +; GFX1250-NEXT: s_add_co_u32 s4, s4, s6 +; GFX1250-NEXT: s_cselect_b32 s6, -1, 0 +; GFX1250-NEXT: v_dual_mov_b32 v2, 0 :: v_dual_mov_b32 v0, s4 +; GFX1250-NEXT: s_cmp_lg_u32 s6, 0 +; GFX1250-NEXT: s_add_co_ci_u32 s5, s5, s7 +; GFX1250-NEXT: s_cselect_b32 s4, -1, 0 +; GFX1250-NEXT: v_mov_b32_e32 v1, s5 ; GFX1250-NEXT: v_cndmask_b32_e64 v3, 0, 1, s4 ; GFX1250-NEXT: s_clause 0x1 ; GFX1250-NEXT: global_store_b64 v2, v[0:1], s[0:1] @@ -841,7 +853,8 @@ define amdgpu_kernel void @suaddo64(ptr addrspace(1) %out, ptr addrspace(1) %car ; GCN-ISEL-LABEL: name: vuaddo64 ; GCN-ISEL-LABEL: body: ; GCN-ISEL-LABEL: bb.0 -; GCN-ISEL: V_ADD_U64_PSEUDO +; GCN-ISEL: V_ADD_CO_U32_e64 +; GCN-ISEL: V_ADDC_U32_e64 define amdgpu_kernel void @vuaddo64(ptr addrspace(1) %out, ptr addrspace(1) %carryout, i64 %a) #0 { ; CISI-LABEL: vuaddo64: @@ -854,9 +867,8 @@ define amdgpu_kernel void @vuaddo64(ptr addrspace(1) %out, ptr addrspace(1) %car ; CISI-NEXT: s_mov_b32 s4, s0 ; CISI-NEXT: v_mov_b32_e32 v1, s9 ; CISI-NEXT: v_add_i32_e32 v0, vcc, s8, v0 -; CISI-NEXT: v_addc_u32_e32 v1, vcc, 0, v1, vcc -; CISI-NEXT: v_cmp_gt_u64_e32 vcc, s[8:9], v[0:1] ; CISI-NEXT: s_mov_b32 s5, s1 +; CISI-NEXT: v_addc_u32_e32 v1, vcc, 0, v1, vcc ; CISI-NEXT: s_mov_b32 s0, s2 ; CISI-NEXT: s_mov_b32 s1, s3 ; CISI-NEXT: s_mov_b32 s2, s6 @@ -876,7 +888,6 @@ define amdgpu_kernel void @vuaddo64(ptr addrspace(1) %out, ptr addrspace(1) %car ; VI-NEXT: v_mov_b32_e32 v6, s5 ; VI-NEXT: v_add_u32_e32 v5, vcc, s4, v0 ; VI-NEXT: v_addc_u32_e32 v6, vcc, 0, v6, vcc -; VI-NEXT: v_cmp_gt_u64_e32 vcc, s[4:5], v[5:6] ; VI-NEXT: v_mov_b32_e32 v2, s1 ; VI-NEXT: v_mov_b32_e32 v3, s2 ; VI-NEXT: v_mov_b32_e32 v4, s3 @@ -894,7 +905,6 @@ define amdgpu_kernel void @vuaddo64(ptr addrspace(1) %out, ptr addrspace(1) %car ; GFX9-NEXT: v_mov_b32_e32 v1, s7 ; GFX9-NEXT: v_add_co_u32_e32 v0, vcc, s6, v0 ; GFX9-NEXT: v_addc_co_u32_e32 v1, vcc, 0, v1, vcc -; GFX9-NEXT: v_cmp_gt_u64_e32 vcc, s[6:7], v[0:1] ; GFX9-NEXT: global_store_dwordx2 v2, v[0:1], s[0:1] ; GFX9-NEXT: v_cndmask_b32_e64 v0, 0, 1, vcc ; GFX9-NEXT: global_store_byte v2, v0, s[2:3] @@ -909,8 +919,7 @@ define amdgpu_kernel void @vuaddo64(ptr addrspace(1) %out, ptr addrspace(1) %car ; GFX1010-NEXT: s_waitcnt lgkmcnt(0) ; GFX1010-NEXT: v_add_co_u32 v0, s4, s6, v0 ; GFX1010-NEXT: v_add_co_ci_u32_e64 v1, s4, s7, 0, s4 -; GFX1010-NEXT: v_cmp_gt_u64_e32 vcc_lo, s[6:7], v[0:1] -; GFX1010-NEXT: v_cndmask_b32_e64 v3, 0, 1, vcc_lo +; GFX1010-NEXT: v_cndmask_b32_e64 v3, 0, 1, s4 ; GFX1010-NEXT: global_store_dwordx2 v2, v[0:1], s[0:1] ; GFX1010-NEXT: global_store_byte v2, v3, s[2:3] ; GFX1010-NEXT: s_endpgm @@ -923,9 +932,8 @@ define amdgpu_kernel void @vuaddo64(ptr addrspace(1) %out, ptr addrspace(1) %car ; GFX1030W32-NEXT: v_mov_b32_e32 v2, 0 ; GFX1030W32-NEXT: s_waitcnt lgkmcnt(0) ; GFX1030W32-NEXT: v_add_co_u32 v0, s4, s6, v0 -; GFX1030W32-NEXT: v_add_co_ci_u32_e64 v1, null, s7, 0, s4 -; GFX1030W32-NEXT: v_cmp_gt_u64_e32 vcc_lo, s[6:7], v[0:1] -; GFX1030W32-NEXT: v_cndmask_b32_e64 v3, 0, 1, vcc_lo +; GFX1030W32-NEXT: v_add_co_ci_u32_e64 v1, s4, s7, 0, s4 +; GFX1030W32-NEXT: v_cndmask_b32_e64 v3, 0, 1, s4 ; GFX1030W32-NEXT: global_store_dwordx2 v2, v[0:1], s[0:1] ; GFX1030W32-NEXT: global_store_byte v2, v3, s[2:3] ; GFX1030W32-NEXT: s_endpgm @@ -938,9 +946,8 @@ define amdgpu_kernel void @vuaddo64(ptr addrspace(1) %out, ptr addrspace(1) %car ; GFX1030W64-NEXT: v_mov_b32_e32 v2, 0 ; GFX1030W64-NEXT: s_waitcnt lgkmcnt(0) ; GFX1030W64-NEXT: v_add_co_u32 v0, s[4:5], s6, v0 -; GFX1030W64-NEXT: v_add_co_ci_u32_e64 v1, null, s7, 0, s[4:5] -; GFX1030W64-NEXT: v_cmp_gt_u64_e32 vcc, s[6:7], v[0:1] -; GFX1030W64-NEXT: v_cndmask_b32_e64 v3, 0, 1, vcc +; GFX1030W64-NEXT: v_add_co_ci_u32_e64 v1, s[4:5], s7, 0, s[4:5] +; GFX1030W64-NEXT: v_cndmask_b32_e64 v3, 0, 1, s[4:5] ; GFX1030W64-NEXT: global_store_dwordx2 v2, v[0:1], s[0:1] ; GFX1030W64-NEXT: global_store_byte v2, v3, s[2:3] ; GFX1030W64-NEXT: s_endpgm @@ -955,10 +962,9 @@ define amdgpu_kernel void @vuaddo64(ptr addrspace(1) %out, ptr addrspace(1) %car ; GFX11-NEXT: s_waitcnt lgkmcnt(0) ; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1) ; GFX11-NEXT: v_add_co_u32 v0, s4, s6, v0 -; GFX11-NEXT: v_add_co_ci_u32_e64 v1, null, s7, 0, s4 +; GFX11-NEXT: v_add_co_ci_u32_e64 v1, s4, s7, 0, s4 ; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) -; GFX11-NEXT: v_cmp_gt_u64_e32 vcc_lo, s[6:7], v[0:1] -; GFX11-NEXT: v_cndmask_b32_e64 v3, 0, 1, vcc_lo +; GFX11-NEXT: v_cndmask_b32_e64 v3, 0, 1, s4 ; GFX11-NEXT: s_clause 0x1 ; GFX11-NEXT: global_store_b64 v2, v[0:1], s[0:1] ; GFX11-NEXT: global_store_b8 v2, v3, s[2:3] @@ -969,16 +975,17 @@ define amdgpu_kernel void @vuaddo64(ptr addrspace(1) %out, ptr addrspace(1) %car ; GFX1250-NEXT: s_clause 0x1 ; GFX1250-NEXT: s_load_b64 s[6:7], s[4:5], 0x34 ; GFX1250-NEXT: s_load_b128 s[0:3], s[4:5], 0x24 -; GFX1250-NEXT: v_mov_b32_e32 v1, 0 ; GFX1250-NEXT: v_and_b32_e32 v0, 0x3ff, v0 +; GFX1250-NEXT: v_mov_b32_e32 v2, 0 ; GFX1250-NEXT: s_wait_kmcnt 0x0 -; GFX1250-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1) -; GFX1250-NEXT: v_add_nc_u64_e32 v[2:3], s[6:7], v[0:1] -; GFX1250-NEXT: v_cmp_gt_u64_e32 vcc_lo, s[6:7], v[2:3] -; GFX1250-NEXT: v_cndmask_b32_e64 v0, 0, 1, vcc_lo +; GFX1250-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1) +; GFX1250-NEXT: v_add_co_u32 v0, s4, s6, v0 +; GFX1250-NEXT: v_add_co_ci_u32_e64 v1, s4, s7, 0, s4 +; GFX1250-NEXT: s_delay_alu instid0(VALU_DEP_1) +; GFX1250-NEXT: v_cndmask_b32_e64 v3, 0, 1, s4 ; GFX1250-NEXT: s_clause 0x1 -; GFX1250-NEXT: global_store_b64 v1, v[2:3], s[0:1] -; GFX1250-NEXT: global_store_b8 v1, v0, s[2:3] +; GFX1250-NEXT: global_store_b64 v2, v[0:1], s[0:1] +; GFX1250-NEXT: global_store_b8 v2, v3, s[2:3] ; GFX1250-NEXT: s_endpgm %tid = call i32 @llvm.amdgcn.workitem.id.x() %tid.ext = sext i32 %tid to i64 @@ -1671,7 +1678,8 @@ define amdgpu_kernel void @usubo32_vcc_user(ptr addrspace(1) %out, ptr addrspace ; GCN-ISEL-LABEL: name: susubo64 ; GCN-ISEL-LABEL: body: ; GCN-ISEL-LABEL: bb.0 -; GCN-ISEL: S_SUB_U64_PSEUDO +; GCN-ISEL: S_USUBO_PSEUDO +; GCN-ISEL: S_SUB_CO_PSEUDO define amdgpu_kernel void @susubo64(ptr addrspace(1) %out, ptr addrspace(1) %carryout, i64 %a, i64 %b) #0 { ; CISI-LABEL: susubo64: @@ -1680,21 +1688,23 @@ define amdgpu_kernel void @susubo64(ptr addrspace(1) %out, ptr addrspace(1) %car ; CISI-NEXT: s_mov_b32 s11, 0xf000 ; CISI-NEXT: s_mov_b32 s10, -1 ; CISI-NEXT: s_waitcnt lgkmcnt(0) -; CISI-NEXT: s_sub_u32 s6, s4, s6 -; CISI-NEXT: v_mov_b32_e32 v0, s4 -; CISI-NEXT: s_subb_u32 s7, s5, s7 -; CISI-NEXT: v_mov_b32_e32 v1, s5 -; CISI-NEXT: v_cmp_gt_u64_e32 vcc, s[6:7], v[0:1] -; CISI-NEXT: v_mov_b32_e32 v2, s6 +; CISI-NEXT: s_sub_u32 s4, s4, s6 +; CISI-NEXT: s_cselect_b64 s[12:13], -1, 0 +; CISI-NEXT: s_or_b32 s6, s12, s13 +; CISI-NEXT: s_cmp_lg_u32 s6, 0 +; CISI-NEXT: s_subb_u32 s5, s5, s7 ; CISI-NEXT: s_mov_b32 s8, s0 ; CISI-NEXT: s_mov_b32 s9, s1 +; CISI-NEXT: v_mov_b32_e32 v0, s4 +; CISI-NEXT: v_mov_b32_e32 v1, s5 +; CISI-NEXT: s_cselect_b64 s[4:5], -1, 0 ; CISI-NEXT: s_mov_b32 s0, s2 ; CISI-NEXT: s_mov_b32 s1, s3 ; CISI-NEXT: s_mov_b32 s2, s10 ; CISI-NEXT: s_mov_b32 s3, s11 -; CISI-NEXT: v_mov_b32_e32 v3, s7 -; CISI-NEXT: v_cndmask_b32_e64 v0, 0, 1, vcc -; CISI-NEXT: buffer_store_dwordx2 v[2:3], off, s[8:11], 0 +; CISI-NEXT: buffer_store_dwordx2 v[0:1], off, s[8:11], 0 +; CISI-NEXT: s_waitcnt expcnt(0) +; CISI-NEXT: v_cndmask_b32_e64 v0, 0, 1, s[4:5] ; CISI-NEXT: buffer_store_byte v0, off, s[0:3], 0 ; CISI-NEXT: s_endpgm ; @@ -1702,37 +1712,37 @@ define amdgpu_kernel void @susubo64(ptr addrspace(1) %out, ptr addrspace(1) %car ; VI: ; %bb.0: ; VI-NEXT: s_load_dwordx8 s[0:7], s[4:5], 0x24 ; VI-NEXT: s_waitcnt lgkmcnt(0) +; VI-NEXT: v_mov_b32_e32 v2, s2 +; VI-NEXT: s_sub_u32 s2, s4, s6 ; VI-NEXT: v_mov_b32_e32 v0, s0 -; VI-NEXT: s_sub_u32 s0, s4, s6 -; VI-NEXT: v_mov_b32_e32 v4, s4 ; VI-NEXT: v_mov_b32_e32 v1, s1 -; VI-NEXT: s_subb_u32 s1, s5, s7 -; VI-NEXT: v_mov_b32_e32 v5, s5 -; VI-NEXT: v_mov_b32_e32 v7, s1 -; VI-NEXT: v_cmp_gt_u64_e32 vcc, s[0:1], v[4:5] -; VI-NEXT: v_mov_b32_e32 v6, s0 -; VI-NEXT: v_mov_b32_e32 v2, s2 +; VI-NEXT: s_cselect_b64 s[0:1], -1, 0 +; VI-NEXT: s_cmp_lg_u64 s[0:1], 0 +; VI-NEXT: s_subb_u32 s0, s5, s7 +; VI-NEXT: v_mov_b32_e32 v4, s2 +; VI-NEXT: v_mov_b32_e32 v5, s0 +; VI-NEXT: s_cselect_b64 s[0:1], -1, 0 ; VI-NEXT: v_mov_b32_e32 v3, s3 -; VI-NEXT: flat_store_dwordx2 v[0:1], v[6:7] -; VI-NEXT: v_cndmask_b32_e64 v0, 0, 1, vcc +; VI-NEXT: flat_store_dwordx2 v[0:1], v[4:5] +; VI-NEXT: v_cndmask_b32_e64 v0, 0, 1, s[0:1] ; VI-NEXT: flat_store_byte v[2:3], v0 ; VI-NEXT: s_endpgm ; ; GFX9-LABEL: susubo64: ; GFX9: ; %bb.0: ; GFX9-NEXT: s_load_dwordx8 s[8:15], s[4:5], 0x24 -; GFX9-NEXT: v_mov_b32_e32 v4, 0 +; GFX9-NEXT: v_mov_b32_e32 v2, 0 ; GFX9-NEXT: s_waitcnt lgkmcnt(0) -; GFX9-NEXT: s_sub_u32 s0, s12, s14 -; GFX9-NEXT: v_mov_b32_e32 v0, s12 -; GFX9-NEXT: v_mov_b32_e32 v1, s13 -; GFX9-NEXT: s_subb_u32 s1, s13, s15 -; GFX9-NEXT: v_mov_b32_e32 v3, s1 -; GFX9-NEXT: v_cmp_gt_u64_e32 vcc, s[0:1], v[0:1] -; GFX9-NEXT: v_mov_b32_e32 v2, s0 -; GFX9-NEXT: v_cndmask_b32_e64 v0, 0, 1, vcc -; GFX9-NEXT: global_store_dwordx2 v4, v[2:3], s[8:9] -; GFX9-NEXT: global_store_byte v4, v0, s[10:11] +; GFX9-NEXT: s_sub_u32 s2, s12, s14 +; GFX9-NEXT: s_cselect_b64 s[0:1], -1, 0 +; GFX9-NEXT: s_cmp_lg_u64 s[0:1], 0 +; GFX9-NEXT: s_subb_u32 s0, s13, s15 +; GFX9-NEXT: v_mov_b32_e32 v0, s2 +; GFX9-NEXT: v_mov_b32_e32 v1, s0 +; GFX9-NEXT: s_cselect_b64 s[0:1], -1, 0 +; GFX9-NEXT: v_cndmask_b32_e64 v3, 0, 1, s[0:1] +; GFX9-NEXT: global_store_dwordx2 v2, v[0:1], s[8:9] +; GFX9-NEXT: global_store_byte v2, v3, s[10:11] ; GFX9-NEXT: s_endpgm ; ; GFX1010-LABEL: susubo64: @@ -1741,10 +1751,12 @@ define amdgpu_kernel void @susubo64(ptr addrspace(1) %out, ptr addrspace(1) %car ; GFX1010-NEXT: v_mov_b32_e32 v2, 0 ; GFX1010-NEXT: s_waitcnt lgkmcnt(0) ; GFX1010-NEXT: s_sub_u32 s0, s12, s14 -; GFX1010-NEXT: s_subb_u32 s1, s13, s15 +; GFX1010-NEXT: s_cselect_b32 s1, -1, 0 ; GFX1010-NEXT: v_mov_b32_e32 v0, s0 +; GFX1010-NEXT: s_cmp_lg_u32 s1, 0 +; GFX1010-NEXT: s_subb_u32 s1, s13, s15 +; GFX1010-NEXT: s_cselect_b32 s0, -1, 0 ; GFX1010-NEXT: v_mov_b32_e32 v1, s1 -; GFX1010-NEXT: v_cmp_gt_u64_e64 s0, s[0:1], s[12:13] ; GFX1010-NEXT: v_cndmask_b32_e64 v3, 0, 1, s0 ; GFX1010-NEXT: global_store_dwordx2 v2, v[0:1], s[8:9] ; GFX1010-NEXT: global_store_byte v2, v3, s[10:11] @@ -1755,11 +1767,13 @@ define amdgpu_kernel void @susubo64(ptr addrspace(1) %out, ptr addrspace(1) %car ; GFX1030W32-NEXT: s_load_dwordx8 s[0:7], s[4:5], 0x24 ; GFX1030W32-NEXT: v_mov_b32_e32 v2, 0 ; GFX1030W32-NEXT: s_waitcnt lgkmcnt(0) -; GFX1030W32-NEXT: s_sub_u32 s6, s4, s6 -; GFX1030W32-NEXT: s_subb_u32 s7, s5, s7 -; GFX1030W32-NEXT: v_mov_b32_e32 v0, s6 -; GFX1030W32-NEXT: v_cmp_gt_u64_e64 s4, s[6:7], s[4:5] -; GFX1030W32-NEXT: v_mov_b32_e32 v1, s7 +; GFX1030W32-NEXT: s_sub_u32 s4, s4, s6 +; GFX1030W32-NEXT: s_cselect_b32 s6, -1, 0 +; GFX1030W32-NEXT: v_mov_b32_e32 v0, s4 +; GFX1030W32-NEXT: s_cmp_lg_u32 s6, 0 +; GFX1030W32-NEXT: s_subb_u32 s5, s5, s7 +; GFX1030W32-NEXT: s_cselect_b32 s4, -1, 0 +; GFX1030W32-NEXT: v_mov_b32_e32 v1, s5 ; GFX1030W32-NEXT: v_cndmask_b32_e64 v3, 0, 1, s4 ; GFX1030W32-NEXT: global_store_dwordx2 v2, v[0:1], s[0:1] ; GFX1030W32-NEXT: global_store_byte v2, v3, s[2:3] @@ -1770,11 +1784,13 @@ define amdgpu_kernel void @susubo64(ptr addrspace(1) %out, ptr addrspace(1) %car ; GFX1030W64-NEXT: s_load_dwordx8 s[0:7], s[4:5], 0x24 ; GFX1030W64-NEXT: v_mov_b32_e32 v2, 0 ; GFX1030W64-NEXT: s_waitcnt lgkmcnt(0) -; GFX1030W64-NEXT: s_sub_u32 s6, s4, s6 -; GFX1030W64-NEXT: s_subb_u32 s7, s5, s7 -; GFX1030W64-NEXT: v_mov_b32_e32 v0, s6 -; GFX1030W64-NEXT: v_cmp_gt_u64_e64 s[4:5], s[6:7], s[4:5] -; GFX1030W64-NEXT: v_mov_b32_e32 v1, s7 +; GFX1030W64-NEXT: s_sub_u32 s4, s4, s6 +; GFX1030W64-NEXT: s_cselect_b64 s[8:9], -1, 0 +; GFX1030W64-NEXT: v_mov_b32_e32 v0, s4 +; GFX1030W64-NEXT: s_cmp_lg_u64 s[8:9], 0 +; GFX1030W64-NEXT: s_subb_u32 s5, s5, s7 +; GFX1030W64-NEXT: v_mov_b32_e32 v1, s5 +; GFX1030W64-NEXT: s_cselect_b64 s[4:5], -1, 0 ; GFX1030W64-NEXT: v_cndmask_b32_e64 v3, 0, 1, s[4:5] ; GFX1030W64-NEXT: global_store_dwordx2 v2, v[0:1], s[0:1] ; GFX1030W64-NEXT: global_store_byte v2, v3, s[2:3] @@ -1784,12 +1800,13 @@ define amdgpu_kernel void @susubo64(ptr addrspace(1) %out, ptr addrspace(1) %car ; GFX11: ; %bb.0: ; GFX11-NEXT: s_load_b256 s[0:7], s[4:5], 0x24 ; GFX11-NEXT: s_waitcnt lgkmcnt(0) -; GFX11-NEXT: s_sub_u32 s6, s4, s6 -; GFX11-NEXT: s_subb_u32 s7, s5, s7 -; GFX11-NEXT: v_mov_b32_e32 v0, s6 -; GFX11-NEXT: v_cmp_gt_u64_e64 s4, s[6:7], s[4:5] -; GFX11-NEXT: v_dual_mov_b32 v2, 0 :: v_dual_mov_b32 v1, s7 -; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_2) +; GFX11-NEXT: s_sub_u32 s4, s4, s6 +; GFX11-NEXT: s_cselect_b32 s6, -1, 0 +; GFX11-NEXT: v_mov_b32_e32 v0, s4 +; GFX11-NEXT: s_cmp_lg_u32 s6, 0 +; GFX11-NEXT: s_subb_u32 s5, s5, s7 +; GFX11-NEXT: s_cselect_b32 s4, -1, 0 +; GFX11-NEXT: v_dual_mov_b32 v2, 0 :: v_dual_mov_b32 v1, s5 ; GFX11-NEXT: v_cndmask_b32_e64 v3, 0, 1, s4 ; GFX11-NEXT: s_clause 0x1 ; GFX11-NEXT: global_store_b64 v2, v[0:1], s[0:1] @@ -1799,12 +1816,14 @@ define amdgpu_kernel void @susubo64(ptr addrspace(1) %out, ptr addrspace(1) %car ; GFX1250-LABEL: susubo64: ; GFX1250: ; %bb.0: ; GFX1250-NEXT: s_load_b256 s[0:7], s[4:5], 0x24 -; GFX1250-NEXT: v_mov_b32_e32 v2, 0 ; GFX1250-NEXT: s_wait_kmcnt 0x0 -; GFX1250-NEXT: s_sub_nc_u64 s[6:7], s[4:5], s[6:7] -; GFX1250-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_1) | instid1(VALU_DEP_2) -; GFX1250-NEXT: v_cmp_gt_u64_e64 s4, s[6:7], s[4:5] -; GFX1250-NEXT: v_mov_b64_e32 v[0:1], s[6:7] +; GFX1250-NEXT: s_sub_co_u32 s4, s4, s6 +; GFX1250-NEXT: s_cselect_b32 s6, -1, 0 +; GFX1250-NEXT: v_dual_mov_b32 v2, 0 :: v_dual_mov_b32 v0, s4 +; GFX1250-NEXT: s_cmp_lg_u32 s6, 0 +; GFX1250-NEXT: s_sub_co_ci_u32 s5, s5, s7 +; GFX1250-NEXT: s_cselect_b32 s4, -1, 0 +; GFX1250-NEXT: v_mov_b32_e32 v1, s5 ; GFX1250-NEXT: v_cndmask_b32_e64 v3, 0, 1, s4 ; GFX1250-NEXT: s_clause 0x1 ; GFX1250-NEXT: global_store_b64 v2, v[0:1], s[0:1] @@ -1821,7 +1840,8 @@ define amdgpu_kernel void @susubo64(ptr addrspace(1) %out, ptr addrspace(1) %car ; GCN-ISEL-LABEL: name: vusubo64 ; GCN-ISEL-LABEL: body: ; GCN-ISEL-LABEL: bb.0 -; GCN-ISEL: V_SUB_U64_PSEUDO +; GCN-ISEL: V_SUB_CO_U32_e64 +; GCN-ISEL: V_SUBB_U32_e64 define amdgpu_kernel void @vusubo64(ptr addrspace(1) %out, ptr addrspace(1) %carryout, i64 %a) #0 { ; CISI-LABEL: vusubo64: @@ -1834,9 +1854,8 @@ define amdgpu_kernel void @vusubo64(ptr addrspace(1) %out, ptr addrspace(1) %car ; CISI-NEXT: s_mov_b32 s4, s0 ; CISI-NEXT: v_mov_b32_e32 v1, s9 ; CISI-NEXT: v_sub_i32_e32 v0, vcc, s8, v0 -; CISI-NEXT: v_subbrev_u32_e32 v1, vcc, 0, v1, vcc -; CISI-NEXT: v_cmp_lt_u64_e32 vcc, s[8:9], v[0:1] ; CISI-NEXT: s_mov_b32 s5, s1 +; CISI-NEXT: v_subbrev_u32_e32 v1, vcc, 0, v1, vcc ; CISI-NEXT: s_mov_b32 s0, s2 ; CISI-NEXT: s_mov_b32 s1, s3 ; CISI-NEXT: s_mov_b32 s2, s6 @@ -1856,7 +1875,6 @@ define amdgpu_kernel void @vusubo64(ptr addrspace(1) %out, ptr addrspace(1) %car ; VI-NEXT: v_mov_b32_e32 v6, s5 ; VI-NEXT: v_sub_u32_e32 v5, vcc, s4, v0 ; VI-NEXT: v_subbrev_u32_e32 v6, vcc, 0, v6, vcc -; VI-NEXT: v_cmp_lt_u64_e32 vcc, s[4:5], v[5:6] ; VI-NEXT: v_mov_b32_e32 v2, s1 ; VI-NEXT: v_mov_b32_e32 v3, s2 ; VI-NEXT: v_mov_b32_e32 v4, s3 @@ -1874,7 +1892,6 @@ define amdgpu_kernel void @vusubo64(ptr addrspace(1) %out, ptr addrspace(1) %car ; GFX9-NEXT: v_mov_b32_e32 v1, s7 ; GFX9-NEXT: v_sub_co_u32_e32 v0, vcc, s6, v0 ; GFX9-NEXT: v_subbrev_co_u32_e32 v1, vcc, 0, v1, vcc -; GFX9-NEXT: v_cmp_lt_u64_e32 vcc, s[6:7], v[0:1] ; GFX9-NEXT: global_store_dwordx2 v2, v[0:1], s[0:1] ; GFX9-NEXT: v_cndmask_b32_e64 v0, 0, 1, vcc ; GFX9-NEXT: global_store_byte v2, v0, s[2:3] @@ -1889,8 +1906,7 @@ define amdgpu_kernel void @vusubo64(ptr addrspace(1) %out, ptr addrspace(1) %car ; GFX1010-NEXT: s_waitcnt lgkmcnt(0) ; GFX1010-NEXT: v_sub_co_u32 v0, s4, s6, v0 ; GFX1010-NEXT: v_sub_co_ci_u32_e64 v1, s4, s7, 0, s4 -; GFX1010-NEXT: v_cmp_lt_u64_e32 vcc_lo, s[6:7], v[0:1] -; GFX1010-NEXT: v_cndmask_b32_e64 v3, 0, 1, vcc_lo +; GFX1010-NEXT: v_cndmask_b32_e64 v3, 0, 1, s4 ; GFX1010-NEXT: global_store_dwordx2 v2, v[0:1], s[0:1] ; GFX1010-NEXT: global_store_byte v2, v3, s[2:3] ; GFX1010-NEXT: s_endpgm @@ -1903,9 +1919,8 @@ define amdgpu_kernel void @vusubo64(ptr addrspace(1) %out, ptr addrspace(1) %car ; GFX1030W32-NEXT: v_mov_b32_e32 v2, 0 ; GFX1030W32-NEXT: s_waitcnt lgkmcnt(0) ; GFX1030W32-NEXT: v_sub_co_u32 v0, s4, s6, v0 -; GFX1030W32-NEXT: v_sub_co_ci_u32_e64 v1, null, s7, 0, s4 -; GFX1030W32-NEXT: v_cmp_lt_u64_e32 vcc_lo, s[6:7], v[0:1] -; GFX1030W32-NEXT: v_cndmask_b32_e64 v3, 0, 1, vcc_lo +; GFX1030W32-NEXT: v_sub_co_ci_u32_e64 v1, s4, s7, 0, s4 +; GFX1030W32-NEXT: v_cndmask_b32_e64 v3, 0, 1, s4 ; GFX1030W32-NEXT: global_store_dwordx2 v2, v[0:1], s[0:1] ; GFX1030W32-NEXT: global_store_byte v2, v3, s[2:3] ; GFX1030W32-NEXT: s_endpgm @@ -1918,9 +1933,8 @@ define amdgpu_kernel void @vusubo64(ptr addrspace(1) %out, ptr addrspace(1) %car ; GFX1030W64-NEXT: v_mov_b32_e32 v2, 0 ; GFX1030W64-NEXT: s_waitcnt lgkmcnt(0) ; GFX1030W64-NEXT: v_sub_co_u32 v0, s[4:5], s6, v0 -; GFX1030W64-NEXT: v_sub_co_ci_u32_e64 v1, null, s7, 0, s[4:5] -; GFX1030W64-NEXT: v_cmp_lt_u64_e32 vcc, s[6:7], v[0:1] -; GFX1030W64-NEXT: v_cndmask_b32_e64 v3, 0, 1, vcc +; GFX1030W64-NEXT: v_sub_co_ci_u32_e64 v1, s[4:5], s7, 0, s[4:5] +; GFX1030W64-NEXT: v_cndmask_b32_e64 v3, 0, 1, s[4:5] ; GFX1030W64-NEXT: global_store_dwordx2 v2, v[0:1], s[0:1] ; GFX1030W64-NEXT: global_store_byte v2, v3, s[2:3] ; GFX1030W64-NEXT: s_endpgm @@ -1935,10 +1949,9 @@ define amdgpu_kernel void @vusubo64(ptr addrspace(1) %out, ptr addrspace(1) %car ; GFX11-NEXT: s_waitcnt lgkmcnt(0) ; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1) ; GFX11-NEXT: v_sub_co_u32 v0, s4, s6, v0 -; GFX11-NEXT: v_sub_co_ci_u32_e64 v1, null, s7, 0, s4 +; GFX11-NEXT: v_sub_co_ci_u32_e64 v1, s4, s7, 0, s4 ; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) -; GFX11-NEXT: v_cmp_lt_u64_e32 vcc_lo, s[6:7], v[0:1] -; GFX11-NEXT: v_cndmask_b32_e64 v3, 0, 1, vcc_lo +; GFX11-NEXT: v_cndmask_b32_e64 v3, 0, 1, s4 ; GFX11-NEXT: s_clause 0x1 ; GFX11-NEXT: global_store_b64 v2, v[0:1], s[0:1] ; GFX11-NEXT: global_store_b8 v2, v3, s[2:3] @@ -1949,16 +1962,17 @@ define amdgpu_kernel void @vusubo64(ptr addrspace(1) %out, ptr addrspace(1) %car ; GFX1250-NEXT: s_clause 0x1 ; GFX1250-NEXT: s_load_b64 s[6:7], s[4:5], 0x34 ; GFX1250-NEXT: s_load_b128 s[0:3], s[4:5], 0x24 -; GFX1250-NEXT: v_mov_b32_e32 v1, 0 ; GFX1250-NEXT: v_and_b32_e32 v0, 0x3ff, v0 +; GFX1250-NEXT: v_mov_b32_e32 v2, 0 ; GFX1250-NEXT: s_wait_kmcnt 0x0 -; GFX1250-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1) -; GFX1250-NEXT: v_sub_nc_u64_e32 v[2:3], s[6:7], v[0:1] -; GFX1250-NEXT: v_cmp_lt_u64_e32 vcc_lo, s[6:7], v[2:3] -; GFX1250-NEXT: v_cndmask_b32_e64 v0, 0, 1, vcc_lo +; GFX1250-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1) +; GFX1250-NEXT: v_sub_co_u32 v0, s4, s6, v0 +; GFX1250-NEXT: v_sub_co_ci_u32_e64 v1, s4, s7, 0, s4 +; GFX1250-NEXT: s_delay_alu instid0(VALU_DEP_1) +; GFX1250-NEXT: v_cndmask_b32_e64 v3, 0, 1, s4 ; GFX1250-NEXT: s_clause 0x1 -; GFX1250-NEXT: global_store_b64 v1, v[2:3], s[0:1] -; GFX1250-NEXT: global_store_b8 v1, v0, s[2:3] +; GFX1250-NEXT: global_store_b64 v2, v[0:1], s[0:1] +; GFX1250-NEXT: global_store_b8 v2, v3, s[2:3] ; GFX1250-NEXT: s_endpgm %tid = call i32 @llvm.amdgcn.workitem.id.x() %tid.ext = sext i32 %tid to i64 diff --git a/llvm/test/CodeGen/AMDGPU/sdiv64.ll b/llvm/test/CodeGen/AMDGPU/sdiv64.ll index ddec6af0af69e..3c40fd23a1c44 100644 --- a/llvm/test/CodeGen/AMDGPU/sdiv64.ll +++ b/llvm/test/CodeGen/AMDGPU/sdiv64.ll @@ -206,8 +206,11 @@ define amdgpu_kernel void @s_test_sdiv(ptr addrspace(1) %out, i64 %x, i64 %y) { ; GCN-IR-NEXT: s_cbranch_vccz .LBB0_5 ; GCN-IR-NEXT: ; %bb.1: ; %udiv-bb1 ; GCN-IR-NEXT: s_add_u32 s18, s16, 1 -; GCN-IR-NEXT: s_addc_u32 s19, s17, 0 -; GCN-IR-NEXT: v_cmp_eq_u64_e64 s[10:11], s[18:19], 0 +; GCN-IR-NEXT: s_cselect_b64 s[10:11], -1, 0 +; GCN-IR-NEXT: s_or_b32 s10, s10, s11 +; GCN-IR-NEXT: s_cmp_lg_u32 s10, 0 +; GCN-IR-NEXT: s_addc_u32 s10, s17, 0 +; GCN-IR-NEXT: s_cselect_b64 s[10:11], -1, 0 ; GCN-IR-NEXT: s_sub_i32 s16, 63, s16 ; GCN-IR-NEXT: s_andn2_b64 vcc, exec, s[10:11] ; GCN-IR-NEXT: s_lshl_b64 s[10:11], s[12:13], s16 @@ -217,9 +220,9 @@ define amdgpu_kernel void @s_test_sdiv(ptr addrspace(1) %out, i64 %x, i64 %y) { ; GCN-IR-NEXT: s_add_u32 s18, s2, -1 ; GCN-IR-NEXT: s_addc_u32 s19, s3, -1 ; GCN-IR-NEXT: s_not_b64 s[8:9], s[14:15] -; GCN-IR-NEXT: s_add_u32 s12, s8, s20 -; GCN-IR-NEXT: s_addc_u32 s13, s9, 0 -; GCN-IR-NEXT: s_mov_b64 s[14:15], 0 +; GCN-IR-NEXT: s_add_u32 s14, s8, s20 +; GCN-IR-NEXT: s_addc_u32 s15, s9, 0 +; GCN-IR-NEXT: s_mov_b64 s[12:13], 0 ; GCN-IR-NEXT: s_mov_b32 s9, 0 ; GCN-IR-NEXT: .LBB0_3: ; %udiv-do-while ; GCN-IR-NEXT: ; =>This Inner Loop Header: Depth=1 @@ -227,19 +230,22 @@ define amdgpu_kernel void @s_test_sdiv(ptr addrspace(1) %out, i64 %x, i64 %y) { ; GCN-IR-NEXT: s_lshr_b32 s8, s11, 31 ; GCN-IR-NEXT: s_lshl_b64 s[10:11], s[10:11], 1 ; GCN-IR-NEXT: s_or_b64 s[16:17], s[16:17], s[8:9] -; GCN-IR-NEXT: s_or_b64 s[10:11], s[14:15], s[10:11] +; GCN-IR-NEXT: s_or_b64 s[10:11], s[12:13], s[10:11] ; GCN-IR-NEXT: s_sub_u32 s8, s18, s16 ; GCN-IR-NEXT: s_subb_u32 s8, s19, s17 -; GCN-IR-NEXT: s_ashr_i32 s14, s8, 31 -; GCN-IR-NEXT: s_mov_b32 s15, s14 -; GCN-IR-NEXT: s_and_b32 s8, s14, 1 -; GCN-IR-NEXT: s_and_b64 s[14:15], s[14:15], s[2:3] -; GCN-IR-NEXT: s_sub_u32 s16, s16, s14 -; GCN-IR-NEXT: s_subb_u32 s17, s17, s15 -; GCN-IR-NEXT: s_add_u32 s12, s12, 1 -; GCN-IR-NEXT: s_addc_u32 s13, s13, 0 -; GCN-IR-NEXT: v_cmp_eq_u64_e64 s[20:21], s[12:13], 0 -; GCN-IR-NEXT: s_mov_b64 s[14:15], s[8:9] +; GCN-IR-NEXT: s_ashr_i32 s12, s8, 31 +; GCN-IR-NEXT: s_mov_b32 s13, s12 +; GCN-IR-NEXT: s_and_b32 s8, s12, 1 +; GCN-IR-NEXT: s_and_b64 s[20:21], s[12:13], s[2:3] +; GCN-IR-NEXT: s_sub_u32 s16, s16, s20 +; GCN-IR-NEXT: s_subb_u32 s17, s17, s21 +; GCN-IR-NEXT: s_add_u32 s14, s14, 1 +; GCN-IR-NEXT: s_cselect_b64 s[20:21], -1, 0 +; GCN-IR-NEXT: s_or_b32 s20, s20, s21 +; GCN-IR-NEXT: s_cmp_lg_u32 s20, 0 +; GCN-IR-NEXT: s_addc_u32 s15, s15, 0 +; GCN-IR-NEXT: s_cselect_b64 s[20:21], -1, 0 +; GCN-IR-NEXT: s_mov_b64 s[12:13], s[8:9] ; GCN-IR-NEXT: s_and_b64 vcc, exec, s[20:21] ; GCN-IR-NEXT: s_cbranch_vccz .LBB0_3 ; GCN-IR-NEXT: .LBB0_4: ; %Flow7 @@ -389,25 +395,25 @@ define i64 @v_test_sdiv(i64 %x, i64 %y) { ; GCN-IR-LABEL: v_test_sdiv: ; GCN-IR: ; %bb.0: ; %_udiv-special-cases ; GCN-IR-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) -; GCN-IR-NEXT: v_ashrrev_i32_e32 v12, 31, v1 -; GCN-IR-NEXT: v_xor_b32_e32 v0, v0, v12 -; GCN-IR-NEXT: v_ashrrev_i32_e32 v13, 31, v3 -; GCN-IR-NEXT: v_xor_b32_e32 v1, v1, v12 -; GCN-IR-NEXT: v_sub_i32_e32 v6, vcc, v0, v12 -; GCN-IR-NEXT: v_subb_u32_e32 v7, vcc, v1, v12, vcc -; GCN-IR-NEXT: v_xor_b32_e32 v0, v2, v13 -; GCN-IR-NEXT: v_xor_b32_e32 v1, v3, v13 -; GCN-IR-NEXT: v_sub_i32_e32 v0, vcc, v0, v13 -; GCN-IR-NEXT: v_subb_u32_e32 v1, vcc, v1, v13, vcc +; GCN-IR-NEXT: v_ashrrev_i32_e32 v10, 31, v1 +; GCN-IR-NEXT: v_xor_b32_e32 v0, v0, v10 +; GCN-IR-NEXT: v_ashrrev_i32_e32 v11, 31, v3 +; GCN-IR-NEXT: v_xor_b32_e32 v1, v1, v10 +; GCN-IR-NEXT: v_sub_i32_e32 v6, vcc, v0, v10 +; GCN-IR-NEXT: v_subb_u32_e32 v7, vcc, v1, v10, vcc +; GCN-IR-NEXT: v_xor_b32_e32 v0, v2, v11 +; GCN-IR-NEXT: v_xor_b32_e32 v1, v3, v11 +; GCN-IR-NEXT: v_sub_i32_e32 v0, vcc, v0, v11 +; GCN-IR-NEXT: v_subb_u32_e32 v1, vcc, v1, v11, vcc ; GCN-IR-NEXT: v_ffbh_u32_e32 v2, v0 ; GCN-IR-NEXT: v_add_i32_e64 v2, s[6:7], 32, v2 ; GCN-IR-NEXT: v_ffbh_u32_e32 v3, v1 -; GCN-IR-NEXT: v_min_u32_e32 v10, v2, v3 +; GCN-IR-NEXT: v_min_u32_e32 v8, v2, v3 ; GCN-IR-NEXT: v_ffbh_u32_e32 v2, v6 ; GCN-IR-NEXT: v_add_i32_e64 v2, s[6:7], 32, v2 ; GCN-IR-NEXT: v_ffbh_u32_e32 v3, v7 -; GCN-IR-NEXT: v_min_u32_e32 v11, v2, v3 -; GCN-IR-NEXT: v_sub_i32_e64 v2, s[6:7], v10, v11 +; GCN-IR-NEXT: v_min_u32_e32 v9, v2, v3 +; GCN-IR-NEXT: v_sub_i32_e64 v2, s[6:7], v8, v9 ; GCN-IR-NEXT: v_cmp_eq_u64_e32 vcc, 0, v[0:1] ; GCN-IR-NEXT: v_cmp_eq_u64_e64 s[4:5], 0, v[6:7] ; GCN-IR-NEXT: v_subb_u32_e64 v3, s[6:7], 0, 0, s[6:7] @@ -416,70 +422,69 @@ define i64 @v_test_sdiv(i64 %x, i64 %y) { ; GCN-IR-NEXT: s_or_b64 s[4:5], s[4:5], s[6:7] ; GCN-IR-NEXT: v_cmp_ne_u64_e32 vcc, 63, v[2:3] ; GCN-IR-NEXT: s_xor_b64 s[6:7], s[4:5], -1 -; GCN-IR-NEXT: v_mov_b32_e32 v14, v12 -; GCN-IR-NEXT: v_mov_b32_e32 v15, v13 +; GCN-IR-NEXT: v_mov_b32_e32 v12, v10 +; GCN-IR-NEXT: v_mov_b32_e32 v13, v11 ; GCN-IR-NEXT: v_cndmask_b32_e64 v5, v7, 0, s[4:5] ; GCN-IR-NEXT: v_cndmask_b32_e64 v4, v6, 0, s[4:5] ; GCN-IR-NEXT: s_and_b64 s[4:5], s[6:7], vcc ; GCN-IR-NEXT: s_and_saveexec_b64 s[6:7], s[4:5] ; GCN-IR-NEXT: s_cbranch_execz .LBB1_6 ; GCN-IR-NEXT: ; %bb.1: ; %udiv-bb1 -; GCN-IR-NEXT: v_add_i32_e32 v8, vcc, 1, v2 -; GCN-IR-NEXT: v_addc_u32_e32 v9, vcc, 0, v3, vcc +; GCN-IR-NEXT: v_add_i32_e32 v14, vcc, 1, v2 +; GCN-IR-NEXT: v_addc_u32_e32 v3, vcc, 0, v3, vcc ; GCN-IR-NEXT: v_sub_i32_e64 v2, s[4:5], 63, v2 -; GCN-IR-NEXT: v_mov_b32_e32 v4, 0 -; GCN-IR-NEXT: v_cmp_ne_u64_e32 vcc, 0, v[8:9] ; GCN-IR-NEXT: v_lshl_b64 v[2:3], v[6:7], v2 +; GCN-IR-NEXT: v_mov_b32_e32 v4, 0 ; GCN-IR-NEXT: v_mov_b32_e32 v5, 0 -; GCN-IR-NEXT: s_and_saveexec_b64 s[4:5], vcc -; GCN-IR-NEXT: s_xor_b64 s[8:9], exec, s[4:5] +; GCN-IR-NEXT: s_xor_b64 s[4:5], vcc, -1 +; GCN-IR-NEXT: s_and_saveexec_b64 s[8:9], s[4:5] +; GCN-IR-NEXT: s_xor_b64 s[4:5], exec, s[8:9] ; GCN-IR-NEXT: s_cbranch_execz .LBB1_5 ; GCN-IR-NEXT: ; %bb.2: ; %udiv-preheader -; GCN-IR-NEXT: v_add_i32_e32 v16, vcc, -1, v0 -; GCN-IR-NEXT: v_addc_u32_e32 v17, vcc, -1, v1, vcc -; GCN-IR-NEXT: v_not_b32_e32 v4, v10 -; GCN-IR-NEXT: v_lshr_b64 v[8:9], v[6:7], v8 -; GCN-IR-NEXT: v_add_i32_e32 v6, vcc, v4, v11 -; GCN-IR-NEXT: v_mov_b32_e32 v10, 0 -; GCN-IR-NEXT: v_addc_u32_e64 v7, s[4:5], -1, 0, vcc -; GCN-IR-NEXT: s_mov_b64 s[10:11], 0 -; GCN-IR-NEXT: v_mov_b32_e32 v11, 0 +; GCN-IR-NEXT: v_lshr_b64 v[6:7], v[6:7], v14 +; GCN-IR-NEXT: v_add_i32_e32 v14, vcc, -1, v0 +; GCN-IR-NEXT: v_addc_u32_e32 v15, vcc, -1, v1, vcc +; GCN-IR-NEXT: v_not_b32_e32 v4, v8 +; GCN-IR-NEXT: v_add_i32_e32 v16, vcc, v4, v9 +; GCN-IR-NEXT: v_addc_u32_e64 v17, s[8:9], -1, 0, vcc +; GCN-IR-NEXT: v_mov_b32_e32 v8, 0 +; GCN-IR-NEXT: s_mov_b64 s[8:9], 0 +; GCN-IR-NEXT: v_mov_b32_e32 v9, 0 ; GCN-IR-NEXT: v_mov_b32_e32 v5, 0 ; GCN-IR-NEXT: .LBB1_3: ; %udiv-do-while ; GCN-IR-NEXT: ; =>This Inner Loop Header: Depth=1 -; GCN-IR-NEXT: v_lshl_b64 v[8:9], v[8:9], 1 +; GCN-IR-NEXT: v_lshl_b64 v[6:7], v[6:7], 1 ; GCN-IR-NEXT: v_lshrrev_b32_e32 v4, 31, v3 -; GCN-IR-NEXT: v_or_b32_e32 v8, v8, v4 +; GCN-IR-NEXT: v_or_b32_e32 v6, v6, v4 ; GCN-IR-NEXT: v_lshl_b64 v[2:3], v[2:3], 1 -; GCN-IR-NEXT: v_sub_i32_e32 v4, vcc, v16, v8 -; GCN-IR-NEXT: v_subb_u32_e32 v4, vcc, v17, v9, vcc -; GCN-IR-NEXT: v_or_b32_e32 v2, v10, v2 -; GCN-IR-NEXT: v_ashrrev_i32_e32 v10, 31, v4 -; GCN-IR-NEXT: v_add_i32_e32 v6, vcc, 1, v6 -; GCN-IR-NEXT: v_or_b32_e32 v3, v11, v3 -; GCN-IR-NEXT: v_and_b32_e32 v4, 1, v10 -; GCN-IR-NEXT: v_and_b32_e32 v11, v10, v1 -; GCN-IR-NEXT: v_and_b32_e32 v10, v10, v0 -; GCN-IR-NEXT: v_addc_u32_e32 v7, vcc, 0, v7, vcc -; GCN-IR-NEXT: v_cmp_eq_u64_e32 vcc, 0, v[6:7] -; GCN-IR-NEXT: v_sub_i32_e64 v8, s[4:5], v8, v10 -; GCN-IR-NEXT: v_subb_u32_e64 v9, s[4:5], v9, v11, s[4:5] -; GCN-IR-NEXT: v_mov_b32_e32 v11, v5 -; GCN-IR-NEXT: s_or_b64 s[10:11], vcc, s[10:11] -; GCN-IR-NEXT: v_mov_b32_e32 v10, v4 -; GCN-IR-NEXT: s_andn2_b64 exec, exec, s[10:11] +; GCN-IR-NEXT: v_sub_i32_e32 v4, vcc, v14, v6 +; GCN-IR-NEXT: v_subb_u32_e32 v4, vcc, v15, v7, vcc +; GCN-IR-NEXT: v_or_b32_e32 v2, v8, v2 +; GCN-IR-NEXT: v_ashrrev_i32_e32 v8, 31, v4 +; GCN-IR-NEXT: v_or_b32_e32 v3, v9, v3 +; GCN-IR-NEXT: v_and_b32_e32 v4, 1, v8 +; GCN-IR-NEXT: v_and_b32_e32 v9, v8, v1 +; GCN-IR-NEXT: v_and_b32_e32 v8, v8, v0 +; GCN-IR-NEXT: v_sub_i32_e32 v6, vcc, v6, v8 +; GCN-IR-NEXT: v_subb_u32_e32 v7, vcc, v7, v9, vcc +; GCN-IR-NEXT: v_add_i32_e32 v16, vcc, 1, v16 +; GCN-IR-NEXT: v_addc_u32_e32 v17, vcc, 0, v17, vcc +; GCN-IR-NEXT: v_mov_b32_e32 v9, v5 +; GCN-IR-NEXT: s_or_b64 s[8:9], vcc, s[8:9] +; GCN-IR-NEXT: v_mov_b32_e32 v8, v4 +; GCN-IR-NEXT: s_andn2_b64 exec, exec, s[8:9] ; GCN-IR-NEXT: s_cbranch_execnz .LBB1_3 ; GCN-IR-NEXT: ; %bb.4: ; %Flow -; GCN-IR-NEXT: s_or_b64 exec, exec, s[10:11] -; GCN-IR-NEXT: .LBB1_5: ; %Flow4 ; GCN-IR-NEXT: s_or_b64 exec, exec, s[8:9] +; GCN-IR-NEXT: .LBB1_5: ; %Flow4 +; GCN-IR-NEXT: s_or_b64 exec, exec, s[4:5] ; GCN-IR-NEXT: v_lshl_b64 v[0:1], v[2:3], 1 ; GCN-IR-NEXT: v_or_b32_e32 v5, v5, v1 ; GCN-IR-NEXT: v_or_b32_e32 v4, v4, v0 ; GCN-IR-NEXT: .LBB1_6: ; %Flow5 ; GCN-IR-NEXT: s_or_b64 exec, exec, s[6:7] -; GCN-IR-NEXT: v_xor_b32_e32 v0, v13, v12 -; GCN-IR-NEXT: v_xor_b32_e32 v1, v15, v14 +; GCN-IR-NEXT: v_xor_b32_e32 v0, v11, v10 +; GCN-IR-NEXT: v_xor_b32_e32 v1, v13, v12 ; GCN-IR-NEXT: v_xor_b32_e32 v3, v4, v0 ; GCN-IR-NEXT: v_xor_b32_e32 v2, v5, v1 ; GCN-IR-NEXT: v_sub_i32_e32 v0, vcc, v3, v0 @@ -1295,34 +1300,37 @@ define amdgpu_kernel void @s_test_sdiv_k_num_i64(ptr addrspace(1) %out, i64 %x) ; GCN-IR-NEXT: s_xor_b64 s[2:3], s[2:3], s[4:5] ; GCN-IR-NEXT: s_sub_u32 s2, s2, s4 ; GCN-IR-NEXT: s_subb_u32 s3, s3, s4 -; GCN-IR-NEXT: s_flbit_i32_b64 s14, s[2:3] -; GCN-IR-NEXT: s_add_u32 s10, s14, 0xffffffc5 +; GCN-IR-NEXT: s_flbit_i32_b64 s16, s[2:3] +; GCN-IR-NEXT: s_add_u32 s10, s16, 0xffffffc5 ; GCN-IR-NEXT: s_addc_u32 s11, 0, -1 ; GCN-IR-NEXT: v_cmp_eq_u64_e64 s[8:9], s[2:3], 0 ; GCN-IR-NEXT: v_cmp_gt_u64_e64 s[12:13], s[10:11], 63 -; GCN-IR-NEXT: v_cmp_eq_u64_e64 s[16:17], s[10:11], 63 +; GCN-IR-NEXT: v_cmp_eq_u64_e64 s[14:15], s[10:11], 63 ; GCN-IR-NEXT: s_or_b64 s[12:13], s[8:9], s[12:13] ; GCN-IR-NEXT: s_and_b64 s[8:9], s[12:13], exec ; GCN-IR-NEXT: s_cselect_b32 s8, 0, 24 -; GCN-IR-NEXT: s_or_b64 s[12:13], s[12:13], s[16:17] +; GCN-IR-NEXT: s_or_b64 s[12:13], s[12:13], s[14:15] ; GCN-IR-NEXT: s_andn2_b64 vcc, exec, s[12:13] ; GCN-IR-NEXT: s_mov_b32 s9, 0 ; GCN-IR-NEXT: s_cbranch_vccz .LBB10_5 ; GCN-IR-NEXT: ; %bb.1: ; %udiv-bb1 ; GCN-IR-NEXT: s_add_u32 s12, s10, 1 -; GCN-IR-NEXT: s_addc_u32 s13, s11, 0 -; GCN-IR-NEXT: v_cmp_eq_u64_e64 s[8:9], s[12:13], 0 +; GCN-IR-NEXT: s_cselect_b64 s[8:9], -1, 0 +; GCN-IR-NEXT: s_or_b32 s8, s8, s9 +; GCN-IR-NEXT: s_cmp_lg_u32 s8, 0 +; GCN-IR-NEXT: s_addc_u32 s8, s11, 0 +; GCN-IR-NEXT: s_cselect_b64 s[8:9], -1, 0 ; GCN-IR-NEXT: s_sub_i32 s10, 63, s10 ; GCN-IR-NEXT: s_andn2_b64 vcc, exec, s[8:9] ; GCN-IR-NEXT: s_lshl_b64 s[8:9], 24, s10 ; GCN-IR-NEXT: s_cbranch_vccz .LBB10_4 ; GCN-IR-NEXT: ; %bb.2: ; %udiv-preheader ; GCN-IR-NEXT: s_lshr_b64 s[12:13], 24, s12 -; GCN-IR-NEXT: s_add_u32 s16, s2, -1 -; GCN-IR-NEXT: s_addc_u32 s17, s3, -1 -; GCN-IR-NEXT: s_sub_u32 s10, 58, s14 -; GCN-IR-NEXT: s_subb_u32 s11, 0, 0 -; GCN-IR-NEXT: s_mov_b64 s[14:15], 0 +; GCN-IR-NEXT: s_add_u32 s14, s2, -1 +; GCN-IR-NEXT: s_addc_u32 s15, s3, -1 +; GCN-IR-NEXT: s_sub_u32 s16, 58, s16 +; GCN-IR-NEXT: s_subb_u32 s17, 0, 0 +; GCN-IR-NEXT: s_mov_b64 s[10:11], 0 ; GCN-IR-NEXT: s_mov_b32 s7, 0 ; GCN-IR-NEXT: .LBB10_3: ; %udiv-do-while ; GCN-IR-NEXT: ; =>This Inner Loop Header: Depth=1 @@ -1330,19 +1338,22 @@ define amdgpu_kernel void @s_test_sdiv_k_num_i64(ptr addrspace(1) %out, i64 %x) ; GCN-IR-NEXT: s_lshr_b32 s6, s9, 31 ; GCN-IR-NEXT: s_lshl_b64 s[8:9], s[8:9], 1 ; GCN-IR-NEXT: s_or_b64 s[12:13], s[12:13], s[6:7] -; GCN-IR-NEXT: s_or_b64 s[8:9], s[14:15], s[8:9] -; GCN-IR-NEXT: s_sub_u32 s6, s16, s12 -; GCN-IR-NEXT: s_subb_u32 s6, s17, s13 -; GCN-IR-NEXT: s_ashr_i32 s14, s6, 31 -; GCN-IR-NEXT: s_mov_b32 s15, s14 -; GCN-IR-NEXT: s_and_b32 s6, s14, 1 -; GCN-IR-NEXT: s_and_b64 s[14:15], s[14:15], s[2:3] -; GCN-IR-NEXT: s_sub_u32 s12, s12, s14 -; GCN-IR-NEXT: s_subb_u32 s13, s13, s15 -; GCN-IR-NEXT: s_add_u32 s10, s10, 1 -; GCN-IR-NEXT: s_addc_u32 s11, s11, 0 -; GCN-IR-NEXT: v_cmp_eq_u64_e64 s[18:19], s[10:11], 0 -; GCN-IR-NEXT: s_mov_b64 s[14:15], s[6:7] +; GCN-IR-NEXT: s_or_b64 s[8:9], s[10:11], s[8:9] +; GCN-IR-NEXT: s_sub_u32 s6, s14, s12 +; GCN-IR-NEXT: s_subb_u32 s6, s15, s13 +; GCN-IR-NEXT: s_ashr_i32 s10, s6, 31 +; GCN-IR-NEXT: s_mov_b32 s11, s10 +; GCN-IR-NEXT: s_and_b32 s6, s10, 1 +; GCN-IR-NEXT: s_and_b64 s[18:19], s[10:11], s[2:3] +; GCN-IR-NEXT: s_sub_u32 s12, s12, s18 +; GCN-IR-NEXT: s_subb_u32 s13, s13, s19 +; GCN-IR-NEXT: s_add_u32 s16, s16, 1 +; GCN-IR-NEXT: s_cselect_b64 s[18:19], -1, 0 +; GCN-IR-NEXT: s_or_b32 s18, s18, s19 +; GCN-IR-NEXT: s_cmp_lg_u32 s18, 0 +; GCN-IR-NEXT: s_addc_u32 s17, s17, 0 +; GCN-IR-NEXT: s_cselect_b64 s[18:19], -1, 0 +; GCN-IR-NEXT: s_mov_b64 s[10:11], s[6:7] ; GCN-IR-NEXT: s_and_b64 vcc, exec, s[18:19] ; GCN-IR-NEXT: s_cbranch_vccz .LBB10_3 ; GCN-IR-NEXT: .LBB10_4: ; %Flow6 @@ -1474,17 +1485,17 @@ define i64 @v_test_sdiv_k_num_i64(i64 %x) { ; GCN-IR-LABEL: v_test_sdiv_k_num_i64: ; GCN-IR: ; %bb.0: ; %_udiv-special-cases ; GCN-IR-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) -; GCN-IR-NEXT: v_ashrrev_i32_e32 v12, 31, v1 -; GCN-IR-NEXT: v_xor_b32_e32 v0, v0, v12 -; GCN-IR-NEXT: v_xor_b32_e32 v1, v1, v12 -; GCN-IR-NEXT: v_sub_i32_e32 v0, vcc, v0, v12 -; GCN-IR-NEXT: v_subb_u32_e32 v1, vcc, v1, v12, vcc +; GCN-IR-NEXT: v_ashrrev_i32_e32 v10, 31, v1 +; GCN-IR-NEXT: v_xor_b32_e32 v0, v0, v10 +; GCN-IR-NEXT: v_xor_b32_e32 v1, v1, v10 +; GCN-IR-NEXT: v_sub_i32_e32 v0, vcc, v0, v10 +; GCN-IR-NEXT: v_subb_u32_e32 v1, vcc, v1, v10, vcc ; GCN-IR-NEXT: v_ffbh_u32_e32 v2, v0 ; GCN-IR-NEXT: v_add_i32_e32 v2, vcc, 32, v2 ; GCN-IR-NEXT: v_ffbh_u32_e32 v3, v1 -; GCN-IR-NEXT: v_min_u32_e32 v10, v2, v3 +; GCN-IR-NEXT: v_min_u32_e32 v8, v2, v3 ; GCN-IR-NEXT: s_movk_i32 s6, 0xffc5 -; GCN-IR-NEXT: v_add_i32_e32 v2, vcc, s6, v10 +; GCN-IR-NEXT: v_add_i32_e32 v2, vcc, s6, v8 ; GCN-IR-NEXT: v_addc_u32_e64 v3, s[6:7], 0, -1, vcc ; GCN-IR-NEXT: v_cmp_eq_u64_e64 s[4:5], 0, v[0:1] ; GCN-IR-NEXT: v_cmp_lt_u64_e32 vcc, 63, v[2:3] @@ -1492,69 +1503,68 @@ define i64 @v_test_sdiv_k_num_i64(i64 %x) { ; GCN-IR-NEXT: s_or_b64 s[4:5], s[4:5], vcc ; GCN-IR-NEXT: v_cndmask_b32_e64 v4, 24, 0, s[4:5] ; GCN-IR-NEXT: s_xor_b64 s[4:5], s[4:5], -1 -; GCN-IR-NEXT: v_mov_b32_e32 v13, v12 +; GCN-IR-NEXT: v_mov_b32_e32 v11, v10 ; GCN-IR-NEXT: v_mov_b32_e32 v5, 0 ; GCN-IR-NEXT: s_and_b64 s[4:5], s[4:5], s[6:7] ; GCN-IR-NEXT: s_and_saveexec_b64 s[6:7], s[4:5] ; GCN-IR-NEXT: s_cbranch_execz .LBB11_6 ; GCN-IR-NEXT: ; %bb.1: ; %udiv-bb1 ; GCN-IR-NEXT: v_add_i32_e32 v6, vcc, 1, v2 -; GCN-IR-NEXT: v_addc_u32_e32 v7, vcc, 0, v3, vcc +; GCN-IR-NEXT: v_addc_u32_e32 v3, vcc, 0, v3, vcc ; GCN-IR-NEXT: v_sub_i32_e64 v2, s[4:5], 63, v2 -; GCN-IR-NEXT: v_mov_b32_e32 v4, 0 -; GCN-IR-NEXT: v_cmp_ne_u64_e32 vcc, 0, v[6:7] ; GCN-IR-NEXT: v_lshl_b64 v[2:3], 24, v2 +; GCN-IR-NEXT: v_mov_b32_e32 v4, 0 ; GCN-IR-NEXT: v_mov_b32_e32 v5, 0 -; GCN-IR-NEXT: s_and_saveexec_b64 s[4:5], vcc -; GCN-IR-NEXT: s_xor_b64 s[8:9], exec, s[4:5] +; GCN-IR-NEXT: s_xor_b64 s[4:5], vcc, -1 +; GCN-IR-NEXT: s_and_saveexec_b64 s[8:9], s[4:5] +; GCN-IR-NEXT: s_xor_b64 s[4:5], exec, s[8:9] ; GCN-IR-NEXT: s_cbranch_execz .LBB11_5 ; GCN-IR-NEXT: ; %bb.2: ; %udiv-preheader -; GCN-IR-NEXT: v_add_i32_e32 v14, vcc, -1, v0 -; GCN-IR-NEXT: v_addc_u32_e32 v15, vcc, -1, v1, vcc -; GCN-IR-NEXT: v_lshr_b64 v[8:9], 24, v6 -; GCN-IR-NEXT: v_sub_i32_e32 v6, vcc, 58, v10 -; GCN-IR-NEXT: v_mov_b32_e32 v10, 0 -; GCN-IR-NEXT: v_subb_u32_e64 v7, s[4:5], 0, 0, vcc -; GCN-IR-NEXT: s_mov_b64 s[10:11], 0 -; GCN-IR-NEXT: v_mov_b32_e32 v11, 0 +; GCN-IR-NEXT: v_add_i32_e32 v12, vcc, -1, v0 +; GCN-IR-NEXT: v_addc_u32_e32 v13, vcc, -1, v1, vcc +; GCN-IR-NEXT: v_sub_i32_e32 v14, vcc, 58, v8 +; GCN-IR-NEXT: v_lshr_b64 v[6:7], 24, v6 +; GCN-IR-NEXT: v_subb_u32_e64 v15, s[8:9], 0, 0, vcc +; GCN-IR-NEXT: v_mov_b32_e32 v8, 0 +; GCN-IR-NEXT: s_mov_b64 s[8:9], 0 +; GCN-IR-NEXT: v_mov_b32_e32 v9, 0 ; GCN-IR-NEXT: v_mov_b32_e32 v5, 0 ; GCN-IR-NEXT: .LBB11_3: ; %udiv-do-while ; GCN-IR-NEXT: ; =>This Inner Loop Header: Depth=1 -; GCN-IR-NEXT: v_lshl_b64 v[8:9], v[8:9], 1 +; GCN-IR-NEXT: v_lshl_b64 v[6:7], v[6:7], 1 ; GCN-IR-NEXT: v_lshrrev_b32_e32 v4, 31, v3 -; GCN-IR-NEXT: v_or_b32_e32 v8, v8, v4 +; GCN-IR-NEXT: v_or_b32_e32 v6, v6, v4 ; GCN-IR-NEXT: v_lshl_b64 v[2:3], v[2:3], 1 -; GCN-IR-NEXT: v_sub_i32_e32 v4, vcc, v14, v8 -; GCN-IR-NEXT: v_subb_u32_e32 v4, vcc, v15, v9, vcc -; GCN-IR-NEXT: v_or_b32_e32 v2, v10, v2 -; GCN-IR-NEXT: v_ashrrev_i32_e32 v10, 31, v4 -; GCN-IR-NEXT: v_add_i32_e32 v6, vcc, 1, v6 -; GCN-IR-NEXT: v_or_b32_e32 v3, v11, v3 -; GCN-IR-NEXT: v_and_b32_e32 v4, 1, v10 -; GCN-IR-NEXT: v_and_b32_e32 v11, v10, v1 -; GCN-IR-NEXT: v_and_b32_e32 v10, v10, v0 -; GCN-IR-NEXT: v_addc_u32_e32 v7, vcc, 0, v7, vcc -; GCN-IR-NEXT: v_cmp_eq_u64_e32 vcc, 0, v[6:7] -; GCN-IR-NEXT: v_sub_i32_e64 v8, s[4:5], v8, v10 -; GCN-IR-NEXT: v_subb_u32_e64 v9, s[4:5], v9, v11, s[4:5] -; GCN-IR-NEXT: v_mov_b32_e32 v11, v5 -; GCN-IR-NEXT: s_or_b64 s[10:11], vcc, s[10:11] -; GCN-IR-NEXT: v_mov_b32_e32 v10, v4 -; GCN-IR-NEXT: s_andn2_b64 exec, exec, s[10:11] +; GCN-IR-NEXT: v_sub_i32_e32 v4, vcc, v12, v6 +; GCN-IR-NEXT: v_subb_u32_e32 v4, vcc, v13, v7, vcc +; GCN-IR-NEXT: v_or_b32_e32 v2, v8, v2 +; GCN-IR-NEXT: v_ashrrev_i32_e32 v8, 31, v4 +; GCN-IR-NEXT: v_or_b32_e32 v3, v9, v3 +; GCN-IR-NEXT: v_and_b32_e32 v4, 1, v8 +; GCN-IR-NEXT: v_and_b32_e32 v9, v8, v1 +; GCN-IR-NEXT: v_and_b32_e32 v8, v8, v0 +; GCN-IR-NEXT: v_sub_i32_e32 v6, vcc, v6, v8 +; GCN-IR-NEXT: v_subb_u32_e32 v7, vcc, v7, v9, vcc +; GCN-IR-NEXT: v_add_i32_e32 v14, vcc, 1, v14 +; GCN-IR-NEXT: v_addc_u32_e32 v15, vcc, 0, v15, vcc +; GCN-IR-NEXT: v_mov_b32_e32 v9, v5 +; GCN-IR-NEXT: s_or_b64 s[8:9], vcc, s[8:9] +; GCN-IR-NEXT: v_mov_b32_e32 v8, v4 +; GCN-IR-NEXT: s_andn2_b64 exec, exec, s[8:9] ; GCN-IR-NEXT: s_cbranch_execnz .LBB11_3 ; GCN-IR-NEXT: ; %bb.4: ; %Flow -; GCN-IR-NEXT: s_or_b64 exec, exec, s[10:11] -; GCN-IR-NEXT: .LBB11_5: ; %Flow4 ; GCN-IR-NEXT: s_or_b64 exec, exec, s[8:9] +; GCN-IR-NEXT: .LBB11_5: ; %Flow4 +; GCN-IR-NEXT: s_or_b64 exec, exec, s[4:5] ; GCN-IR-NEXT: v_lshl_b64 v[0:1], v[2:3], 1 ; GCN-IR-NEXT: v_or_b32_e32 v5, v5, v1 ; GCN-IR-NEXT: v_or_b32_e32 v4, v4, v0 ; GCN-IR-NEXT: .LBB11_6: ; %Flow5 ; GCN-IR-NEXT: s_or_b64 exec, exec, s[6:7] -; GCN-IR-NEXT: v_xor_b32_e32 v0, v4, v12 -; GCN-IR-NEXT: v_xor_b32_e32 v1, v5, v13 -; GCN-IR-NEXT: v_sub_i32_e32 v0, vcc, v0, v12 -; GCN-IR-NEXT: v_subb_u32_e32 v1, vcc, v1, v13, vcc +; GCN-IR-NEXT: v_xor_b32_e32 v0, v4, v10 +; GCN-IR-NEXT: v_xor_b32_e32 v1, v5, v11 +; GCN-IR-NEXT: v_sub_i32_e32 v0, vcc, v0, v10 +; GCN-IR-NEXT: v_subb_u32_e32 v1, vcc, v1, v11, vcc ; GCN-IR-NEXT: s_setpc_b64 s[30:31] %result = sdiv i64 24, %x ret i64 %result @@ -1667,17 +1677,17 @@ define i64 @v_test_sdiv_pow2_k_num_i64(i64 %x) { ; GCN-IR-LABEL: v_test_sdiv_pow2_k_num_i64: ; GCN-IR: ; %bb.0: ; %_udiv-special-cases ; GCN-IR-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) -; GCN-IR-NEXT: v_ashrrev_i32_e32 v12, 31, v1 -; GCN-IR-NEXT: v_xor_b32_e32 v0, v0, v12 -; GCN-IR-NEXT: v_xor_b32_e32 v1, v1, v12 -; GCN-IR-NEXT: v_sub_i32_e32 v0, vcc, v0, v12 -; GCN-IR-NEXT: v_subb_u32_e32 v1, vcc, v1, v12, vcc +; GCN-IR-NEXT: v_ashrrev_i32_e32 v10, 31, v1 +; GCN-IR-NEXT: v_xor_b32_e32 v0, v0, v10 +; GCN-IR-NEXT: v_xor_b32_e32 v1, v1, v10 +; GCN-IR-NEXT: v_sub_i32_e32 v0, vcc, v0, v10 +; GCN-IR-NEXT: v_subb_u32_e32 v1, vcc, v1, v10, vcc ; GCN-IR-NEXT: v_ffbh_u32_e32 v2, v0 ; GCN-IR-NEXT: v_add_i32_e32 v2, vcc, 32, v2 ; GCN-IR-NEXT: v_ffbh_u32_e32 v3, v1 -; GCN-IR-NEXT: v_min_u32_e32 v10, v2, v3 +; GCN-IR-NEXT: v_min_u32_e32 v8, v2, v3 ; GCN-IR-NEXT: s_movk_i32 s6, 0xffd0 -; GCN-IR-NEXT: v_add_i32_e32 v2, vcc, s6, v10 +; GCN-IR-NEXT: v_add_i32_e32 v2, vcc, s6, v8 ; GCN-IR-NEXT: v_addc_u32_e64 v3, s[6:7], 0, -1, vcc ; GCN-IR-NEXT: v_cmp_eq_u64_e64 s[4:5], 0, v[0:1] ; GCN-IR-NEXT: v_cmp_lt_u64_e32 vcc, 63, v[2:3] @@ -1686,70 +1696,69 @@ define i64 @v_test_sdiv_pow2_k_num_i64(i64 %x) { ; GCN-IR-NEXT: s_or_b64 s[4:5], s[4:5], vcc ; GCN-IR-NEXT: v_cndmask_b32_e64 v4, v4, 0, s[4:5] ; GCN-IR-NEXT: s_xor_b64 s[4:5], s[4:5], -1 -; GCN-IR-NEXT: v_mov_b32_e32 v13, v12 +; GCN-IR-NEXT: v_mov_b32_e32 v11, v10 ; GCN-IR-NEXT: v_mov_b32_e32 v5, 0 ; GCN-IR-NEXT: s_and_b64 s[4:5], s[4:5], s[6:7] ; GCN-IR-NEXT: s_and_saveexec_b64 s[6:7], s[4:5] ; GCN-IR-NEXT: s_cbranch_execz .LBB12_6 ; GCN-IR-NEXT: ; %bb.1: ; %udiv-bb1 ; GCN-IR-NEXT: v_add_i32_e32 v6, vcc, 1, v2 +; GCN-IR-NEXT: v_addc_u32_e32 v3, vcc, 0, v3, vcc ; GCN-IR-NEXT: v_sub_i32_e64 v2, s[4:5], 63, v2 -; GCN-IR-NEXT: v_addc_u32_e32 v7, vcc, 0, v3, vcc -; GCN-IR-NEXT: s_mov_b64 s[4:5], 0x8000 +; GCN-IR-NEXT: s_mov_b64 s[8:9], 0x8000 +; GCN-IR-NEXT: v_lshl_b64 v[2:3], s[8:9], v2 ; GCN-IR-NEXT: v_mov_b32_e32 v4, 0 -; GCN-IR-NEXT: v_cmp_ne_u64_e32 vcc, 0, v[6:7] -; GCN-IR-NEXT: v_lshl_b64 v[2:3], s[4:5], v2 ; GCN-IR-NEXT: v_mov_b32_e32 v5, 0 -; GCN-IR-NEXT: s_and_saveexec_b64 s[8:9], vcc -; GCN-IR-NEXT: s_xor_b64 s[8:9], exec, s[8:9] +; GCN-IR-NEXT: s_xor_b64 s[4:5], vcc, -1 +; GCN-IR-NEXT: s_and_saveexec_b64 s[10:11], s[4:5] +; GCN-IR-NEXT: s_xor_b64 s[4:5], exec, s[10:11] ; GCN-IR-NEXT: s_cbranch_execz .LBB12_5 ; GCN-IR-NEXT: ; %bb.2: ; %udiv-preheader -; GCN-IR-NEXT: v_add_i32_e32 v14, vcc, -1, v0 -; GCN-IR-NEXT: v_addc_u32_e32 v15, vcc, -1, v1, vcc -; GCN-IR-NEXT: v_lshr_b64 v[8:9], s[4:5], v6 -; GCN-IR-NEXT: v_sub_i32_e32 v6, vcc, 47, v10 -; GCN-IR-NEXT: v_mov_b32_e32 v10, 0 -; GCN-IR-NEXT: v_subb_u32_e64 v7, s[4:5], 0, 0, vcc -; GCN-IR-NEXT: s_mov_b64 s[10:11], 0 -; GCN-IR-NEXT: v_mov_b32_e32 v11, 0 +; GCN-IR-NEXT: v_add_i32_e32 v12, vcc, -1, v0 +; GCN-IR-NEXT: v_addc_u32_e32 v13, vcc, -1, v1, vcc +; GCN-IR-NEXT: v_sub_i32_e32 v14, vcc, 47, v8 +; GCN-IR-NEXT: v_lshr_b64 v[6:7], s[8:9], v6 +; GCN-IR-NEXT: v_subb_u32_e64 v15, s[8:9], 0, 0, vcc +; GCN-IR-NEXT: v_mov_b32_e32 v8, 0 +; GCN-IR-NEXT: s_mov_b64 s[8:9], 0 +; GCN-IR-NEXT: v_mov_b32_e32 v9, 0 ; GCN-IR-NEXT: v_mov_b32_e32 v5, 0 ; GCN-IR-NEXT: .LBB12_3: ; %udiv-do-while ; GCN-IR-NEXT: ; =>This Inner Loop Header: Depth=1 -; GCN-IR-NEXT: v_lshl_b64 v[8:9], v[8:9], 1 +; GCN-IR-NEXT: v_lshl_b64 v[6:7], v[6:7], 1 ; GCN-IR-NEXT: v_lshrrev_b32_e32 v4, 31, v3 -; GCN-IR-NEXT: v_or_b32_e32 v8, v8, v4 +; GCN-IR-NEXT: v_or_b32_e32 v6, v6, v4 ; GCN-IR-NEXT: v_lshl_b64 v[2:3], v[2:3], 1 -; GCN-IR-NEXT: v_sub_i32_e32 v4, vcc, v14, v8 -; GCN-IR-NEXT: v_subb_u32_e32 v4, vcc, v15, v9, vcc -; GCN-IR-NEXT: v_or_b32_e32 v2, v10, v2 -; GCN-IR-NEXT: v_ashrrev_i32_e32 v10, 31, v4 -; GCN-IR-NEXT: v_add_i32_e32 v6, vcc, 1, v6 -; GCN-IR-NEXT: v_or_b32_e32 v3, v11, v3 -; GCN-IR-NEXT: v_and_b32_e32 v4, 1, v10 -; GCN-IR-NEXT: v_and_b32_e32 v11, v10, v1 -; GCN-IR-NEXT: v_and_b32_e32 v10, v10, v0 -; GCN-IR-NEXT: v_addc_u32_e32 v7, vcc, 0, v7, vcc -; GCN-IR-NEXT: v_cmp_eq_u64_e32 vcc, 0, v[6:7] -; GCN-IR-NEXT: v_sub_i32_e64 v8, s[4:5], v8, v10 -; GCN-IR-NEXT: v_subb_u32_e64 v9, s[4:5], v9, v11, s[4:5] -; GCN-IR-NEXT: v_mov_b32_e32 v11, v5 -; GCN-IR-NEXT: s_or_b64 s[10:11], vcc, s[10:11] -; GCN-IR-NEXT: v_mov_b32_e32 v10, v4 -; GCN-IR-NEXT: s_andn2_b64 exec, exec, s[10:11] +; GCN-IR-NEXT: v_sub_i32_e32 v4, vcc, v12, v6 +; GCN-IR-NEXT: v_subb_u32_e32 v4, vcc, v13, v7, vcc +; GCN-IR-NEXT: v_or_b32_e32 v2, v8, v2 +; GCN-IR-NEXT: v_ashrrev_i32_e32 v8, 31, v4 +; GCN-IR-NEXT: v_or_b32_e32 v3, v9, v3 +; GCN-IR-NEXT: v_and_b32_e32 v4, 1, v8 +; GCN-IR-NEXT: v_and_b32_e32 v9, v8, v1 +; GCN-IR-NEXT: v_and_b32_e32 v8, v8, v0 +; GCN-IR-NEXT: v_sub_i32_e32 v6, vcc, v6, v8 +; GCN-IR-NEXT: v_subb_u32_e32 v7, vcc, v7, v9, vcc +; GCN-IR-NEXT: v_add_i32_e32 v14, vcc, 1, v14 +; GCN-IR-NEXT: v_addc_u32_e32 v15, vcc, 0, v15, vcc +; GCN-IR-NEXT: v_mov_b32_e32 v9, v5 +; GCN-IR-NEXT: s_or_b64 s[8:9], vcc, s[8:9] +; GCN-IR-NEXT: v_mov_b32_e32 v8, v4 +; GCN-IR-NEXT: s_andn2_b64 exec, exec, s[8:9] ; GCN-IR-NEXT: s_cbranch_execnz .LBB12_3 ; GCN-IR-NEXT: ; %bb.4: ; %Flow -; GCN-IR-NEXT: s_or_b64 exec, exec, s[10:11] -; GCN-IR-NEXT: .LBB12_5: ; %Flow4 ; GCN-IR-NEXT: s_or_b64 exec, exec, s[8:9] +; GCN-IR-NEXT: .LBB12_5: ; %Flow4 +; GCN-IR-NEXT: s_or_b64 exec, exec, s[4:5] ; GCN-IR-NEXT: v_lshl_b64 v[0:1], v[2:3], 1 ; GCN-IR-NEXT: v_or_b32_e32 v5, v5, v1 ; GCN-IR-NEXT: v_or_b32_e32 v4, v4, v0 ; GCN-IR-NEXT: .LBB12_6: ; %Flow5 ; GCN-IR-NEXT: s_or_b64 exec, exec, s[6:7] -; GCN-IR-NEXT: v_xor_b32_e32 v0, v4, v12 -; GCN-IR-NEXT: v_xor_b32_e32 v1, v5, v13 -; GCN-IR-NEXT: v_sub_i32_e32 v0, vcc, v0, v12 -; GCN-IR-NEXT: v_subb_u32_e32 v1, vcc, v1, v13, vcc +; GCN-IR-NEXT: v_xor_b32_e32 v0, v4, v10 +; GCN-IR-NEXT: v_xor_b32_e32 v1, v5, v11 +; GCN-IR-NEXT: v_sub_i32_e32 v0, vcc, v0, v10 +; GCN-IR-NEXT: v_subb_u32_e32 v1, vcc, v1, v11, vcc ; GCN-IR-NEXT: s_setpc_b64 s[30:31] %result = sdiv i64 32768, %x ret i64 %result @@ -1769,20 +1778,20 @@ define i64 @v_test_sdiv_pow2_k_den_i64(i64 %x) { ; GCN-IR-LABEL: v_test_sdiv_pow2_k_den_i64: ; GCN-IR: ; %bb.0: ; %_udiv-special-cases ; GCN-IR-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) -; GCN-IR-NEXT: v_ashrrev_i32_e32 v10, 31, v1 -; GCN-IR-NEXT: v_xor_b32_e32 v0, v0, v10 -; GCN-IR-NEXT: v_xor_b32_e32 v1, v1, v10 -; GCN-IR-NEXT: v_sub_i32_e32 v4, vcc, v0, v10 -; GCN-IR-NEXT: v_subb_u32_e32 v5, vcc, v1, v10, vcc +; GCN-IR-NEXT: v_ashrrev_i32_e32 v8, 31, v1 +; GCN-IR-NEXT: v_xor_b32_e32 v0, v0, v8 +; GCN-IR-NEXT: v_xor_b32_e32 v1, v1, v8 +; GCN-IR-NEXT: v_sub_i32_e32 v4, vcc, v0, v8 +; GCN-IR-NEXT: v_subb_u32_e32 v5, vcc, v1, v8, vcc ; GCN-IR-NEXT: v_ffbh_u32_e32 v0, v4 ; GCN-IR-NEXT: v_add_i32_e64 v0, s[4:5], 32, v0 ; GCN-IR-NEXT: v_ffbh_u32_e32 v1, v5 -; GCN-IR-NEXT: v_min_u32_e32 v8, v0, v1 -; GCN-IR-NEXT: v_sub_i32_e64 v0, s[4:5], 48, v8 +; GCN-IR-NEXT: v_min_u32_e32 v6, v0, v1 +; GCN-IR-NEXT: v_sub_i32_e64 v0, s[4:5], 48, v6 ; GCN-IR-NEXT: v_subb_u32_e64 v1, s[4:5], 0, 0, s[4:5] ; GCN-IR-NEXT: v_cmp_eq_u64_e32 vcc, 0, v[4:5] ; GCN-IR-NEXT: v_cmp_lt_u64_e64 s[4:5], 63, v[0:1] -; GCN-IR-NEXT: v_mov_b32_e32 v11, v10 +; GCN-IR-NEXT: v_mov_b32_e32 v9, v8 ; GCN-IR-NEXT: s_or_b64 s[4:5], vcc, s[4:5] ; GCN-IR-NEXT: v_cmp_ne_u64_e32 vcc, 63, v[0:1] ; GCN-IR-NEXT: s_xor_b64 s[6:7], s[4:5], -1 @@ -1792,61 +1801,60 @@ define i64 @v_test_sdiv_pow2_k_den_i64(i64 %x) { ; GCN-IR-NEXT: s_and_saveexec_b64 s[6:7], s[4:5] ; GCN-IR-NEXT: s_cbranch_execz .LBB13_6 ; GCN-IR-NEXT: ; %bb.1: ; %udiv-bb1 -; GCN-IR-NEXT: v_add_i32_e32 v6, vcc, 1, v0 -; GCN-IR-NEXT: v_addc_u32_e32 v7, vcc, 0, v1, vcc +; GCN-IR-NEXT: v_add_i32_e32 v7, vcc, 1, v0 +; GCN-IR-NEXT: v_addc_u32_e32 v1, vcc, 0, v1, vcc ; GCN-IR-NEXT: v_sub_i32_e64 v0, s[4:5], 63, v0 -; GCN-IR-NEXT: v_mov_b32_e32 v2, 0 -; GCN-IR-NEXT: v_cmp_ne_u64_e32 vcc, 0, v[6:7] ; GCN-IR-NEXT: v_lshl_b64 v[0:1], v[4:5], v0 +; GCN-IR-NEXT: v_mov_b32_e32 v2, 0 ; GCN-IR-NEXT: v_mov_b32_e32 v3, 0 -; GCN-IR-NEXT: s_and_saveexec_b64 s[4:5], vcc -; GCN-IR-NEXT: s_xor_b64 s[8:9], exec, s[4:5] +; GCN-IR-NEXT: s_xor_b64 s[4:5], vcc, -1 +; GCN-IR-NEXT: s_and_saveexec_b64 s[8:9], s[4:5] +; GCN-IR-NEXT: s_xor_b64 s[4:5], exec, s[8:9] ; GCN-IR-NEXT: s_cbranch_execz .LBB13_5 ; GCN-IR-NEXT: ; %bb.2: ; %udiv-preheader -; GCN-IR-NEXT: v_lshr_b64 v[6:7], v[4:5], v6 -; GCN-IR-NEXT: v_add_i32_e32 v4, vcc, 0xffffffcf, v8 -; GCN-IR-NEXT: v_mov_b32_e32 v8, 0 -; GCN-IR-NEXT: v_addc_u32_e64 v5, s[4:5], 0, -1, vcc -; GCN-IR-NEXT: s_mov_b64 s[10:11], 0 -; GCN-IR-NEXT: v_mov_b32_e32 v9, 0 +; GCN-IR-NEXT: v_add_i32_e32 v10, vcc, 0xffffffcf, v6 +; GCN-IR-NEXT: v_lshr_b64 v[4:5], v[4:5], v7 +; GCN-IR-NEXT: v_addc_u32_e64 v11, s[8:9], 0, -1, vcc +; GCN-IR-NEXT: v_mov_b32_e32 v6, 0 +; GCN-IR-NEXT: s_mov_b64 s[8:9], 0 +; GCN-IR-NEXT: v_mov_b32_e32 v7, 0 ; GCN-IR-NEXT: v_mov_b32_e32 v3, 0 -; GCN-IR-NEXT: s_movk_i32 s12, 0x7fff +; GCN-IR-NEXT: s_movk_i32 s10, 0x7fff ; GCN-IR-NEXT: .LBB13_3: ; %udiv-do-while ; GCN-IR-NEXT: ; =>This Inner Loop Header: Depth=1 -; GCN-IR-NEXT: v_lshl_b64 v[6:7], v[6:7], 1 +; GCN-IR-NEXT: v_lshl_b64 v[4:5], v[4:5], 1 ; GCN-IR-NEXT: v_lshrrev_b32_e32 v2, 31, v1 -; GCN-IR-NEXT: v_or_b32_e32 v6, v6, v2 -; GCN-IR-NEXT: v_sub_i32_e32 v2, vcc, s12, v6 +; GCN-IR-NEXT: v_or_b32_e32 v4, v4, v2 ; GCN-IR-NEXT: v_lshl_b64 v[0:1], v[0:1], 1 -; GCN-IR-NEXT: v_subb_u32_e32 v2, vcc, 0, v7, vcc -; GCN-IR-NEXT: v_add_i32_e32 v4, vcc, 1, v4 -; GCN-IR-NEXT: v_or_b32_e32 v0, v8, v0 -; GCN-IR-NEXT: v_ashrrev_i32_e32 v8, 31, v2 -; GCN-IR-NEXT: v_addc_u32_e32 v5, vcc, 0, v5, vcc -; GCN-IR-NEXT: v_and_b32_e32 v2, 1, v8 -; GCN-IR-NEXT: v_and_b32_e32 v8, 0x8000, v8 -; GCN-IR-NEXT: v_cmp_eq_u64_e32 vcc, 0, v[4:5] -; GCN-IR-NEXT: v_or_b32_e32 v1, v9, v1 -; GCN-IR-NEXT: v_sub_i32_e64 v6, s[4:5], v6, v8 -; GCN-IR-NEXT: v_mov_b32_e32 v9, v3 -; GCN-IR-NEXT: v_subbrev_u32_e64 v7, s[4:5], 0, v7, s[4:5] -; GCN-IR-NEXT: s_or_b64 s[10:11], vcc, s[10:11] -; GCN-IR-NEXT: v_mov_b32_e32 v8, v2 -; GCN-IR-NEXT: s_andn2_b64 exec, exec, s[10:11] +; GCN-IR-NEXT: v_sub_i32_e32 v2, vcc, s10, v4 +; GCN-IR-NEXT: v_subb_u32_e32 v2, vcc, 0, v5, vcc +; GCN-IR-NEXT: v_or_b32_e32 v0, v6, v0 +; GCN-IR-NEXT: v_ashrrev_i32_e32 v6, 31, v2 +; GCN-IR-NEXT: v_and_b32_e32 v2, 1, v6 +; GCN-IR-NEXT: v_and_b32_e32 v6, 0x8000, v6 +; GCN-IR-NEXT: v_sub_i32_e32 v4, vcc, v4, v6 +; GCN-IR-NEXT: v_subbrev_u32_e32 v5, vcc, 0, v5, vcc +; GCN-IR-NEXT: v_add_i32_e32 v10, vcc, 1, v10 +; GCN-IR-NEXT: v_or_b32_e32 v1, v7, v1 +; GCN-IR-NEXT: v_addc_u32_e32 v11, vcc, 0, v11, vcc +; GCN-IR-NEXT: v_mov_b32_e32 v7, v3 +; GCN-IR-NEXT: s_or_b64 s[8:9], vcc, s[8:9] +; GCN-IR-NEXT: v_mov_b32_e32 v6, v2 +; GCN-IR-NEXT: s_andn2_b64 exec, exec, s[8:9] ; GCN-IR-NEXT: s_cbranch_execnz .LBB13_3 ; GCN-IR-NEXT: ; %bb.4: ; %Flow -; GCN-IR-NEXT: s_or_b64 exec, exec, s[10:11] -; GCN-IR-NEXT: .LBB13_5: ; %Flow4 ; GCN-IR-NEXT: s_or_b64 exec, exec, s[8:9] +; GCN-IR-NEXT: .LBB13_5: ; %Flow4 +; GCN-IR-NEXT: s_or_b64 exec, exec, s[4:5] ; GCN-IR-NEXT: v_lshl_b64 v[0:1], v[0:1], 1 ; GCN-IR-NEXT: v_or_b32_e32 v3, v3, v1 ; GCN-IR-NEXT: v_or_b32_e32 v2, v2, v0 ; GCN-IR-NEXT: .LBB13_6: ; %Flow5 ; GCN-IR-NEXT: s_or_b64 exec, exec, s[6:7] -; GCN-IR-NEXT: v_xor_b32_e32 v0, v2, v10 -; GCN-IR-NEXT: v_xor_b32_e32 v1, v3, v11 -; GCN-IR-NEXT: v_sub_i32_e32 v0, vcc, v0, v10 -; GCN-IR-NEXT: v_subb_u32_e32 v1, vcc, v1, v11, vcc +; GCN-IR-NEXT: v_xor_b32_e32 v0, v2, v8 +; GCN-IR-NEXT: v_xor_b32_e32 v1, v3, v9 +; GCN-IR-NEXT: v_sub_i32_e32 v0, vcc, v0, v8 +; GCN-IR-NEXT: v_subb_u32_e32 v1, vcc, v1, v9, vcc ; GCN-IR-NEXT: s_setpc_b64 s[30:31] %result = sdiv i64 %x, 32768 ret i64 %result diff --git a/llvm/test/CodeGen/AMDGPU/srem64.ll b/llvm/test/CodeGen/AMDGPU/srem64.ll index 2d95875cad882..fd067024fac1f 100644 --- a/llvm/test/CodeGen/AMDGPU/srem64.ll +++ b/llvm/test/CodeGen/AMDGPU/srem64.ll @@ -170,35 +170,38 @@ define amdgpu_kernel void @s_test_srem(ptr addrspace(1) %out, i64 %x, i64 %y) { ; GCN-IR-NEXT: v_cmp_eq_u64_e64 s[8:9], s[6:7], 0 ; GCN-IR-NEXT: v_cmp_eq_u64_e64 s[12:13], s[2:3], 0 ; GCN-IR-NEXT: s_flbit_i32_b64 s10, s[6:7] -; GCN-IR-NEXT: s_flbit_i32_b64 s18, s[2:3] +; GCN-IR-NEXT: s_flbit_i32_b64 s16, s[2:3] ; GCN-IR-NEXT: s_or_b64 s[8:9], s[8:9], s[12:13] -; GCN-IR-NEXT: s_sub_u32 s12, s10, s18 +; GCN-IR-NEXT: s_sub_u32 s12, s10, s16 ; GCN-IR-NEXT: s_subb_u32 s13, 0, 0 ; GCN-IR-NEXT: v_cmp_gt_u64_e64 s[14:15], s[12:13], 63 -; GCN-IR-NEXT: v_cmp_eq_u64_e64 s[16:17], s[12:13], 63 +; GCN-IR-NEXT: v_cmp_eq_u64_e64 s[18:19], s[12:13], 63 ; GCN-IR-NEXT: s_or_b64 s[14:15], s[8:9], s[14:15] ; GCN-IR-NEXT: s_and_b64 s[8:9], s[14:15], exec ; GCN-IR-NEXT: s_cselect_b32 s9, 0, s3 ; GCN-IR-NEXT: s_cselect_b32 s8, 0, s2 -; GCN-IR-NEXT: s_or_b64 s[14:15], s[14:15], s[16:17] +; GCN-IR-NEXT: s_or_b64 s[14:15], s[14:15], s[18:19] ; GCN-IR-NEXT: s_andn2_b64 vcc, exec, s[14:15] ; GCN-IR-NEXT: s_cbranch_vccz .LBB0_5 ; GCN-IR-NEXT: ; %bb.1: ; %udiv-bb1 ; GCN-IR-NEXT: s_add_u32 s14, s12, 1 -; GCN-IR-NEXT: s_addc_u32 s15, s13, 0 -; GCN-IR-NEXT: v_cmp_eq_u64_e64 s[8:9], s[14:15], 0 +; GCN-IR-NEXT: s_cselect_b64 s[8:9], -1, 0 +; GCN-IR-NEXT: s_or_b32 s8, s8, s9 +; GCN-IR-NEXT: s_cmp_lg_u32 s8, 0 +; GCN-IR-NEXT: s_addc_u32 s8, s13, 0 +; GCN-IR-NEXT: s_cselect_b64 s[8:9], -1, 0 ; GCN-IR-NEXT: s_sub_i32 s12, 63, s12 ; GCN-IR-NEXT: s_andn2_b64 vcc, exec, s[8:9] ; GCN-IR-NEXT: s_lshl_b64 s[8:9], s[2:3], s12 ; GCN-IR-NEXT: s_cbranch_vccz .LBB0_4 ; GCN-IR-NEXT: ; %bb.2: ; %udiv-preheader ; GCN-IR-NEXT: s_lshr_b64 s[12:13], s[2:3], s14 -; GCN-IR-NEXT: s_add_u32 s16, s6, -1 -; GCN-IR-NEXT: s_addc_u32 s17, s7, -1 +; GCN-IR-NEXT: s_add_u32 s14, s6, -1 +; GCN-IR-NEXT: s_addc_u32 s15, s7, -1 ; GCN-IR-NEXT: s_not_b64 s[4:5], s[10:11] -; GCN-IR-NEXT: s_add_u32 s10, s4, s18 -; GCN-IR-NEXT: s_addc_u32 s11, s5, 0 -; GCN-IR-NEXT: s_mov_b64 s[14:15], 0 +; GCN-IR-NEXT: s_add_u32 s16, s4, s16 +; GCN-IR-NEXT: s_addc_u32 s17, s5, 0 +; GCN-IR-NEXT: s_mov_b64 s[10:11], 0 ; GCN-IR-NEXT: s_mov_b32 s5, 0 ; GCN-IR-NEXT: .LBB0_3: ; %udiv-do-while ; GCN-IR-NEXT: ; =>This Inner Loop Header: Depth=1 @@ -206,19 +209,22 @@ define amdgpu_kernel void @s_test_srem(ptr addrspace(1) %out, i64 %x, i64 %y) { ; GCN-IR-NEXT: s_lshr_b32 s4, s9, 31 ; GCN-IR-NEXT: s_lshl_b64 s[8:9], s[8:9], 1 ; GCN-IR-NEXT: s_or_b64 s[12:13], s[12:13], s[4:5] -; GCN-IR-NEXT: s_or_b64 s[8:9], s[14:15], s[8:9] -; GCN-IR-NEXT: s_sub_u32 s4, s16, s12 -; GCN-IR-NEXT: s_subb_u32 s4, s17, s13 -; GCN-IR-NEXT: s_ashr_i32 s14, s4, 31 -; GCN-IR-NEXT: s_mov_b32 s15, s14 -; GCN-IR-NEXT: s_and_b32 s4, s14, 1 -; GCN-IR-NEXT: s_and_b64 s[14:15], s[14:15], s[6:7] -; GCN-IR-NEXT: s_sub_u32 s12, s12, s14 -; GCN-IR-NEXT: s_subb_u32 s13, s13, s15 -; GCN-IR-NEXT: s_add_u32 s10, s10, 1 -; GCN-IR-NEXT: s_addc_u32 s11, s11, 0 -; GCN-IR-NEXT: v_cmp_eq_u64_e64 s[18:19], s[10:11], 0 -; GCN-IR-NEXT: s_mov_b64 s[14:15], s[4:5] +; GCN-IR-NEXT: s_or_b64 s[8:9], s[10:11], s[8:9] +; GCN-IR-NEXT: s_sub_u32 s4, s14, s12 +; GCN-IR-NEXT: s_subb_u32 s4, s15, s13 +; GCN-IR-NEXT: s_ashr_i32 s10, s4, 31 +; GCN-IR-NEXT: s_mov_b32 s11, s10 +; GCN-IR-NEXT: s_and_b32 s4, s10, 1 +; GCN-IR-NEXT: s_and_b64 s[18:19], s[10:11], s[6:7] +; GCN-IR-NEXT: s_sub_u32 s12, s12, s18 +; GCN-IR-NEXT: s_subb_u32 s13, s13, s19 +; GCN-IR-NEXT: s_add_u32 s16, s16, 1 +; GCN-IR-NEXT: s_cselect_b64 s[18:19], -1, 0 +; GCN-IR-NEXT: s_or_b32 s18, s18, s19 +; GCN-IR-NEXT: s_cmp_lg_u32 s18, 0 +; GCN-IR-NEXT: s_addc_u32 s17, s17, 0 +; GCN-IR-NEXT: s_cselect_b64 s[18:19], -1, 0 +; GCN-IR-NEXT: s_mov_b64 s[10:11], s[4:5] ; GCN-IR-NEXT: s_and_b64 vcc, exec, s[18:19] ; GCN-IR-NEXT: s_cbranch_vccz .LBB0_3 ; GCN-IR-NEXT: .LBB0_4: ; %Flow7 @@ -373,12 +379,12 @@ define i64 @v_test_srem(i64 %x, i64 %y) { ; GCN-IR-LABEL: v_test_srem: ; GCN-IR: ; %bb.0: ; %_udiv-special-cases ; GCN-IR-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) -; GCN-IR-NEXT: v_ashrrev_i32_e32 v14, 31, v1 -; GCN-IR-NEXT: v_xor_b32_e32 v0, v0, v14 -; GCN-IR-NEXT: v_xor_b32_e32 v1, v1, v14 -; GCN-IR-NEXT: v_sub_i32_e32 v0, vcc, v0, v14 +; GCN-IR-NEXT: v_ashrrev_i32_e32 v12, 31, v1 +; GCN-IR-NEXT: v_xor_b32_e32 v0, v0, v12 +; GCN-IR-NEXT: v_xor_b32_e32 v1, v1, v12 +; GCN-IR-NEXT: v_sub_i32_e32 v0, vcc, v0, v12 ; GCN-IR-NEXT: v_ashrrev_i32_e32 v4, 31, v3 -; GCN-IR-NEXT: v_subb_u32_e32 v1, vcc, v1, v14, vcc +; GCN-IR-NEXT: v_subb_u32_e32 v1, vcc, v1, v12, vcc ; GCN-IR-NEXT: v_xor_b32_e32 v2, v2, v4 ; GCN-IR-NEXT: v_xor_b32_e32 v3, v3, v4 ; GCN-IR-NEXT: v_sub_i32_e32 v2, vcc, v2, v4 @@ -386,12 +392,12 @@ define i64 @v_test_srem(i64 %x, i64 %y) { ; GCN-IR-NEXT: v_ffbh_u32_e32 v4, v2 ; GCN-IR-NEXT: v_add_i32_e64 v4, s[6:7], 32, v4 ; GCN-IR-NEXT: v_ffbh_u32_e32 v5, v3 -; GCN-IR-NEXT: v_min_u32_e32 v12, v4, v5 +; GCN-IR-NEXT: v_min_u32_e32 v10, v4, v5 ; GCN-IR-NEXT: v_ffbh_u32_e32 v4, v0 ; GCN-IR-NEXT: v_add_i32_e64 v4, s[6:7], 32, v4 ; GCN-IR-NEXT: v_ffbh_u32_e32 v5, v1 -; GCN-IR-NEXT: v_min_u32_e32 v13, v4, v5 -; GCN-IR-NEXT: v_sub_i32_e64 v4, s[6:7], v12, v13 +; GCN-IR-NEXT: v_min_u32_e32 v11, v4, v5 +; GCN-IR-NEXT: v_sub_i32_e64 v4, s[6:7], v10, v11 ; GCN-IR-NEXT: v_cmp_eq_u64_e32 vcc, 0, v[2:3] ; GCN-IR-NEXT: v_cmp_eq_u64_e64 s[4:5], 0, v[0:1] ; GCN-IR-NEXT: v_subb_u32_e64 v5, s[6:7], 0, 0, s[6:7] @@ -400,7 +406,7 @@ define i64 @v_test_srem(i64 %x, i64 %y) { ; GCN-IR-NEXT: s_or_b64 s[4:5], s[4:5], s[6:7] ; GCN-IR-NEXT: v_cmp_ne_u64_e32 vcc, 63, v[4:5] ; GCN-IR-NEXT: s_xor_b64 s[6:7], s[4:5], -1 -; GCN-IR-NEXT: v_mov_b32_e32 v15, v14 +; GCN-IR-NEXT: v_mov_b32_e32 v13, v12 ; GCN-IR-NEXT: v_cndmask_b32_e64 v7, v1, 0, s[4:5] ; GCN-IR-NEXT: v_cndmask_b32_e64 v6, v0, 0, s[4:5] ; GCN-IR-NEXT: s_and_b64 s[4:5], s[6:7], vcc @@ -408,54 +414,53 @@ define i64 @v_test_srem(i64 %x, i64 %y) { ; GCN-IR-NEXT: s_cbranch_execz .LBB1_6 ; GCN-IR-NEXT: ; %bb.1: ; %udiv-bb1 ; GCN-IR-NEXT: v_add_i32_e32 v8, vcc, 1, v4 -; GCN-IR-NEXT: v_addc_u32_e32 v9, vcc, 0, v5, vcc +; GCN-IR-NEXT: v_addc_u32_e32 v5, vcc, 0, v5, vcc ; GCN-IR-NEXT: v_sub_i32_e64 v4, s[4:5], 63, v4 -; GCN-IR-NEXT: v_mov_b32_e32 v6, 0 -; GCN-IR-NEXT: v_cmp_ne_u64_e32 vcc, 0, v[8:9] ; GCN-IR-NEXT: v_lshl_b64 v[4:5], v[0:1], v4 +; GCN-IR-NEXT: v_mov_b32_e32 v6, 0 ; GCN-IR-NEXT: v_mov_b32_e32 v7, 0 -; GCN-IR-NEXT: s_and_saveexec_b64 s[4:5], vcc -; GCN-IR-NEXT: s_xor_b64 s[8:9], exec, s[4:5] +; GCN-IR-NEXT: s_xor_b64 s[4:5], vcc, -1 +; GCN-IR-NEXT: s_and_saveexec_b64 s[8:9], s[4:5] +; GCN-IR-NEXT: s_xor_b64 s[4:5], exec, s[8:9] ; GCN-IR-NEXT: s_cbranch_execz .LBB1_5 ; GCN-IR-NEXT: ; %bb.2: ; %udiv-preheader -; GCN-IR-NEXT: v_add_i32_e32 v16, vcc, -1, v2 -; GCN-IR-NEXT: v_addc_u32_e32 v17, vcc, -1, v3, vcc -; GCN-IR-NEXT: v_not_b32_e32 v6, v12 -; GCN-IR-NEXT: v_lshr_b64 v[10:11], v[0:1], v8 -; GCN-IR-NEXT: v_add_i32_e32 v8, vcc, v6, v13 -; GCN-IR-NEXT: v_mov_b32_e32 v12, 0 -; GCN-IR-NEXT: v_addc_u32_e64 v9, s[4:5], -1, 0, vcc -; GCN-IR-NEXT: s_mov_b64 s[10:11], 0 -; GCN-IR-NEXT: v_mov_b32_e32 v13, 0 +; GCN-IR-NEXT: v_add_i32_e32 v14, vcc, -1, v2 +; GCN-IR-NEXT: v_addc_u32_e32 v15, vcc, -1, v3, vcc +; GCN-IR-NEXT: v_not_b32_e32 v6, v10 +; GCN-IR-NEXT: v_add_i32_e32 v16, vcc, v6, v11 +; GCN-IR-NEXT: v_lshr_b64 v[8:9], v[0:1], v8 +; GCN-IR-NEXT: v_addc_u32_e64 v17, s[8:9], -1, 0, vcc +; GCN-IR-NEXT: v_mov_b32_e32 v10, 0 +; GCN-IR-NEXT: s_mov_b64 s[8:9], 0 +; GCN-IR-NEXT: v_mov_b32_e32 v11, 0 ; GCN-IR-NEXT: v_mov_b32_e32 v7, 0 ; GCN-IR-NEXT: .LBB1_3: ; %udiv-do-while ; GCN-IR-NEXT: ; =>This Inner Loop Header: Depth=1 -; GCN-IR-NEXT: v_lshl_b64 v[10:11], v[10:11], 1 +; GCN-IR-NEXT: v_lshl_b64 v[8:9], v[8:9], 1 ; GCN-IR-NEXT: v_lshrrev_b32_e32 v6, 31, v5 -; GCN-IR-NEXT: v_or_b32_e32 v10, v10, v6 +; GCN-IR-NEXT: v_or_b32_e32 v8, v8, v6 ; GCN-IR-NEXT: v_lshl_b64 v[4:5], v[4:5], 1 -; GCN-IR-NEXT: v_sub_i32_e32 v6, vcc, v16, v10 -; GCN-IR-NEXT: v_subb_u32_e32 v6, vcc, v17, v11, vcc -; GCN-IR-NEXT: v_or_b32_e32 v4, v12, v4 -; GCN-IR-NEXT: v_ashrrev_i32_e32 v12, 31, v6 -; GCN-IR-NEXT: v_add_i32_e32 v8, vcc, 1, v8 -; GCN-IR-NEXT: v_or_b32_e32 v5, v13, v5 -; GCN-IR-NEXT: v_and_b32_e32 v6, 1, v12 -; GCN-IR-NEXT: v_and_b32_e32 v13, v12, v3 -; GCN-IR-NEXT: v_and_b32_e32 v12, v12, v2 -; GCN-IR-NEXT: v_addc_u32_e32 v9, vcc, 0, v9, vcc -; GCN-IR-NEXT: v_cmp_eq_u64_e32 vcc, 0, v[8:9] -; GCN-IR-NEXT: v_sub_i32_e64 v10, s[4:5], v10, v12 -; GCN-IR-NEXT: v_subb_u32_e64 v11, s[4:5], v11, v13, s[4:5] -; GCN-IR-NEXT: v_mov_b32_e32 v13, v7 -; GCN-IR-NEXT: s_or_b64 s[10:11], vcc, s[10:11] -; GCN-IR-NEXT: v_mov_b32_e32 v12, v6 -; GCN-IR-NEXT: s_andn2_b64 exec, exec, s[10:11] +; GCN-IR-NEXT: v_sub_i32_e32 v6, vcc, v14, v8 +; GCN-IR-NEXT: v_subb_u32_e32 v6, vcc, v15, v9, vcc +; GCN-IR-NEXT: v_or_b32_e32 v4, v10, v4 +; GCN-IR-NEXT: v_ashrrev_i32_e32 v10, 31, v6 +; GCN-IR-NEXT: v_or_b32_e32 v5, v11, v5 +; GCN-IR-NEXT: v_and_b32_e32 v6, 1, v10 +; GCN-IR-NEXT: v_and_b32_e32 v11, v10, v3 +; GCN-IR-NEXT: v_and_b32_e32 v10, v10, v2 +; GCN-IR-NEXT: v_sub_i32_e32 v8, vcc, v8, v10 +; GCN-IR-NEXT: v_subb_u32_e32 v9, vcc, v9, v11, vcc +; GCN-IR-NEXT: v_add_i32_e32 v16, vcc, 1, v16 +; GCN-IR-NEXT: v_addc_u32_e32 v17, vcc, 0, v17, vcc +; GCN-IR-NEXT: v_mov_b32_e32 v11, v7 +; GCN-IR-NEXT: s_or_b64 s[8:9], vcc, s[8:9] +; GCN-IR-NEXT: v_mov_b32_e32 v10, v6 +; GCN-IR-NEXT: s_andn2_b64 exec, exec, s[8:9] ; GCN-IR-NEXT: s_cbranch_execnz .LBB1_3 ; GCN-IR-NEXT: ; %bb.4: ; %Flow -; GCN-IR-NEXT: s_or_b64 exec, exec, s[10:11] -; GCN-IR-NEXT: .LBB1_5: ; %Flow4 ; GCN-IR-NEXT: s_or_b64 exec, exec, s[8:9] +; GCN-IR-NEXT: .LBB1_5: ; %Flow4 +; GCN-IR-NEXT: s_or_b64 exec, exec, s[4:5] ; GCN-IR-NEXT: v_lshl_b64 v[4:5], v[4:5], 1 ; GCN-IR-NEXT: v_or_b32_e32 v7, v7, v5 ; GCN-IR-NEXT: v_or_b32_e32 v6, v6, v4 @@ -469,10 +474,10 @@ define i64 @v_test_srem(i64 %x, i64 %y) { ; GCN-IR-NEXT: v_add_i32_e32 v3, vcc, v4, v3 ; GCN-IR-NEXT: v_sub_i32_e32 v0, vcc, v0, v2 ; GCN-IR-NEXT: v_subb_u32_e32 v1, vcc, v1, v3, vcc -; GCN-IR-NEXT: v_xor_b32_e32 v0, v0, v14 -; GCN-IR-NEXT: v_xor_b32_e32 v1, v1, v15 -; GCN-IR-NEXT: v_sub_i32_e32 v0, vcc, v0, v14 -; GCN-IR-NEXT: v_subb_u32_e32 v1, vcc, v1, v15, vcc +; GCN-IR-NEXT: v_xor_b32_e32 v0, v0, v12 +; GCN-IR-NEXT: v_xor_b32_e32 v1, v1, v13 +; GCN-IR-NEXT: v_sub_i32_e32 v0, vcc, v0, v12 +; GCN-IR-NEXT: v_subb_u32_e32 v1, vcc, v1, v13, vcc ; GCN-IR-NEXT: s_setpc_b64 s[30:31] %result = srem i64 %x, %y ret i64 %result @@ -1148,35 +1153,38 @@ define amdgpu_kernel void @s_test_srem33_64(ptr addrspace(1) %out, i64 %x, i64 % ; GCN-IR-NEXT: v_cmp_eq_u64_e64 s[2:3], s[8:9], 0 ; GCN-IR-NEXT: s_flbit_i32_b64 s12, s[8:9] ; GCN-IR-NEXT: s_or_b64 s[10:11], s[2:3], s[10:11] -; GCN-IR-NEXT: s_flbit_i32_b64 s20, s[6:7] -; GCN-IR-NEXT: s_sub_u32 s14, s12, s20 +; GCN-IR-NEXT: s_flbit_i32_b64 s18, s[6:7] +; GCN-IR-NEXT: s_sub_u32 s14, s12, s18 ; GCN-IR-NEXT: s_subb_u32 s15, 0, 0 ; GCN-IR-NEXT: v_cmp_gt_u64_e64 s[16:17], s[14:15], 63 -; GCN-IR-NEXT: v_cmp_eq_u64_e64 s[18:19], s[14:15], 63 +; GCN-IR-NEXT: v_cmp_eq_u64_e64 s[20:21], s[14:15], 63 ; GCN-IR-NEXT: s_or_b64 s[16:17], s[10:11], s[16:17] ; GCN-IR-NEXT: s_and_b64 s[10:11], s[16:17], exec ; GCN-IR-NEXT: s_cselect_b32 s11, 0, s7 ; GCN-IR-NEXT: s_cselect_b32 s10, 0, s6 -; GCN-IR-NEXT: s_or_b64 s[16:17], s[16:17], s[18:19] +; GCN-IR-NEXT: s_or_b64 s[16:17], s[16:17], s[20:21] ; GCN-IR-NEXT: s_mov_b64 s[2:3], 0 ; GCN-IR-NEXT: s_andn2_b64 vcc, exec, s[16:17] ; GCN-IR-NEXT: s_cbranch_vccz .LBB8_5 ; GCN-IR-NEXT: ; %bb.1: ; %udiv-bb1 ; GCN-IR-NEXT: s_add_u32 s16, s14, 1 -; GCN-IR-NEXT: s_addc_u32 s17, s15, 0 -; GCN-IR-NEXT: v_cmp_eq_u64_e64 s[10:11], s[16:17], 0 +; GCN-IR-NEXT: s_cselect_b64 s[10:11], -1, 0 +; GCN-IR-NEXT: s_or_b32 s10, s10, s11 +; GCN-IR-NEXT: s_cmp_lg_u32 s10, 0 +; GCN-IR-NEXT: s_addc_u32 s10, s15, 0 +; GCN-IR-NEXT: s_cselect_b64 s[10:11], -1, 0 ; GCN-IR-NEXT: s_sub_i32 s14, 63, s14 ; GCN-IR-NEXT: s_andn2_b64 vcc, exec, s[10:11] ; GCN-IR-NEXT: s_lshl_b64 s[10:11], s[6:7], s14 ; GCN-IR-NEXT: s_cbranch_vccz .LBB8_4 ; GCN-IR-NEXT: ; %bb.2: ; %udiv-preheader ; GCN-IR-NEXT: s_lshr_b64 s[14:15], s[6:7], s16 -; GCN-IR-NEXT: s_add_u32 s18, s8, -1 -; GCN-IR-NEXT: s_addc_u32 s19, s9, -1 +; GCN-IR-NEXT: s_add_u32 s16, s8, -1 +; GCN-IR-NEXT: s_addc_u32 s17, s9, -1 ; GCN-IR-NEXT: s_not_b64 s[2:3], s[12:13] -; GCN-IR-NEXT: s_add_u32 s12, s2, s20 -; GCN-IR-NEXT: s_addc_u32 s13, s3, 0 -; GCN-IR-NEXT: s_mov_b64 s[16:17], 0 +; GCN-IR-NEXT: s_add_u32 s18, s2, s18 +; GCN-IR-NEXT: s_addc_u32 s19, s3, 0 +; GCN-IR-NEXT: s_mov_b64 s[12:13], 0 ; GCN-IR-NEXT: s_mov_b32 s3, 0 ; GCN-IR-NEXT: .LBB8_3: ; %udiv-do-while ; GCN-IR-NEXT: ; =>This Inner Loop Header: Depth=1 @@ -1184,19 +1192,22 @@ define amdgpu_kernel void @s_test_srem33_64(ptr addrspace(1) %out, i64 %x, i64 % ; GCN-IR-NEXT: s_lshr_b32 s2, s11, 31 ; GCN-IR-NEXT: s_lshl_b64 s[10:11], s[10:11], 1 ; GCN-IR-NEXT: s_or_b64 s[14:15], s[14:15], s[2:3] -; GCN-IR-NEXT: s_or_b64 s[10:11], s[16:17], s[10:11] -; GCN-IR-NEXT: s_sub_u32 s2, s18, s14 -; GCN-IR-NEXT: s_subb_u32 s2, s19, s15 -; GCN-IR-NEXT: s_ashr_i32 s16, s2, 31 -; GCN-IR-NEXT: s_mov_b32 s17, s16 -; GCN-IR-NEXT: s_and_b32 s2, s16, 1 -; GCN-IR-NEXT: s_and_b64 s[16:17], s[16:17], s[8:9] -; GCN-IR-NEXT: s_sub_u32 s14, s14, s16 -; GCN-IR-NEXT: s_subb_u32 s15, s15, s17 -; GCN-IR-NEXT: s_add_u32 s12, s12, 1 -; GCN-IR-NEXT: s_addc_u32 s13, s13, 0 -; GCN-IR-NEXT: v_cmp_eq_u64_e64 s[20:21], s[12:13], 0 -; GCN-IR-NEXT: s_mov_b64 s[16:17], s[2:3] +; GCN-IR-NEXT: s_or_b64 s[10:11], s[12:13], s[10:11] +; GCN-IR-NEXT: s_sub_u32 s2, s16, s14 +; GCN-IR-NEXT: s_subb_u32 s2, s17, s15 +; GCN-IR-NEXT: s_ashr_i32 s12, s2, 31 +; GCN-IR-NEXT: s_mov_b32 s13, s12 +; GCN-IR-NEXT: s_and_b32 s2, s12, 1 +; GCN-IR-NEXT: s_and_b64 s[20:21], s[12:13], s[8:9] +; GCN-IR-NEXT: s_sub_u32 s14, s14, s20 +; GCN-IR-NEXT: s_subb_u32 s15, s15, s21 +; GCN-IR-NEXT: s_add_u32 s18, s18, 1 +; GCN-IR-NEXT: s_cselect_b64 s[20:21], -1, 0 +; GCN-IR-NEXT: s_or_b32 s20, s20, s21 +; GCN-IR-NEXT: s_cmp_lg_u32 s20, 0 +; GCN-IR-NEXT: s_addc_u32 s19, s19, 0 +; GCN-IR-NEXT: s_cselect_b64 s[20:21], -1, 0 +; GCN-IR-NEXT: s_mov_b64 s[12:13], s[2:3] ; GCN-IR-NEXT: s_and_b64 vcc, exec, s[20:21] ; GCN-IR-NEXT: s_cbranch_vccz .LBB8_3 ; GCN-IR-NEXT: .LBB8_4: ; %Flow7 @@ -1463,34 +1474,37 @@ define amdgpu_kernel void @s_test_srem_k_num_i64(ptr addrspace(1) %out, i64 %x) ; GCN-IR-NEXT: s_xor_b64 s[2:3], s[2:3], s[8:9] ; GCN-IR-NEXT: s_sub_u32 s4, s2, s8 ; GCN-IR-NEXT: s_subb_u32 s5, s3, s8 -; GCN-IR-NEXT: s_flbit_i32_b64 s12, s[4:5] -; GCN-IR-NEXT: s_add_u32 s2, s12, 0xffffffc5 +; GCN-IR-NEXT: s_flbit_i32_b64 s14, s[4:5] +; GCN-IR-NEXT: s_add_u32 s2, s14, 0xffffffc5 ; GCN-IR-NEXT: s_addc_u32 s3, 0, -1 ; GCN-IR-NEXT: v_cmp_eq_u64_e64 s[8:9], s[4:5], 0 ; GCN-IR-NEXT: v_cmp_gt_u64_e64 s[10:11], s[2:3], 63 -; GCN-IR-NEXT: v_cmp_eq_u64_e64 s[14:15], s[2:3], 63 +; GCN-IR-NEXT: v_cmp_eq_u64_e64 s[12:13], s[2:3], 63 ; GCN-IR-NEXT: s_or_b64 s[10:11], s[8:9], s[10:11] ; GCN-IR-NEXT: s_and_b64 s[8:9], s[10:11], exec ; GCN-IR-NEXT: s_cselect_b32 s8, 0, 24 -; GCN-IR-NEXT: s_or_b64 s[10:11], s[10:11], s[14:15] +; GCN-IR-NEXT: s_or_b64 s[10:11], s[10:11], s[12:13] ; GCN-IR-NEXT: s_andn2_b64 vcc, exec, s[10:11] ; GCN-IR-NEXT: s_mov_b32 s9, 0 ; GCN-IR-NEXT: s_cbranch_vccz .LBB10_5 ; GCN-IR-NEXT: ; %bb.1: ; %udiv-bb1 ; GCN-IR-NEXT: s_add_u32 s8, s2, 1 -; GCN-IR-NEXT: s_addc_u32 s9, s3, 0 -; GCN-IR-NEXT: v_cmp_eq_u64_e64 s[10:11], s[8:9], 0 +; GCN-IR-NEXT: s_cselect_b64 s[10:11], -1, 0 +; GCN-IR-NEXT: s_or_b32 s9, s10, s11 +; GCN-IR-NEXT: s_cmp_lg_u32 s9, 0 +; GCN-IR-NEXT: s_addc_u32 s3, s3, 0 +; GCN-IR-NEXT: s_cselect_b64 s[10:11], -1, 0 ; GCN-IR-NEXT: s_sub_i32 s2, 63, s2 ; GCN-IR-NEXT: s_andn2_b64 vcc, exec, s[10:11] ; GCN-IR-NEXT: s_lshl_b64 s[2:3], 24, s2 ; GCN-IR-NEXT: s_cbranch_vccz .LBB10_4 ; GCN-IR-NEXT: ; %bb.2: ; %udiv-preheader ; GCN-IR-NEXT: s_lshr_b64 s[10:11], 24, s8 -; GCN-IR-NEXT: s_add_u32 s14, s4, -1 -; GCN-IR-NEXT: s_addc_u32 s15, s5, -1 -; GCN-IR-NEXT: s_sub_u32 s8, 58, s12 -; GCN-IR-NEXT: s_subb_u32 s9, 0, 0 -; GCN-IR-NEXT: s_mov_b64 s[12:13], 0 +; GCN-IR-NEXT: s_add_u32 s12, s4, -1 +; GCN-IR-NEXT: s_addc_u32 s13, s5, -1 +; GCN-IR-NEXT: s_sub_u32 s14, 58, s14 +; GCN-IR-NEXT: s_subb_u32 s15, 0, 0 +; GCN-IR-NEXT: s_mov_b64 s[8:9], 0 ; GCN-IR-NEXT: s_mov_b32 s7, 0 ; GCN-IR-NEXT: .LBB10_3: ; %udiv-do-while ; GCN-IR-NEXT: ; =>This Inner Loop Header: Depth=1 @@ -1498,19 +1512,22 @@ define amdgpu_kernel void @s_test_srem_k_num_i64(ptr addrspace(1) %out, i64 %x) ; GCN-IR-NEXT: s_lshr_b32 s6, s3, 31 ; GCN-IR-NEXT: s_lshl_b64 s[2:3], s[2:3], 1 ; GCN-IR-NEXT: s_or_b64 s[10:11], s[10:11], s[6:7] -; GCN-IR-NEXT: s_or_b64 s[2:3], s[12:13], s[2:3] -; GCN-IR-NEXT: s_sub_u32 s6, s14, s10 -; GCN-IR-NEXT: s_subb_u32 s6, s15, s11 -; GCN-IR-NEXT: s_ashr_i32 s12, s6, 31 -; GCN-IR-NEXT: s_mov_b32 s13, s12 -; GCN-IR-NEXT: s_and_b32 s6, s12, 1 -; GCN-IR-NEXT: s_and_b64 s[12:13], s[12:13], s[4:5] -; GCN-IR-NEXT: s_sub_u32 s10, s10, s12 -; GCN-IR-NEXT: s_subb_u32 s11, s11, s13 -; GCN-IR-NEXT: s_add_u32 s8, s8, 1 -; GCN-IR-NEXT: s_addc_u32 s9, s9, 0 -; GCN-IR-NEXT: v_cmp_eq_u64_e64 s[16:17], s[8:9], 0 -; GCN-IR-NEXT: s_mov_b64 s[12:13], s[6:7] +; GCN-IR-NEXT: s_or_b64 s[2:3], s[8:9], s[2:3] +; GCN-IR-NEXT: s_sub_u32 s6, s12, s10 +; GCN-IR-NEXT: s_subb_u32 s6, s13, s11 +; GCN-IR-NEXT: s_ashr_i32 s8, s6, 31 +; GCN-IR-NEXT: s_mov_b32 s9, s8 +; GCN-IR-NEXT: s_and_b32 s6, s8, 1 +; GCN-IR-NEXT: s_and_b64 s[16:17], s[8:9], s[4:5] +; GCN-IR-NEXT: s_sub_u32 s10, s10, s16 +; GCN-IR-NEXT: s_subb_u32 s11, s11, s17 +; GCN-IR-NEXT: s_add_u32 s14, s14, 1 +; GCN-IR-NEXT: s_cselect_b64 s[16:17], -1, 0 +; GCN-IR-NEXT: s_or_b32 s16, s16, s17 +; GCN-IR-NEXT: s_cmp_lg_u32 s16, 0 +; GCN-IR-NEXT: s_addc_u32 s15, s15, 0 +; GCN-IR-NEXT: s_cselect_b64 s[16:17], -1, 0 +; GCN-IR-NEXT: s_mov_b64 s[8:9], s[6:7] ; GCN-IR-NEXT: s_and_b64 vcc, exec, s[16:17] ; GCN-IR-NEXT: s_cbranch_vccz .LBB10_3 ; GCN-IR-NEXT: .LBB10_4: ; %Flow6 @@ -1649,9 +1666,9 @@ define i64 @v_test_srem_k_num_i64(i64 %x) { ; GCN-IR-NEXT: v_ffbh_u32_e32 v2, v0 ; GCN-IR-NEXT: v_add_i32_e32 v2, vcc, 32, v2 ; GCN-IR-NEXT: v_ffbh_u32_e32 v3, v1 -; GCN-IR-NEXT: v_min_u32_e32 v10, v2, v3 +; GCN-IR-NEXT: v_min_u32_e32 v8, v2, v3 ; GCN-IR-NEXT: s_movk_i32 s6, 0xffc5 -; GCN-IR-NEXT: v_add_i32_e32 v2, vcc, s6, v10 +; GCN-IR-NEXT: v_add_i32_e32 v2, vcc, s6, v8 ; GCN-IR-NEXT: v_addc_u32_e64 v3, s[6:7], 0, -1, vcc ; GCN-IR-NEXT: v_cmp_eq_u64_e64 s[4:5], 0, v[0:1] ; GCN-IR-NEXT: v_cmp_lt_u64_e32 vcc, 63, v[2:3] @@ -1665,53 +1682,52 @@ define i64 @v_test_srem_k_num_i64(i64 %x) { ; GCN-IR-NEXT: s_cbranch_execz .LBB11_6 ; GCN-IR-NEXT: ; %bb.1: ; %udiv-bb1 ; GCN-IR-NEXT: v_add_i32_e32 v6, vcc, 1, v2 -; GCN-IR-NEXT: v_addc_u32_e32 v7, vcc, 0, v3, vcc +; GCN-IR-NEXT: v_addc_u32_e32 v3, vcc, 0, v3, vcc ; GCN-IR-NEXT: v_sub_i32_e64 v2, s[4:5], 63, v2 -; GCN-IR-NEXT: v_mov_b32_e32 v4, 0 -; GCN-IR-NEXT: v_cmp_ne_u64_e32 vcc, 0, v[6:7] ; GCN-IR-NEXT: v_lshl_b64 v[2:3], 24, v2 +; GCN-IR-NEXT: v_mov_b32_e32 v4, 0 ; GCN-IR-NEXT: v_mov_b32_e32 v5, 0 -; GCN-IR-NEXT: s_and_saveexec_b64 s[4:5], vcc -; GCN-IR-NEXT: s_xor_b64 s[8:9], exec, s[4:5] +; GCN-IR-NEXT: s_xor_b64 s[4:5], vcc, -1 +; GCN-IR-NEXT: s_and_saveexec_b64 s[8:9], s[4:5] +; GCN-IR-NEXT: s_xor_b64 s[4:5], exec, s[8:9] ; GCN-IR-NEXT: s_cbranch_execz .LBB11_5 ; GCN-IR-NEXT: ; %bb.2: ; %udiv-preheader -; GCN-IR-NEXT: v_add_i32_e32 v12, vcc, -1, v0 -; GCN-IR-NEXT: v_addc_u32_e32 v13, vcc, -1, v1, vcc -; GCN-IR-NEXT: v_lshr_b64 v[8:9], 24, v6 -; GCN-IR-NEXT: v_sub_i32_e32 v6, vcc, 58, v10 -; GCN-IR-NEXT: v_mov_b32_e32 v10, 0 -; GCN-IR-NEXT: v_subb_u32_e64 v7, s[4:5], 0, 0, vcc -; GCN-IR-NEXT: s_mov_b64 s[10:11], 0 -; GCN-IR-NEXT: v_mov_b32_e32 v11, 0 +; GCN-IR-NEXT: v_add_i32_e32 v10, vcc, -1, v0 +; GCN-IR-NEXT: v_addc_u32_e32 v11, vcc, -1, v1, vcc +; GCN-IR-NEXT: v_sub_i32_e32 v12, vcc, 58, v8 +; GCN-IR-NEXT: v_lshr_b64 v[6:7], 24, v6 +; GCN-IR-NEXT: v_subb_u32_e64 v13, s[8:9], 0, 0, vcc +; GCN-IR-NEXT: v_mov_b32_e32 v8, 0 +; GCN-IR-NEXT: s_mov_b64 s[8:9], 0 +; GCN-IR-NEXT: v_mov_b32_e32 v9, 0 ; GCN-IR-NEXT: v_mov_b32_e32 v5, 0 ; GCN-IR-NEXT: .LBB11_3: ; %udiv-do-while ; GCN-IR-NEXT: ; =>This Inner Loop Header: Depth=1 -; GCN-IR-NEXT: v_lshl_b64 v[8:9], v[8:9], 1 +; GCN-IR-NEXT: v_lshl_b64 v[6:7], v[6:7], 1 ; GCN-IR-NEXT: v_lshrrev_b32_e32 v4, 31, v3 -; GCN-IR-NEXT: v_or_b32_e32 v8, v8, v4 +; GCN-IR-NEXT: v_or_b32_e32 v6, v6, v4 ; GCN-IR-NEXT: v_lshl_b64 v[2:3], v[2:3], 1 -; GCN-IR-NEXT: v_sub_i32_e32 v4, vcc, v12, v8 -; GCN-IR-NEXT: v_subb_u32_e32 v4, vcc, v13, v9, vcc -; GCN-IR-NEXT: v_or_b32_e32 v2, v10, v2 -; GCN-IR-NEXT: v_ashrrev_i32_e32 v10, 31, v4 -; GCN-IR-NEXT: v_add_i32_e32 v6, vcc, 1, v6 -; GCN-IR-NEXT: v_or_b32_e32 v3, v11, v3 -; GCN-IR-NEXT: v_and_b32_e32 v4, 1, v10 -; GCN-IR-NEXT: v_and_b32_e32 v11, v10, v1 -; GCN-IR-NEXT: v_and_b32_e32 v10, v10, v0 -; GCN-IR-NEXT: v_addc_u32_e32 v7, vcc, 0, v7, vcc -; GCN-IR-NEXT: v_cmp_eq_u64_e32 vcc, 0, v[6:7] -; GCN-IR-NEXT: v_sub_i32_e64 v8, s[4:5], v8, v10 -; GCN-IR-NEXT: v_subb_u32_e64 v9, s[4:5], v9, v11, s[4:5] -; GCN-IR-NEXT: v_mov_b32_e32 v11, v5 -; GCN-IR-NEXT: s_or_b64 s[10:11], vcc, s[10:11] -; GCN-IR-NEXT: v_mov_b32_e32 v10, v4 -; GCN-IR-NEXT: s_andn2_b64 exec, exec, s[10:11] +; GCN-IR-NEXT: v_sub_i32_e32 v4, vcc, v10, v6 +; GCN-IR-NEXT: v_subb_u32_e32 v4, vcc, v11, v7, vcc +; GCN-IR-NEXT: v_or_b32_e32 v2, v8, v2 +; GCN-IR-NEXT: v_ashrrev_i32_e32 v8, 31, v4 +; GCN-IR-NEXT: v_or_b32_e32 v3, v9, v3 +; GCN-IR-NEXT: v_and_b32_e32 v4, 1, v8 +; GCN-IR-NEXT: v_and_b32_e32 v9, v8, v1 +; GCN-IR-NEXT: v_and_b32_e32 v8, v8, v0 +; GCN-IR-NEXT: v_sub_i32_e32 v6, vcc, v6, v8 +; GCN-IR-NEXT: v_subb_u32_e32 v7, vcc, v7, v9, vcc +; GCN-IR-NEXT: v_add_i32_e32 v12, vcc, 1, v12 +; GCN-IR-NEXT: v_addc_u32_e32 v13, vcc, 0, v13, vcc +; GCN-IR-NEXT: v_mov_b32_e32 v9, v5 +; GCN-IR-NEXT: s_or_b64 s[8:9], vcc, s[8:9] +; GCN-IR-NEXT: v_mov_b32_e32 v8, v4 +; GCN-IR-NEXT: s_andn2_b64 exec, exec, s[8:9] ; GCN-IR-NEXT: s_cbranch_execnz .LBB11_3 ; GCN-IR-NEXT: ; %bb.4: ; %Flow -; GCN-IR-NEXT: s_or_b64 exec, exec, s[10:11] -; GCN-IR-NEXT: .LBB11_5: ; %Flow4 ; GCN-IR-NEXT: s_or_b64 exec, exec, s[8:9] +; GCN-IR-NEXT: .LBB11_5: ; %Flow4 +; GCN-IR-NEXT: s_or_b64 exec, exec, s[4:5] ; GCN-IR-NEXT: v_lshl_b64 v[2:3], v[2:3], 1 ; GCN-IR-NEXT: v_or_b32_e32 v5, v5, v3 ; GCN-IR-NEXT: v_or_b32_e32 v4, v4, v2 @@ -1840,9 +1856,9 @@ define i64 @v_test_srem_pow2_k_num_i64(i64 %x) { ; GCN-IR-NEXT: v_ffbh_u32_e32 v2, v0 ; GCN-IR-NEXT: v_add_i32_e32 v2, vcc, 32, v2 ; GCN-IR-NEXT: v_ffbh_u32_e32 v3, v1 -; GCN-IR-NEXT: v_min_u32_e32 v10, v2, v3 +; GCN-IR-NEXT: v_min_u32_e32 v8, v2, v3 ; GCN-IR-NEXT: s_movk_i32 s6, 0xffd0 -; GCN-IR-NEXT: v_add_i32_e32 v2, vcc, s6, v10 +; GCN-IR-NEXT: v_add_i32_e32 v2, vcc, s6, v8 ; GCN-IR-NEXT: v_addc_u32_e64 v3, s[6:7], 0, -1, vcc ; GCN-IR-NEXT: v_cmp_eq_u64_e64 s[4:5], 0, v[0:1] ; GCN-IR-NEXT: v_cmp_lt_u64_e32 vcc, 63, v[2:3] @@ -1857,54 +1873,53 @@ define i64 @v_test_srem_pow2_k_num_i64(i64 %x) { ; GCN-IR-NEXT: s_cbranch_execz .LBB12_6 ; GCN-IR-NEXT: ; %bb.1: ; %udiv-bb1 ; GCN-IR-NEXT: v_add_i32_e32 v6, vcc, 1, v2 +; GCN-IR-NEXT: v_addc_u32_e32 v3, vcc, 0, v3, vcc ; GCN-IR-NEXT: v_sub_i32_e64 v2, s[4:5], 63, v2 -; GCN-IR-NEXT: v_addc_u32_e32 v7, vcc, 0, v3, vcc -; GCN-IR-NEXT: s_mov_b64 s[4:5], 0x8000 +; GCN-IR-NEXT: s_mov_b64 s[8:9], 0x8000 +; GCN-IR-NEXT: v_lshl_b64 v[2:3], s[8:9], v2 ; GCN-IR-NEXT: v_mov_b32_e32 v4, 0 -; GCN-IR-NEXT: v_cmp_ne_u64_e32 vcc, 0, v[6:7] -; GCN-IR-NEXT: v_lshl_b64 v[2:3], s[4:5], v2 ; GCN-IR-NEXT: v_mov_b32_e32 v5, 0 -; GCN-IR-NEXT: s_and_saveexec_b64 s[8:9], vcc -; GCN-IR-NEXT: s_xor_b64 s[8:9], exec, s[8:9] +; GCN-IR-NEXT: s_xor_b64 s[4:5], vcc, -1 +; GCN-IR-NEXT: s_and_saveexec_b64 s[10:11], s[4:5] +; GCN-IR-NEXT: s_xor_b64 s[4:5], exec, s[10:11] ; GCN-IR-NEXT: s_cbranch_execz .LBB12_5 ; GCN-IR-NEXT: ; %bb.2: ; %udiv-preheader -; GCN-IR-NEXT: v_add_i32_e32 v12, vcc, -1, v0 -; GCN-IR-NEXT: v_addc_u32_e32 v13, vcc, -1, v1, vcc -; GCN-IR-NEXT: v_lshr_b64 v[8:9], s[4:5], v6 -; GCN-IR-NEXT: v_sub_i32_e32 v6, vcc, 47, v10 -; GCN-IR-NEXT: v_mov_b32_e32 v10, 0 -; GCN-IR-NEXT: v_subb_u32_e64 v7, s[4:5], 0, 0, vcc -; GCN-IR-NEXT: s_mov_b64 s[10:11], 0 -; GCN-IR-NEXT: v_mov_b32_e32 v11, 0 +; GCN-IR-NEXT: v_add_i32_e32 v10, vcc, -1, v0 +; GCN-IR-NEXT: v_addc_u32_e32 v11, vcc, -1, v1, vcc +; GCN-IR-NEXT: v_sub_i32_e32 v12, vcc, 47, v8 +; GCN-IR-NEXT: v_lshr_b64 v[6:7], s[8:9], v6 +; GCN-IR-NEXT: v_subb_u32_e64 v13, s[8:9], 0, 0, vcc +; GCN-IR-NEXT: v_mov_b32_e32 v8, 0 +; GCN-IR-NEXT: s_mov_b64 s[8:9], 0 +; GCN-IR-NEXT: v_mov_b32_e32 v9, 0 ; GCN-IR-NEXT: v_mov_b32_e32 v5, 0 ; GCN-IR-NEXT: .LBB12_3: ; %udiv-do-while ; GCN-IR-NEXT: ; =>This Inner Loop Header: Depth=1 -; GCN-IR-NEXT: v_lshl_b64 v[8:9], v[8:9], 1 +; GCN-IR-NEXT: v_lshl_b64 v[6:7], v[6:7], 1 ; GCN-IR-NEXT: v_lshrrev_b32_e32 v4, 31, v3 -; GCN-IR-NEXT: v_or_b32_e32 v8, v8, v4 +; GCN-IR-NEXT: v_or_b32_e32 v6, v6, v4 ; GCN-IR-NEXT: v_lshl_b64 v[2:3], v[2:3], 1 -; GCN-IR-NEXT: v_sub_i32_e32 v4, vcc, v12, v8 -; GCN-IR-NEXT: v_subb_u32_e32 v4, vcc, v13, v9, vcc -; GCN-IR-NEXT: v_or_b32_e32 v2, v10, v2 -; GCN-IR-NEXT: v_ashrrev_i32_e32 v10, 31, v4 -; GCN-IR-NEXT: v_add_i32_e32 v6, vcc, 1, v6 -; GCN-IR-NEXT: v_or_b32_e32 v3, v11, v3 -; GCN-IR-NEXT: v_and_b32_e32 v4, 1, v10 -; GCN-IR-NEXT: v_and_b32_e32 v11, v10, v1 -; GCN-IR-NEXT: v_and_b32_e32 v10, v10, v0 -; GCN-IR-NEXT: v_addc_u32_e32 v7, vcc, 0, v7, vcc -; GCN-IR-NEXT: v_cmp_eq_u64_e32 vcc, 0, v[6:7] -; GCN-IR-NEXT: v_sub_i32_e64 v8, s[4:5], v8, v10 -; GCN-IR-NEXT: v_subb_u32_e64 v9, s[4:5], v9, v11, s[4:5] -; GCN-IR-NEXT: v_mov_b32_e32 v11, v5 -; GCN-IR-NEXT: s_or_b64 s[10:11], vcc, s[10:11] -; GCN-IR-NEXT: v_mov_b32_e32 v10, v4 -; GCN-IR-NEXT: s_andn2_b64 exec, exec, s[10:11] +; GCN-IR-NEXT: v_sub_i32_e32 v4, vcc, v10, v6 +; GCN-IR-NEXT: v_subb_u32_e32 v4, vcc, v11, v7, vcc +; GCN-IR-NEXT: v_or_b32_e32 v2, v8, v2 +; GCN-IR-NEXT: v_ashrrev_i32_e32 v8, 31, v4 +; GCN-IR-NEXT: v_or_b32_e32 v3, v9, v3 +; GCN-IR-NEXT: v_and_b32_e32 v4, 1, v8 +; GCN-IR-NEXT: v_and_b32_e32 v9, v8, v1 +; GCN-IR-NEXT: v_and_b32_e32 v8, v8, v0 +; GCN-IR-NEXT: v_sub_i32_e32 v6, vcc, v6, v8 +; GCN-IR-NEXT: v_subb_u32_e32 v7, vcc, v7, v9, vcc +; GCN-IR-NEXT: v_add_i32_e32 v12, vcc, 1, v12 +; GCN-IR-NEXT: v_addc_u32_e32 v13, vcc, 0, v13, vcc +; GCN-IR-NEXT: v_mov_b32_e32 v9, v5 +; GCN-IR-NEXT: s_or_b64 s[8:9], vcc, s[8:9] +; GCN-IR-NEXT: v_mov_b32_e32 v8, v4 +; GCN-IR-NEXT: s_andn2_b64 exec, exec, s[8:9] ; GCN-IR-NEXT: s_cbranch_execnz .LBB12_3 ; GCN-IR-NEXT: ; %bb.4: ; %Flow -; GCN-IR-NEXT: s_or_b64 exec, exec, s[10:11] -; GCN-IR-NEXT: .LBB12_5: ; %Flow4 ; GCN-IR-NEXT: s_or_b64 exec, exec, s[8:9] +; GCN-IR-NEXT: .LBB12_5: ; %Flow4 +; GCN-IR-NEXT: s_or_b64 exec, exec, s[4:5] ; GCN-IR-NEXT: v_lshl_b64 v[2:3], v[2:3], 1 ; GCN-IR-NEXT: v_or_b32_e32 v5, v5, v3 ; GCN-IR-NEXT: v_or_b32_e32 v4, v4, v2 @@ -1939,20 +1954,20 @@ define i64 @v_test_srem_pow2_k_den_i64(i64 %x) { ; GCN-IR-LABEL: v_test_srem_pow2_k_den_i64: ; GCN-IR: ; %bb.0: ; %_udiv-special-cases ; GCN-IR-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) -; GCN-IR-NEXT: v_ashrrev_i32_e32 v12, 31, v1 -; GCN-IR-NEXT: v_xor_b32_e32 v0, v0, v12 -; GCN-IR-NEXT: v_xor_b32_e32 v1, v1, v12 -; GCN-IR-NEXT: v_sub_i32_e32 v0, vcc, v0, v12 -; GCN-IR-NEXT: v_subb_u32_e32 v1, vcc, v1, v12, vcc +; GCN-IR-NEXT: v_ashrrev_i32_e32 v10, 31, v1 +; GCN-IR-NEXT: v_xor_b32_e32 v0, v0, v10 +; GCN-IR-NEXT: v_xor_b32_e32 v1, v1, v10 +; GCN-IR-NEXT: v_sub_i32_e32 v0, vcc, v0, v10 +; GCN-IR-NEXT: v_subb_u32_e32 v1, vcc, v1, v10, vcc ; GCN-IR-NEXT: v_ffbh_u32_e32 v2, v0 ; GCN-IR-NEXT: v_add_i32_e64 v2, s[4:5], 32, v2 ; GCN-IR-NEXT: v_ffbh_u32_e32 v3, v1 -; GCN-IR-NEXT: v_min_u32_e32 v10, v2, v3 -; GCN-IR-NEXT: v_sub_i32_e64 v2, s[4:5], 48, v10 +; GCN-IR-NEXT: v_min_u32_e32 v8, v2, v3 +; GCN-IR-NEXT: v_sub_i32_e64 v2, s[4:5], 48, v8 ; GCN-IR-NEXT: v_subb_u32_e64 v3, s[4:5], 0, 0, s[4:5] ; GCN-IR-NEXT: v_cmp_eq_u64_e32 vcc, 0, v[0:1] ; GCN-IR-NEXT: v_cmp_lt_u64_e64 s[4:5], 63, v[2:3] -; GCN-IR-NEXT: v_mov_b32_e32 v13, v12 +; GCN-IR-NEXT: v_mov_b32_e32 v11, v10 ; GCN-IR-NEXT: s_or_b64 s[4:5], vcc, s[4:5] ; GCN-IR-NEXT: v_cmp_ne_u64_e32 vcc, 63, v[2:3] ; GCN-IR-NEXT: s_xor_b64 s[6:7], s[4:5], -1 @@ -1963,51 +1978,50 @@ define i64 @v_test_srem_pow2_k_den_i64(i64 %x) { ; GCN-IR-NEXT: s_cbranch_execz .LBB13_6 ; GCN-IR-NEXT: ; %bb.1: ; %udiv-bb1 ; GCN-IR-NEXT: v_add_i32_e32 v6, vcc, 1, v2 -; GCN-IR-NEXT: v_addc_u32_e32 v7, vcc, 0, v3, vcc +; GCN-IR-NEXT: v_addc_u32_e32 v3, vcc, 0, v3, vcc ; GCN-IR-NEXT: v_sub_i32_e64 v2, s[4:5], 63, v2 -; GCN-IR-NEXT: v_mov_b32_e32 v4, 0 -; GCN-IR-NEXT: v_cmp_ne_u64_e32 vcc, 0, v[6:7] ; GCN-IR-NEXT: v_lshl_b64 v[2:3], v[0:1], v2 +; GCN-IR-NEXT: v_mov_b32_e32 v4, 0 ; GCN-IR-NEXT: v_mov_b32_e32 v5, 0 -; GCN-IR-NEXT: s_and_saveexec_b64 s[4:5], vcc -; GCN-IR-NEXT: s_xor_b64 s[8:9], exec, s[4:5] +; GCN-IR-NEXT: s_xor_b64 s[4:5], vcc, -1 +; GCN-IR-NEXT: s_and_saveexec_b64 s[8:9], s[4:5] +; GCN-IR-NEXT: s_xor_b64 s[4:5], exec, s[8:9] ; GCN-IR-NEXT: s_cbranch_execz .LBB13_5 ; GCN-IR-NEXT: ; %bb.2: ; %udiv-preheader -; GCN-IR-NEXT: v_lshr_b64 v[8:9], v[0:1], v6 -; GCN-IR-NEXT: v_add_i32_e32 v6, vcc, 0xffffffcf, v10 -; GCN-IR-NEXT: v_mov_b32_e32 v10, 0 -; GCN-IR-NEXT: v_addc_u32_e64 v7, s[4:5], 0, -1, vcc -; GCN-IR-NEXT: s_mov_b64 s[10:11], 0 -; GCN-IR-NEXT: v_mov_b32_e32 v11, 0 +; GCN-IR-NEXT: v_add_i32_e32 v12, vcc, 0xffffffcf, v8 +; GCN-IR-NEXT: v_lshr_b64 v[6:7], v[0:1], v6 +; GCN-IR-NEXT: v_addc_u32_e64 v13, s[8:9], 0, -1, vcc +; GCN-IR-NEXT: v_mov_b32_e32 v8, 0 +; GCN-IR-NEXT: s_mov_b64 s[8:9], 0 +; GCN-IR-NEXT: v_mov_b32_e32 v9, 0 ; GCN-IR-NEXT: v_mov_b32_e32 v5, 0 -; GCN-IR-NEXT: s_movk_i32 s12, 0x7fff +; GCN-IR-NEXT: s_movk_i32 s10, 0x7fff ; GCN-IR-NEXT: .LBB13_3: ; %udiv-do-while ; GCN-IR-NEXT: ; =>This Inner Loop Header: Depth=1 -; GCN-IR-NEXT: v_lshl_b64 v[8:9], v[8:9], 1 +; GCN-IR-NEXT: v_lshl_b64 v[6:7], v[6:7], 1 ; GCN-IR-NEXT: v_lshrrev_b32_e32 v4, 31, v3 -; GCN-IR-NEXT: v_or_b32_e32 v8, v8, v4 -; GCN-IR-NEXT: v_sub_i32_e32 v4, vcc, s12, v8 +; GCN-IR-NEXT: v_or_b32_e32 v6, v6, v4 ; GCN-IR-NEXT: v_lshl_b64 v[2:3], v[2:3], 1 -; GCN-IR-NEXT: v_subb_u32_e32 v4, vcc, 0, v9, vcc -; GCN-IR-NEXT: v_add_i32_e32 v6, vcc, 1, v6 -; GCN-IR-NEXT: v_or_b32_e32 v2, v10, v2 -; GCN-IR-NEXT: v_ashrrev_i32_e32 v10, 31, v4 -; GCN-IR-NEXT: v_addc_u32_e32 v7, vcc, 0, v7, vcc -; GCN-IR-NEXT: v_and_b32_e32 v4, 1, v10 -; GCN-IR-NEXT: v_and_b32_e32 v10, 0x8000, v10 -; GCN-IR-NEXT: v_cmp_eq_u64_e32 vcc, 0, v[6:7] -; GCN-IR-NEXT: v_or_b32_e32 v3, v11, v3 -; GCN-IR-NEXT: v_sub_i32_e64 v8, s[4:5], v8, v10 -; GCN-IR-NEXT: v_mov_b32_e32 v11, v5 -; GCN-IR-NEXT: v_subbrev_u32_e64 v9, s[4:5], 0, v9, s[4:5] -; GCN-IR-NEXT: s_or_b64 s[10:11], vcc, s[10:11] -; GCN-IR-NEXT: v_mov_b32_e32 v10, v4 -; GCN-IR-NEXT: s_andn2_b64 exec, exec, s[10:11] +; GCN-IR-NEXT: v_sub_i32_e32 v4, vcc, s10, v6 +; GCN-IR-NEXT: v_subb_u32_e32 v4, vcc, 0, v7, vcc +; GCN-IR-NEXT: v_or_b32_e32 v2, v8, v2 +; GCN-IR-NEXT: v_ashrrev_i32_e32 v8, 31, v4 +; GCN-IR-NEXT: v_and_b32_e32 v4, 1, v8 +; GCN-IR-NEXT: v_and_b32_e32 v8, 0x8000, v8 +; GCN-IR-NEXT: v_sub_i32_e32 v6, vcc, v6, v8 +; GCN-IR-NEXT: v_subbrev_u32_e32 v7, vcc, 0, v7, vcc +; GCN-IR-NEXT: v_add_i32_e32 v12, vcc, 1, v12 +; GCN-IR-NEXT: v_or_b32_e32 v3, v9, v3 +; GCN-IR-NEXT: v_addc_u32_e32 v13, vcc, 0, v13, vcc +; GCN-IR-NEXT: v_mov_b32_e32 v9, v5 +; GCN-IR-NEXT: s_or_b64 s[8:9], vcc, s[8:9] +; GCN-IR-NEXT: v_mov_b32_e32 v8, v4 +; GCN-IR-NEXT: s_andn2_b64 exec, exec, s[8:9] ; GCN-IR-NEXT: s_cbranch_execnz .LBB13_3 ; GCN-IR-NEXT: ; %bb.4: ; %Flow -; GCN-IR-NEXT: s_or_b64 exec, exec, s[10:11] -; GCN-IR-NEXT: .LBB13_5: ; %Flow4 ; GCN-IR-NEXT: s_or_b64 exec, exec, s[8:9] +; GCN-IR-NEXT: .LBB13_5: ; %Flow4 +; GCN-IR-NEXT: s_or_b64 exec, exec, s[4:5] ; GCN-IR-NEXT: v_lshl_b64 v[2:3], v[2:3], 1 ; GCN-IR-NEXT: v_or_b32_e32 v5, v5, v3 ; GCN-IR-NEXT: v_or_b32_e32 v4, v4, v2 @@ -2016,10 +2030,10 @@ define i64 @v_test_srem_pow2_k_den_i64(i64 %x) { ; GCN-IR-NEXT: v_lshl_b64 v[2:3], v[4:5], 15 ; GCN-IR-NEXT: v_sub_i32_e32 v0, vcc, v0, v2 ; GCN-IR-NEXT: v_subb_u32_e32 v1, vcc, v1, v3, vcc -; GCN-IR-NEXT: v_xor_b32_e32 v0, v0, v12 -; GCN-IR-NEXT: v_xor_b32_e32 v1, v1, v13 -; GCN-IR-NEXT: v_sub_i32_e32 v0, vcc, v0, v12 -; GCN-IR-NEXT: v_subb_u32_e32 v1, vcc, v1, v13, vcc +; GCN-IR-NEXT: v_xor_b32_e32 v0, v0, v10 +; GCN-IR-NEXT: v_xor_b32_e32 v1, v1, v11 +; GCN-IR-NEXT: v_sub_i32_e32 v0, vcc, v0, v10 +; GCN-IR-NEXT: v_subb_u32_e32 v1, vcc, v1, v11, vcc ; GCN-IR-NEXT: s_setpc_b64 s[30:31] %result = srem i64 %x, 32768 ret i64 %result diff --git a/llvm/test/CodeGen/AMDGPU/uaddo.ll b/llvm/test/CodeGen/AMDGPU/uaddo.ll index e1574dcd45462..bb5918b256d28 100644 --- a/llvm/test/CodeGen/AMDGPU/uaddo.ll +++ b/llvm/test/CodeGen/AMDGPU/uaddo.ll @@ -14,15 +14,16 @@ define amdgpu_kernel void @s_uaddo_i64_zext(ptr addrspace(1) %out, i64 %a, i64 % ; SI-NEXT: s_mov_b32 s6, -1 ; SI-NEXT: s_waitcnt lgkmcnt(0) ; SI-NEXT: s_mov_b32 s4, s0 -; SI-NEXT: s_add_u32 s0, s2, s8 -; SI-NEXT: v_mov_b32_e32 v0, s2 +; SI-NEXT: s_add_u32 s2, s2, s8 ; SI-NEXT: s_mov_b32 s5, s1 -; SI-NEXT: s_addc_u32 s1, s3, s9 +; SI-NEXT: s_cselect_b64 s[0:1], -1, 0 +; SI-NEXT: s_or_b32 s0, s0, s1 +; SI-NEXT: s_cmp_lg_u32 s0, 0 +; SI-NEXT: s_addc_u32 s3, s3, s9 +; SI-NEXT: s_cselect_b64 s[0:1], -1, 0 +; SI-NEXT: v_cndmask_b32_e64 v0, 0, 1, s[0:1] ; SI-NEXT: v_mov_b32_e32 v1, s3 -; SI-NEXT: v_cmp_lt_u64_e32 vcc, s[0:1], v[0:1] -; SI-NEXT: v_mov_b32_e32 v1, s1 -; SI-NEXT: v_cndmask_b32_e64 v0, 0, 1, vcc -; SI-NEXT: v_add_i32_e32 v0, vcc, s0, v0 +; SI-NEXT: v_add_i32_e32 v0, vcc, s2, v0 ; SI-NEXT: v_addc_u32_e32 v1, vcc, 0, v1, vcc ; SI-NEXT: buffer_store_dwordx2 v[0:1], off, s[4:7], 0 ; SI-NEXT: s_endpgm @@ -33,15 +34,15 @@ define amdgpu_kernel void @s_uaddo_i64_zext(ptr addrspace(1) %out, i64 %a, i64 % ; VI-NEXT: s_load_dwordx2 s[4:5], s[4:5], 0x34 ; VI-NEXT: s_waitcnt lgkmcnt(0) ; VI-NEXT: v_mov_b32_e32 v0, s0 -; VI-NEXT: s_add_u32 s0, s2, s4 -; VI-NEXT: v_mov_b32_e32 v2, s2 +; VI-NEXT: s_add_u32 s2, s2, s4 ; VI-NEXT: v_mov_b32_e32 v1, s1 +; VI-NEXT: s_cselect_b64 s[0:1], -1, 0 +; VI-NEXT: s_cmp_lg_u64 s[0:1], 0 +; VI-NEXT: s_addc_u32 s3, s3, s5 +; VI-NEXT: s_cselect_b64 s[0:1], -1, 0 +; VI-NEXT: v_cndmask_b32_e64 v2, 0, 1, s[0:1] ; VI-NEXT: v_mov_b32_e32 v3, s3 -; VI-NEXT: s_addc_u32 s1, s3, s5 -; VI-NEXT: v_cmp_lt_u64_e32 vcc, s[0:1], v[2:3] -; VI-NEXT: v_mov_b32_e32 v3, s1 -; VI-NEXT: v_cndmask_b32_e64 v2, 0, 1, vcc -; VI-NEXT: v_add_u32_e32 v2, vcc, s0, v2 +; VI-NEXT: v_add_u32_e32 v2, vcc, s2, v2 ; VI-NEXT: v_addc_u32_e32 v3, vcc, 0, v3, vcc ; VI-NEXT: flat_store_dwordx2 v[0:1], v[2:3] ; VI-NEXT: s_endpgm @@ -52,14 +53,14 @@ define amdgpu_kernel void @s_uaddo_i64_zext(ptr addrspace(1) %out, i64 %a, i64 % ; GFX9-NEXT: s_load_dwordx2 s[6:7], s[4:5], 0x34 ; GFX9-NEXT: v_mov_b32_e32 v2, 0 ; GFX9-NEXT: s_waitcnt lgkmcnt(0) -; GFX9-NEXT: v_mov_b32_e32 v0, s2 -; GFX9-NEXT: s_add_u32 s4, s2, s6 -; GFX9-NEXT: v_mov_b32_e32 v1, s3 -; GFX9-NEXT: s_addc_u32 s5, s3, s7 -; GFX9-NEXT: v_cmp_lt_u64_e32 vcc, s[4:5], v[0:1] -; GFX9-NEXT: v_mov_b32_e32 v1, s5 -; GFX9-NEXT: v_cndmask_b32_e64 v0, 0, 1, vcc -; GFX9-NEXT: v_add_co_u32_e32 v0, vcc, s4, v0 +; GFX9-NEXT: s_add_u32 s6, s2, s6 +; GFX9-NEXT: s_cselect_b64 s[4:5], -1, 0 +; GFX9-NEXT: s_cmp_lg_u64 s[4:5], 0 +; GFX9-NEXT: s_addc_u32 s4, s3, s7 +; GFX9-NEXT: s_cselect_b64 s[2:3], -1, 0 +; GFX9-NEXT: v_cndmask_b32_e64 v0, 0, 1, s[2:3] +; GFX9-NEXT: v_mov_b32_e32 v1, s4 +; GFX9-NEXT: v_add_co_u32_e32 v0, vcc, s6, v0 ; GFX9-NEXT: v_addc_co_u32_e32 v1, vcc, 0, v1, vcc ; GFX9-NEXT: global_store_dwordx2 v2, v[0:1], s[0:1] ; GFX9-NEXT: s_endpgm @@ -71,12 +72,14 @@ define amdgpu_kernel void @s_uaddo_i64_zext(ptr addrspace(1) %out, i64 %a, i64 % ; GFX10-NEXT: s_load_dwordx2 s[6:7], s[4:5], 0x34 ; GFX10-NEXT: v_mov_b32_e32 v2, 0 ; GFX10-NEXT: s_waitcnt lgkmcnt(0) -; GFX10-NEXT: s_add_u32 s4, s2, s6 -; GFX10-NEXT: s_addc_u32 s5, s3, s7 -; GFX10-NEXT: v_cmp_lt_u64_e64 s2, s[4:5], s[2:3] -; GFX10-NEXT: v_cndmask_b32_e64 v0, 0, 1, s2 -; GFX10-NEXT: v_add_co_u32 v0, s2, s4, v0 -; GFX10-NEXT: v_add_co_ci_u32_e64 v1, s2, s5, 0, s2 +; GFX10-NEXT: s_add_u32 s2, s2, s6 +; GFX10-NEXT: s_cselect_b32 s4, -1, 0 +; GFX10-NEXT: s_cmp_lg_u32 s4, 0 +; GFX10-NEXT: s_addc_u32 s3, s3, s7 +; GFX10-NEXT: s_cselect_b32 s4, -1, 0 +; GFX10-NEXT: v_cndmask_b32_e64 v0, 0, 1, s4 +; GFX10-NEXT: v_add_co_u32 v0, s2, s2, v0 +; GFX10-NEXT: v_add_co_ci_u32_e64 v1, s2, s3, 0, s2 ; GFX10-NEXT: global_store_dwordx2 v2, v[0:1], s[0:1] ; GFX10-NEXT: s_endpgm ; @@ -87,14 +90,16 @@ define amdgpu_kernel void @s_uaddo_i64_zext(ptr addrspace(1) %out, i64 %a, i64 % ; GFX11-NEXT: s_load_b64 s[4:5], s[4:5], 0x34 ; GFX11-NEXT: v_mov_b32_e32 v2, 0 ; GFX11-NEXT: s_waitcnt lgkmcnt(0) -; GFX11-NEXT: s_add_u32 s4, s2, s4 -; GFX11-NEXT: s_addc_u32 s5, s3, s5 -; GFX11-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(VALU_DEP_1) -; GFX11-NEXT: v_cmp_lt_u64_e64 s2, s[4:5], s[2:3] -; GFX11-NEXT: v_cndmask_b32_e64 v0, 0, 1, s2 +; GFX11-NEXT: s_add_u32 s2, s2, s4 +; GFX11-NEXT: s_cselect_b32 s4, -1, 0 +; GFX11-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_2) | instid1(SALU_CYCLE_1) +; GFX11-NEXT: s_cmp_lg_u32 s4, 0 +; GFX11-NEXT: s_addc_u32 s3, s3, s5 +; GFX11-NEXT: s_cselect_b32 s4, -1, 0 +; GFX11-NEXT: v_cndmask_b32_e64 v0, 0, 1, s4 ; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1) -; GFX11-NEXT: v_add_co_u32 v0, s2, s4, v0 -; GFX11-NEXT: v_add_co_ci_u32_e64 v1, null, s5, 0, s2 +; GFX11-NEXT: v_add_co_u32 v0, s2, s2, v0 +; GFX11-NEXT: v_add_co_ci_u32_e64 v1, null, s3, 0, s2 ; GFX11-NEXT: global_store_b64 v2, v[0:1], s[0:1] ; GFX11-NEXT: s_endpgm %uadd = call { i64, i1 } @llvm.uadd.with.overflow.i64(i64 %a, i64 %b) @@ -436,21 +441,23 @@ define amdgpu_kernel void @s_uaddo_i64(ptr addrspace(1) %out, ptr addrspace(1) % ; SI-NEXT: s_mov_b32 s11, 0xf000 ; SI-NEXT: s_mov_b32 s10, -1 ; SI-NEXT: s_waitcnt lgkmcnt(0) -; SI-NEXT: s_add_u32 s6, s4, s6 -; SI-NEXT: v_mov_b32_e32 v0, s4 -; SI-NEXT: s_addc_u32 s7, s5, s7 -; SI-NEXT: v_mov_b32_e32 v1, s5 -; SI-NEXT: v_cmp_lt_u64_e32 vcc, s[6:7], v[0:1] -; SI-NEXT: v_mov_b32_e32 v2, s6 +; SI-NEXT: s_add_u32 s4, s4, s6 +; SI-NEXT: s_cselect_b64 s[12:13], -1, 0 +; SI-NEXT: s_or_b32 s6, s12, s13 +; SI-NEXT: s_cmp_lg_u32 s6, 0 +; SI-NEXT: s_addc_u32 s5, s5, s7 ; SI-NEXT: s_mov_b32 s8, s0 ; SI-NEXT: s_mov_b32 s9, s1 +; SI-NEXT: v_mov_b32_e32 v0, s4 +; SI-NEXT: v_mov_b32_e32 v1, s5 +; SI-NEXT: s_cselect_b64 s[4:5], -1, 0 ; SI-NEXT: s_mov_b32 s0, s2 ; SI-NEXT: s_mov_b32 s1, s3 ; SI-NEXT: s_mov_b32 s2, s10 ; SI-NEXT: s_mov_b32 s3, s11 -; SI-NEXT: v_mov_b32_e32 v3, s7 -; SI-NEXT: v_cndmask_b32_e64 v0, 0, 1, vcc -; SI-NEXT: buffer_store_dwordx2 v[2:3], off, s[8:11], 0 +; SI-NEXT: buffer_store_dwordx2 v[0:1], off, s[8:11], 0 +; SI-NEXT: s_waitcnt expcnt(0) +; SI-NEXT: v_cndmask_b32_e64 v0, 0, 1, s[4:5] ; SI-NEXT: buffer_store_byte v0, off, s[0:3], 0 ; SI-NEXT: s_endpgm ; @@ -458,37 +465,37 @@ define amdgpu_kernel void @s_uaddo_i64(ptr addrspace(1) %out, ptr addrspace(1) % ; VI: ; %bb.0: ; VI-NEXT: s_load_dwordx8 s[0:7], s[4:5], 0x24 ; VI-NEXT: s_waitcnt lgkmcnt(0) +; VI-NEXT: v_mov_b32_e32 v2, s2 +; VI-NEXT: s_add_u32 s2, s4, s6 ; VI-NEXT: v_mov_b32_e32 v0, s0 -; VI-NEXT: s_add_u32 s0, s4, s6 -; VI-NEXT: v_mov_b32_e32 v4, s4 ; VI-NEXT: v_mov_b32_e32 v1, s1 -; VI-NEXT: s_addc_u32 s1, s5, s7 -; VI-NEXT: v_mov_b32_e32 v5, s5 -; VI-NEXT: v_mov_b32_e32 v7, s1 -; VI-NEXT: v_cmp_lt_u64_e32 vcc, s[0:1], v[4:5] -; VI-NEXT: v_mov_b32_e32 v6, s0 -; VI-NEXT: v_mov_b32_e32 v2, s2 +; VI-NEXT: s_cselect_b64 s[0:1], -1, 0 +; VI-NEXT: s_cmp_lg_u64 s[0:1], 0 +; VI-NEXT: s_addc_u32 s0, s5, s7 +; VI-NEXT: v_mov_b32_e32 v4, s2 +; VI-NEXT: v_mov_b32_e32 v5, s0 +; VI-NEXT: s_cselect_b64 s[0:1], -1, 0 ; VI-NEXT: v_mov_b32_e32 v3, s3 -; VI-NEXT: flat_store_dwordx2 v[0:1], v[6:7] -; VI-NEXT: v_cndmask_b32_e64 v0, 0, 1, vcc +; VI-NEXT: flat_store_dwordx2 v[0:1], v[4:5] +; VI-NEXT: v_cndmask_b32_e64 v0, 0, 1, s[0:1] ; VI-NEXT: flat_store_byte v[2:3], v0 ; VI-NEXT: s_endpgm ; ; GFX9-LABEL: s_uaddo_i64: ; GFX9: ; %bb.0: ; GFX9-NEXT: s_load_dwordx8 s[8:15], s[4:5], 0x24 -; GFX9-NEXT: v_mov_b32_e32 v4, 0 +; GFX9-NEXT: v_mov_b32_e32 v2, 0 ; GFX9-NEXT: s_waitcnt lgkmcnt(0) -; GFX9-NEXT: s_add_u32 s0, s12, s14 -; GFX9-NEXT: v_mov_b32_e32 v0, s12 -; GFX9-NEXT: v_mov_b32_e32 v1, s13 -; GFX9-NEXT: s_addc_u32 s1, s13, s15 -; GFX9-NEXT: v_mov_b32_e32 v3, s1 -; GFX9-NEXT: v_cmp_lt_u64_e32 vcc, s[0:1], v[0:1] -; GFX9-NEXT: v_mov_b32_e32 v2, s0 -; GFX9-NEXT: v_cndmask_b32_e64 v0, 0, 1, vcc -; GFX9-NEXT: global_store_dwordx2 v4, v[2:3], s[8:9] -; GFX9-NEXT: global_store_byte v4, v0, s[10:11] +; GFX9-NEXT: s_add_u32 s2, s12, s14 +; GFX9-NEXT: s_cselect_b64 s[0:1], -1, 0 +; GFX9-NEXT: s_cmp_lg_u64 s[0:1], 0 +; GFX9-NEXT: s_addc_u32 s0, s13, s15 +; GFX9-NEXT: v_mov_b32_e32 v0, s2 +; GFX9-NEXT: v_mov_b32_e32 v1, s0 +; GFX9-NEXT: s_cselect_b64 s[0:1], -1, 0 +; GFX9-NEXT: v_cndmask_b32_e64 v3, 0, 1, s[0:1] +; GFX9-NEXT: global_store_dwordx2 v2, v[0:1], s[8:9] +; GFX9-NEXT: global_store_byte v2, v3, s[10:11] ; GFX9-NEXT: s_endpgm ; ; GFX10-LABEL: s_uaddo_i64: @@ -497,10 +504,12 @@ define amdgpu_kernel void @s_uaddo_i64(ptr addrspace(1) %out, ptr addrspace(1) % ; GFX10-NEXT: v_mov_b32_e32 v2, 0 ; GFX10-NEXT: s_waitcnt lgkmcnt(0) ; GFX10-NEXT: s_add_u32 s0, s12, s14 -; GFX10-NEXT: s_addc_u32 s1, s13, s15 +; GFX10-NEXT: s_cselect_b32 s1, -1, 0 ; GFX10-NEXT: v_mov_b32_e32 v0, s0 +; GFX10-NEXT: s_cmp_lg_u32 s1, 0 +; GFX10-NEXT: s_addc_u32 s1, s13, s15 +; GFX10-NEXT: s_cselect_b32 s0, -1, 0 ; GFX10-NEXT: v_mov_b32_e32 v1, s1 -; GFX10-NEXT: v_cmp_lt_u64_e64 s0, s[0:1], s[12:13] ; GFX10-NEXT: v_cndmask_b32_e64 v3, 0, 1, s0 ; GFX10-NEXT: global_store_dwordx2 v2, v[0:1], s[8:9] ; GFX10-NEXT: global_store_byte v2, v3, s[10:11] @@ -510,12 +519,13 @@ define amdgpu_kernel void @s_uaddo_i64(ptr addrspace(1) %out, ptr addrspace(1) % ; GFX11: ; %bb.0: ; GFX11-NEXT: s_load_b256 s[0:7], s[4:5], 0x24 ; GFX11-NEXT: s_waitcnt lgkmcnt(0) -; GFX11-NEXT: s_add_u32 s6, s4, s6 -; GFX11-NEXT: s_addc_u32 s7, s5, s7 -; GFX11-NEXT: v_mov_b32_e32 v0, s6 -; GFX11-NEXT: v_cmp_lt_u64_e64 s4, s[6:7], s[4:5] -; GFX11-NEXT: v_dual_mov_b32 v2, 0 :: v_dual_mov_b32 v1, s7 -; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_2) +; GFX11-NEXT: s_add_u32 s4, s4, s6 +; GFX11-NEXT: s_cselect_b32 s6, -1, 0 +; GFX11-NEXT: v_mov_b32_e32 v0, s4 +; GFX11-NEXT: s_cmp_lg_u32 s6, 0 +; GFX11-NEXT: s_addc_u32 s5, s5, s7 +; GFX11-NEXT: s_cselect_b32 s4, -1, 0 +; GFX11-NEXT: v_dual_mov_b32 v2, 0 :: v_dual_mov_b32 v1, s5 ; GFX11-NEXT: v_cndmask_b32_e64 v3, 0, 1, s4 ; GFX11-NEXT: s_clause 0x1 ; GFX11-NEXT: global_store_b64 v2, v[0:1], s[0:1] @@ -551,10 +561,10 @@ define amdgpu_kernel void @v_uaddo_i64(ptr addrspace(1) %out, ptr addrspace(1) % ; SI-NEXT: s_mov_b32 s4, s2 ; SI-NEXT: s_mov_b32 s5, s3 ; SI-NEXT: s_waitcnt vmcnt(0) -; SI-NEXT: v_add_i32_e32 v2, vcc, v0, v2 -; SI-NEXT: v_addc_u32_e32 v3, vcc, v1, v3, vcc -; SI-NEXT: v_cmp_lt_u64_e32 vcc, v[2:3], v[0:1] -; SI-NEXT: buffer_store_dwordx2 v[2:3], off, s[8:11], 0 +; SI-NEXT: v_add_i32_e32 v0, vcc, v0, v2 +; SI-NEXT: v_addc_u32_e32 v1, vcc, v1, v3, vcc +; SI-NEXT: buffer_store_dwordx2 v[0:1], off, s[8:11], 0 +; SI-NEXT: s_waitcnt expcnt(0) ; SI-NEXT: v_cndmask_b32_e64 v0, 0, 1, vcc ; SI-NEXT: buffer_store_byte v0, off, s[4:7], 0 ; SI-NEXT: s_endpgm @@ -574,10 +584,9 @@ define amdgpu_kernel void @v_uaddo_i64(ptr addrspace(1) %out, ptr addrspace(1) % ; VI-NEXT: v_mov_b32_e32 v6, s2 ; VI-NEXT: v_mov_b32_e32 v7, s3 ; VI-NEXT: s_waitcnt vmcnt(0) -; VI-NEXT: v_add_u32_e32 v2, vcc, v0, v2 -; VI-NEXT: v_addc_u32_e32 v3, vcc, v1, v3, vcc -; VI-NEXT: v_cmp_lt_u64_e32 vcc, v[2:3], v[0:1] -; VI-NEXT: flat_store_dwordx2 v[4:5], v[2:3] +; VI-NEXT: v_add_u32_e32 v0, vcc, v0, v2 +; VI-NEXT: v_addc_u32_e32 v1, vcc, v1, v3, vcc +; VI-NEXT: flat_store_dwordx2 v[4:5], v[0:1] ; VI-NEXT: v_cndmask_b32_e64 v0, 0, 1, vcc ; VI-NEXT: flat_store_byte v[6:7], v0 ; VI-NEXT: s_endpgm @@ -590,10 +599,9 @@ define amdgpu_kernel void @v_uaddo_i64(ptr addrspace(1) %out, ptr addrspace(1) % ; GFX9-NEXT: global_load_dwordx2 v[0:1], v4, s[12:13] ; GFX9-NEXT: global_load_dwordx2 v[2:3], v4, s[14:15] ; GFX9-NEXT: s_waitcnt vmcnt(0) -; GFX9-NEXT: v_add_co_u32_e32 v2, vcc, v0, v2 -; GFX9-NEXT: v_addc_co_u32_e32 v3, vcc, v1, v3, vcc -; GFX9-NEXT: v_cmp_lt_u64_e32 vcc, v[2:3], v[0:1] -; GFX9-NEXT: global_store_dwordx2 v4, v[2:3], s[8:9] +; GFX9-NEXT: v_add_co_u32_e32 v0, vcc, v0, v2 +; GFX9-NEXT: v_addc_co_u32_e32 v1, vcc, v1, v3, vcc +; GFX9-NEXT: global_store_dwordx2 v4, v[0:1], s[8:9] ; GFX9-NEXT: v_cndmask_b32_e64 v0, 0, 1, vcc ; GFX9-NEXT: global_store_byte v4, v0, s[10:11] ; GFX9-NEXT: s_endpgm @@ -607,12 +615,11 @@ define amdgpu_kernel void @v_uaddo_i64(ptr addrspace(1) %out, ptr addrspace(1) % ; GFX10-NEXT: global_load_dwordx2 v[0:1], v4, s[12:13] ; GFX10-NEXT: global_load_dwordx2 v[2:3], v4, s[14:15] ; GFX10-NEXT: s_waitcnt vmcnt(0) -; GFX10-NEXT: v_add_co_u32 v2, vcc_lo, v0, v2 -; GFX10-NEXT: v_add_co_ci_u32_e32 v3, vcc_lo, v1, v3, vcc_lo -; GFX10-NEXT: v_cmp_lt_u64_e32 vcc_lo, v[2:3], v[0:1] -; GFX10-NEXT: v_cndmask_b32_e64 v0, 0, 1, vcc_lo -; GFX10-NEXT: global_store_dwordx2 v4, v[2:3], s[8:9] -; GFX10-NEXT: global_store_byte v4, v0, s[10:11] +; GFX10-NEXT: v_add_co_u32 v0, vcc_lo, v0, v2 +; GFX10-NEXT: v_add_co_ci_u32_e32 v1, vcc_lo, v1, v3, vcc_lo +; GFX10-NEXT: v_cndmask_b32_e64 v2, 0, 1, vcc_lo +; GFX10-NEXT: global_store_dwordx2 v4, v[0:1], s[8:9] +; GFX10-NEXT: global_store_byte v4, v2, s[10:11] ; GFX10-NEXT: s_endpgm ; ; GFX11-LABEL: v_uaddo_i64: @@ -624,14 +631,12 @@ define amdgpu_kernel void @v_uaddo_i64(ptr addrspace(1) %out, ptr addrspace(1) % ; GFX11-NEXT: global_load_b64 v[0:1], v4, s[4:5] ; GFX11-NEXT: global_load_b64 v[2:3], v4, s[6:7] ; GFX11-NEXT: s_waitcnt vmcnt(0) -; GFX11-NEXT: v_add_co_u32 v2, vcc_lo, v0, v2 -; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1) -; GFX11-NEXT: v_add_co_ci_u32_e64 v3, null, v1, v3, vcc_lo -; GFX11-NEXT: v_cmp_lt_u64_e32 vcc_lo, v[2:3], v[0:1] -; GFX11-NEXT: v_cndmask_b32_e64 v0, 0, 1, vcc_lo +; GFX11-NEXT: v_add_co_u32 v0, vcc_lo, v0, v2 +; GFX11-NEXT: v_add_co_ci_u32_e32 v1, vcc_lo, v1, v3, vcc_lo +; GFX11-NEXT: v_cndmask_b32_e64 v2, 0, 1, vcc_lo ; GFX11-NEXT: s_clause 0x1 -; GFX11-NEXT: global_store_b64 v4, v[2:3], s[0:1] -; GFX11-NEXT: global_store_b8 v4, v0, s[2:3] +; GFX11-NEXT: global_store_b64 v4, v[0:1], s[0:1] +; GFX11-NEXT: global_store_b8 v4, v2, s[2:3] ; GFX11-NEXT: s_endpgm %tid = call i32 @llvm.amdgcn.workitem.id.x() %tid.ext = sext i32 %tid to i64 diff --git a/llvm/test/CodeGen/AMDGPU/uaddsat.ll b/llvm/test/CodeGen/AMDGPU/uaddsat.ll index 923017400adb1..7f89581d00fde 100644 --- a/llvm/test/CodeGen/AMDGPU/uaddsat.ll +++ b/llvm/test/CodeGen/AMDGPU/uaddsat.ll @@ -693,52 +693,47 @@ define i64 @v_uaddsat_i64(i64 %lhs, i64 %rhs) { ; GFX6-LABEL: v_uaddsat_i64: ; GFX6: ; %bb.0: ; GFX6-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) -; GFX6-NEXT: v_add_i32_e32 v2, vcc, v0, v2 -; GFX6-NEXT: v_addc_u32_e32 v3, vcc, v1, v3, vcc -; GFX6-NEXT: v_cmp_lt_u64_e32 vcc, v[2:3], v[0:1] -; GFX6-NEXT: v_cndmask_b32_e64 v0, v2, -1, vcc -; GFX6-NEXT: v_cndmask_b32_e64 v1, v3, -1, vcc +; GFX6-NEXT: v_add_i32_e32 v0, vcc, v0, v2 +; GFX6-NEXT: v_addc_u32_e32 v1, vcc, v1, v3, vcc +; GFX6-NEXT: v_cndmask_b32_e64 v0, v0, -1, vcc +; GFX6-NEXT: v_cndmask_b32_e64 v1, v1, -1, vcc ; GFX6-NEXT: s_setpc_b64 s[30:31] ; ; GFX8-LABEL: v_uaddsat_i64: ; GFX8: ; %bb.0: ; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) -; GFX8-NEXT: v_add_u32_e32 v2, vcc, v0, v2 -; GFX8-NEXT: v_addc_u32_e32 v3, vcc, v1, v3, vcc -; GFX8-NEXT: v_cmp_lt_u64_e32 vcc, v[2:3], v[0:1] -; GFX8-NEXT: v_cndmask_b32_e64 v0, v2, -1, vcc -; GFX8-NEXT: v_cndmask_b32_e64 v1, v3, -1, vcc +; GFX8-NEXT: v_add_u32_e32 v0, vcc, v0, v2 +; GFX8-NEXT: v_addc_u32_e32 v1, vcc, v1, v3, vcc +; GFX8-NEXT: v_cndmask_b32_e64 v0, v0, -1, vcc +; GFX8-NEXT: v_cndmask_b32_e64 v1, v1, -1, vcc ; GFX8-NEXT: s_setpc_b64 s[30:31] ; ; GFX9-LABEL: v_uaddsat_i64: ; GFX9: ; %bb.0: ; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) -; GFX9-NEXT: v_add_co_u32_e32 v2, vcc, v0, v2 -; GFX9-NEXT: v_addc_co_u32_e32 v3, vcc, v1, v3, vcc -; GFX9-NEXT: v_cmp_lt_u64_e32 vcc, v[2:3], v[0:1] -; GFX9-NEXT: v_cndmask_b32_e64 v0, v2, -1, vcc -; GFX9-NEXT: v_cndmask_b32_e64 v1, v3, -1, vcc +; GFX9-NEXT: v_add_co_u32_e32 v0, vcc, v0, v2 +; GFX9-NEXT: v_addc_co_u32_e32 v1, vcc, v1, v3, vcc +; GFX9-NEXT: v_cndmask_b32_e64 v0, v0, -1, vcc +; GFX9-NEXT: v_cndmask_b32_e64 v1, v1, -1, vcc ; GFX9-NEXT: s_setpc_b64 s[30:31] ; ; GFX10-LABEL: v_uaddsat_i64: ; GFX10: ; %bb.0: ; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) -; GFX10-NEXT: v_add_co_u32 v2, vcc_lo, v0, v2 -; GFX10-NEXT: v_add_co_ci_u32_e32 v3, vcc_lo, v1, v3, vcc_lo -; GFX10-NEXT: v_cmp_lt_u64_e32 vcc_lo, v[2:3], v[0:1] -; GFX10-NEXT: v_cndmask_b32_e64 v0, v2, -1, vcc_lo -; GFX10-NEXT: v_cndmask_b32_e64 v1, v3, -1, vcc_lo +; GFX10-NEXT: v_add_co_u32 v0, vcc_lo, v0, v2 +; GFX10-NEXT: v_add_co_ci_u32_e32 v1, vcc_lo, v1, v3, vcc_lo +; GFX10-NEXT: v_cndmask_b32_e64 v0, v0, -1, vcc_lo +; GFX10-NEXT: v_cndmask_b32_e64 v1, v1, -1, vcc_lo ; GFX10-NEXT: s_setpc_b64 s[30:31] ; ; GFX11-LABEL: v_uaddsat_i64: ; GFX11: ; %bb.0: ; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) -; GFX11-NEXT: v_add_co_u32 v2, vcc_lo, v0, v2 -; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1) -; GFX11-NEXT: v_add_co_ci_u32_e64 v3, null, v1, v3, vcc_lo -; GFX11-NEXT: v_cmp_lt_u64_e32 vcc_lo, v[2:3], v[0:1] -; GFX11-NEXT: v_cndmask_b32_e64 v0, v2, -1, vcc_lo -; GFX11-NEXT: v_cndmask_b32_e64 v1, v3, -1, vcc_lo +; GFX11-NEXT: v_add_co_u32 v0, vcc_lo, v0, v2 +; GFX11-NEXT: v_add_co_ci_u32_e32 v1, vcc_lo, v1, v3, vcc_lo +; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2) +; GFX11-NEXT: v_cndmask_b32_e64 v0, v0, -1, vcc_lo +; GFX11-NEXT: v_cndmask_b32_e64 v1, v1, -1, vcc_lo ; GFX11-NEXT: s_setpc_b64 s[30:31] %result = call i64 @llvm.uadd.sat.i64(i64 %lhs, i64 %rhs) ret i64 %result diff --git a/llvm/test/CodeGen/AMDGPU/udiv64.ll b/llvm/test/CodeGen/AMDGPU/udiv64.ll index dc11e81476a7e..09e79102d7564 100644 --- a/llvm/test/CodeGen/AMDGPU/udiv64.ll +++ b/llvm/test/CodeGen/AMDGPU/udiv64.ll @@ -146,8 +146,11 @@ define amdgpu_kernel void @s_test_udiv_i64(ptr addrspace(1) %out, i64 %x, i64 %y ; GCN-IR-NEXT: s_cbranch_vccz .LBB0_5 ; GCN-IR-NEXT: ; %bb.1: ; %udiv-bb1 ; GCN-IR-NEXT: s_add_u32 s14, s12, 1 -; GCN-IR-NEXT: s_addc_u32 s15, s13, 0 -; GCN-IR-NEXT: v_cmp_eq_u64_e64 s[8:9], s[14:15], 0 +; GCN-IR-NEXT: s_cselect_b64 s[8:9], -1, 0 +; GCN-IR-NEXT: s_or_b32 s8, s8, s9 +; GCN-IR-NEXT: s_cmp_lg_u32 s8, 0 +; GCN-IR-NEXT: s_addc_u32 s8, s13, 0 +; GCN-IR-NEXT: s_cselect_b64 s[8:9], -1, 0 ; GCN-IR-NEXT: s_sub_i32 s12, 63, s12 ; GCN-IR-NEXT: s_andn2_b64 vcc, exec, s[8:9] ; GCN-IR-NEXT: s_lshl_b64 s[8:9], s[2:3], s12 @@ -157,9 +160,9 @@ define amdgpu_kernel void @s_test_udiv_i64(ptr addrspace(1) %out, i64 %x, i64 %y ; GCN-IR-NEXT: s_add_u32 s14, s6, -1 ; GCN-IR-NEXT: s_addc_u32 s15, s7, -1 ; GCN-IR-NEXT: s_not_b64 s[2:3], s[10:11] -; GCN-IR-NEXT: s_add_u32 s2, s2, s16 -; GCN-IR-NEXT: s_addc_u32 s3, s3, 0 -; GCN-IR-NEXT: s_mov_b64 s[10:11], 0 +; GCN-IR-NEXT: s_add_u32 s10, s2, s16 +; GCN-IR-NEXT: s_addc_u32 s11, s3, 0 +; GCN-IR-NEXT: s_mov_b64 s[2:3], 0 ; GCN-IR-NEXT: s_mov_b32 s5, 0 ; GCN-IR-NEXT: .LBB0_3: ; %udiv-do-while ; GCN-IR-NEXT: ; =>This Inner Loop Header: Depth=1 @@ -167,19 +170,22 @@ define amdgpu_kernel void @s_test_udiv_i64(ptr addrspace(1) %out, i64 %x, i64 %y ; GCN-IR-NEXT: s_lshr_b32 s4, s9, 31 ; GCN-IR-NEXT: s_lshl_b64 s[8:9], s[8:9], 1 ; GCN-IR-NEXT: s_or_b64 s[12:13], s[12:13], s[4:5] -; GCN-IR-NEXT: s_or_b64 s[8:9], s[10:11], s[8:9] -; GCN-IR-NEXT: s_sub_u32 s4, s14, s12 -; GCN-IR-NEXT: s_subb_u32 s4, s15, s13 -; GCN-IR-NEXT: s_ashr_i32 s10, s4, 31 -; GCN-IR-NEXT: s_mov_b32 s11, s10 -; GCN-IR-NEXT: s_and_b32 s4, s10, 1 -; GCN-IR-NEXT: s_and_b64 s[10:11], s[10:11], s[6:7] -; GCN-IR-NEXT: s_sub_u32 s12, s12, s10 -; GCN-IR-NEXT: s_subb_u32 s13, s13, s11 -; GCN-IR-NEXT: s_add_u32 s2, s2, 1 -; GCN-IR-NEXT: s_addc_u32 s3, s3, 0 -; GCN-IR-NEXT: v_cmp_eq_u64_e64 s[16:17], s[2:3], 0 -; GCN-IR-NEXT: s_mov_b64 s[10:11], s[4:5] +; GCN-IR-NEXT: s_or_b64 s[8:9], s[2:3], s[8:9] +; GCN-IR-NEXT: s_sub_u32 s2, s14, s12 +; GCN-IR-NEXT: s_subb_u32 s2, s15, s13 +; GCN-IR-NEXT: s_ashr_i32 s2, s2, 31 +; GCN-IR-NEXT: s_mov_b32 s3, s2 +; GCN-IR-NEXT: s_and_b32 s4, s2, 1 +; GCN-IR-NEXT: s_and_b64 s[16:17], s[2:3], s[6:7] +; GCN-IR-NEXT: s_sub_u32 s12, s12, s16 +; GCN-IR-NEXT: s_subb_u32 s13, s13, s17 +; GCN-IR-NEXT: s_add_u32 s10, s10, 1 +; GCN-IR-NEXT: s_cselect_b64 s[16:17], -1, 0 +; GCN-IR-NEXT: s_or_b32 s16, s16, s17 +; GCN-IR-NEXT: s_cmp_lg_u32 s16, 0 +; GCN-IR-NEXT: s_addc_u32 s11, s11, 0 +; GCN-IR-NEXT: s_cselect_b64 s[16:17], -1, 0 +; GCN-IR-NEXT: s_mov_b64 s[2:3], s[4:5] ; GCN-IR-NEXT: s_and_b64 vcc, exec, s[16:17] ; GCN-IR-NEXT: s_cbranch_vccz .LBB0_3 ; GCN-IR-NEXT: .LBB0_4: ; %Flow7 @@ -313,19 +319,19 @@ define i64 @v_test_udiv_i64(i64 %x, i64 %y) { ; GCN-IR-NEXT: v_ffbh_u32_e32 v4, v2 ; GCN-IR-NEXT: v_add_i32_e64 v4, s[6:7], 32, v4 ; GCN-IR-NEXT: v_ffbh_u32_e32 v5, v3 -; GCN-IR-NEXT: v_min_u32_e32 v14, v4, v5 +; GCN-IR-NEXT: v_min_u32_e32 v8, v4, v5 ; GCN-IR-NEXT: v_ffbh_u32_e32 v4, v0 ; GCN-IR-NEXT: v_add_i32_e64 v4, s[6:7], 32, v4 ; GCN-IR-NEXT: v_ffbh_u32_e32 v5, v1 -; GCN-IR-NEXT: v_min_u32_e32 v15, v4, v5 -; GCN-IR-NEXT: v_sub_i32_e64 v8, s[6:7], v14, v15 +; GCN-IR-NEXT: v_min_u32_e32 v9, v4, v5 +; GCN-IR-NEXT: v_sub_i32_e64 v6, s[6:7], v8, v9 ; GCN-IR-NEXT: v_cmp_eq_u64_e32 vcc, 0, v[2:3] ; GCN-IR-NEXT: v_cmp_eq_u64_e64 s[4:5], 0, v[0:1] -; GCN-IR-NEXT: v_subb_u32_e64 v9, s[6:7], 0, 0, s[6:7] -; GCN-IR-NEXT: v_cmp_lt_u64_e64 s[6:7], 63, v[8:9] +; GCN-IR-NEXT: v_subb_u32_e64 v7, s[6:7], 0, 0, s[6:7] +; GCN-IR-NEXT: v_cmp_lt_u64_e64 s[6:7], 63, v[6:7] ; GCN-IR-NEXT: s_or_b64 s[4:5], vcc, s[4:5] ; GCN-IR-NEXT: s_or_b64 s[4:5], s[4:5], s[6:7] -; GCN-IR-NEXT: v_cmp_ne_u64_e32 vcc, 63, v[8:9] +; GCN-IR-NEXT: v_cmp_ne_u64_e32 vcc, 63, v[6:7] ; GCN-IR-NEXT: s_xor_b64 s[6:7], s[4:5], -1 ; GCN-IR-NEXT: v_cndmask_b32_e64 v4, v1, 0, s[4:5] ; GCN-IR-NEXT: v_cndmask_b32_e64 v5, v0, 0, s[4:5] @@ -333,55 +339,54 @@ define i64 @v_test_udiv_i64(i64 %x, i64 %y) { ; GCN-IR-NEXT: s_and_saveexec_b64 s[6:7], s[4:5] ; GCN-IR-NEXT: s_cbranch_execz .LBB1_6 ; GCN-IR-NEXT: ; %bb.1: ; %udiv-bb1 -; GCN-IR-NEXT: v_add_i32_e32 v10, vcc, 1, v8 -; GCN-IR-NEXT: v_addc_u32_e32 v11, vcc, 0, v9, vcc -; GCN-IR-NEXT: v_sub_i32_e64 v4, s[4:5], 63, v8 -; GCN-IR-NEXT: v_mov_b32_e32 v6, 0 -; GCN-IR-NEXT: v_cmp_ne_u64_e32 vcc, 0, v[10:11] +; GCN-IR-NEXT: v_add_i32_e32 v10, vcc, 1, v6 +; GCN-IR-NEXT: v_addc_u32_e32 v4, vcc, 0, v7, vcc +; GCN-IR-NEXT: v_sub_i32_e64 v4, s[4:5], 63, v6 ; GCN-IR-NEXT: v_lshl_b64 v[4:5], v[0:1], v4 +; GCN-IR-NEXT: v_mov_b32_e32 v6, 0 ; GCN-IR-NEXT: v_mov_b32_e32 v7, 0 -; GCN-IR-NEXT: s_and_saveexec_b64 s[4:5], vcc -; GCN-IR-NEXT: s_xor_b64 s[8:9], exec, s[4:5] +; GCN-IR-NEXT: s_xor_b64 s[4:5], vcc, -1 +; GCN-IR-NEXT: s_and_saveexec_b64 s[8:9], s[4:5] +; GCN-IR-NEXT: s_xor_b64 s[4:5], exec, s[8:9] ; GCN-IR-NEXT: s_cbranch_execz .LBB1_5 ; GCN-IR-NEXT: ; %bb.2: ; %udiv-preheader -; GCN-IR-NEXT: v_add_i32_e32 v12, vcc, -1, v2 -; GCN-IR-NEXT: v_lshr_b64 v[8:9], v[0:1], v10 -; GCN-IR-NEXT: v_addc_u32_e32 v13, vcc, -1, v3, vcc -; GCN-IR-NEXT: v_not_b32_e32 v0, v14 -; GCN-IR-NEXT: v_add_i32_e32 v0, vcc, v0, v15 -; GCN-IR-NEXT: v_mov_b32_e32 v10, 0 -; GCN-IR-NEXT: v_addc_u32_e64 v1, s[4:5], -1, 0, vcc -; GCN-IR-NEXT: s_mov_b64 s[10:11], 0 -; GCN-IR-NEXT: v_mov_b32_e32 v11, 0 +; GCN-IR-NEXT: v_lshr_b64 v[0:1], v[0:1], v10 +; GCN-IR-NEXT: v_add_i32_e32 v10, vcc, -1, v2 +; GCN-IR-NEXT: v_addc_u32_e32 v11, vcc, -1, v3, vcc +; GCN-IR-NEXT: v_not_b32_e32 v6, v8 +; GCN-IR-NEXT: v_add_i32_e32 v12, vcc, v6, v9 +; GCN-IR-NEXT: v_addc_u32_e64 v13, s[8:9], -1, 0, vcc +; GCN-IR-NEXT: v_mov_b32_e32 v8, 0 +; GCN-IR-NEXT: s_mov_b64 s[8:9], 0 +; GCN-IR-NEXT: v_mov_b32_e32 v9, 0 ; GCN-IR-NEXT: v_mov_b32_e32 v7, 0 ; GCN-IR-NEXT: .LBB1_3: ; %udiv-do-while ; GCN-IR-NEXT: ; =>This Inner Loop Header: Depth=1 -; GCN-IR-NEXT: v_lshl_b64 v[8:9], v[8:9], 1 +; GCN-IR-NEXT: v_lshl_b64 v[0:1], v[0:1], 1 ; GCN-IR-NEXT: v_lshrrev_b32_e32 v6, 31, v5 -; GCN-IR-NEXT: v_or_b32_e32 v8, v8, v6 +; GCN-IR-NEXT: v_or_b32_e32 v0, v0, v6 ; GCN-IR-NEXT: v_lshl_b64 v[4:5], v[4:5], 1 -; GCN-IR-NEXT: v_sub_i32_e32 v6, vcc, v12, v8 -; GCN-IR-NEXT: v_subb_u32_e32 v6, vcc, v13, v9, vcc -; GCN-IR-NEXT: v_or_b32_e32 v4, v10, v4 -; GCN-IR-NEXT: v_ashrrev_i32_e32 v10, 31, v6 -; GCN-IR-NEXT: v_add_i32_e32 v0, vcc, 1, v0 -; GCN-IR-NEXT: v_or_b32_e32 v5, v11, v5 -; GCN-IR-NEXT: v_and_b32_e32 v6, 1, v10 -; GCN-IR-NEXT: v_and_b32_e32 v11, v10, v3 -; GCN-IR-NEXT: v_and_b32_e32 v10, v10, v2 -; GCN-IR-NEXT: v_addc_u32_e32 v1, vcc, 0, v1, vcc -; GCN-IR-NEXT: v_cmp_eq_u64_e32 vcc, 0, v[0:1] -; GCN-IR-NEXT: v_sub_i32_e64 v8, s[4:5], v8, v10 -; GCN-IR-NEXT: v_subb_u32_e64 v9, s[4:5], v9, v11, s[4:5] -; GCN-IR-NEXT: v_mov_b32_e32 v11, v7 -; GCN-IR-NEXT: s_or_b64 s[10:11], vcc, s[10:11] -; GCN-IR-NEXT: v_mov_b32_e32 v10, v6 -; GCN-IR-NEXT: s_andn2_b64 exec, exec, s[10:11] +; GCN-IR-NEXT: v_sub_i32_e32 v6, vcc, v10, v0 +; GCN-IR-NEXT: v_subb_u32_e32 v6, vcc, v11, v1, vcc +; GCN-IR-NEXT: v_or_b32_e32 v4, v8, v4 +; GCN-IR-NEXT: v_ashrrev_i32_e32 v8, 31, v6 +; GCN-IR-NEXT: v_or_b32_e32 v5, v9, v5 +; GCN-IR-NEXT: v_and_b32_e32 v6, 1, v8 +; GCN-IR-NEXT: v_and_b32_e32 v9, v8, v3 +; GCN-IR-NEXT: v_and_b32_e32 v8, v8, v2 +; GCN-IR-NEXT: v_sub_i32_e32 v0, vcc, v0, v8 +; GCN-IR-NEXT: v_subb_u32_e32 v1, vcc, v1, v9, vcc +; GCN-IR-NEXT: v_add_i32_e32 v12, vcc, 1, v12 +; GCN-IR-NEXT: v_addc_u32_e32 v13, vcc, 0, v13, vcc +; GCN-IR-NEXT: v_mov_b32_e32 v9, v7 +; GCN-IR-NEXT: s_or_b64 s[8:9], vcc, s[8:9] +; GCN-IR-NEXT: v_mov_b32_e32 v8, v6 +; GCN-IR-NEXT: s_andn2_b64 exec, exec, s[8:9] ; GCN-IR-NEXT: s_cbranch_execnz .LBB1_3 ; GCN-IR-NEXT: ; %bb.4: ; %Flow -; GCN-IR-NEXT: s_or_b64 exec, exec, s[10:11] -; GCN-IR-NEXT: .LBB1_5: ; %Flow4 ; GCN-IR-NEXT: s_or_b64 exec, exec, s[8:9] +; GCN-IR-NEXT: .LBB1_5: ; %Flow4 +; GCN-IR-NEXT: s_or_b64 exec, exec, s[4:5] ; GCN-IR-NEXT: v_lshl_b64 v[0:1], v[4:5], 1 ; GCN-IR-NEXT: v_or_b32_e32 v4, v7, v1 ; GCN-IR-NEXT: v_or_b32_e32 v5, v6, v0 @@ -927,34 +932,37 @@ define amdgpu_kernel void @s_test_udiv_k_num_i64(ptr addrspace(1) %out, i64 %x) ; GCN-IR-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x9 ; GCN-IR-NEXT: s_mov_b64 s[4:5], 0 ; GCN-IR-NEXT: s_waitcnt lgkmcnt(0) -; GCN-IR-NEXT: s_flbit_i32_b64 s12, s[2:3] -; GCN-IR-NEXT: s_add_u32 s8, s12, 0xffffffc5 +; GCN-IR-NEXT: s_flbit_i32_b64 s14, s[2:3] +; GCN-IR-NEXT: s_add_u32 s8, s14, 0xffffffc5 ; GCN-IR-NEXT: s_addc_u32 s9, 0, -1 ; GCN-IR-NEXT: v_cmp_eq_u64_e64 s[6:7], s[2:3], 0 ; GCN-IR-NEXT: v_cmp_gt_u64_e64 s[10:11], s[8:9], 63 -; GCN-IR-NEXT: v_cmp_eq_u64_e64 s[14:15], s[8:9], 63 +; GCN-IR-NEXT: v_cmp_eq_u64_e64 s[12:13], s[8:9], 63 ; GCN-IR-NEXT: s_or_b64 s[10:11], s[6:7], s[10:11] ; GCN-IR-NEXT: s_and_b64 s[6:7], s[10:11], exec ; GCN-IR-NEXT: s_cselect_b32 s6, 0, 24 -; GCN-IR-NEXT: s_or_b64 s[10:11], s[10:11], s[14:15] +; GCN-IR-NEXT: s_or_b64 s[10:11], s[10:11], s[12:13] ; GCN-IR-NEXT: s_andn2_b64 vcc, exec, s[10:11] ; GCN-IR-NEXT: s_mov_b32 s7, 0 ; GCN-IR-NEXT: s_cbranch_vccz .LBB8_5 ; GCN-IR-NEXT: ; %bb.1: ; %udiv-bb1 ; GCN-IR-NEXT: s_add_u32 s10, s8, 1 -; GCN-IR-NEXT: s_addc_u32 s11, s9, 0 -; GCN-IR-NEXT: v_cmp_eq_u64_e64 s[6:7], s[10:11], 0 +; GCN-IR-NEXT: s_cselect_b64 s[6:7], -1, 0 +; GCN-IR-NEXT: s_or_b32 s6, s6, s7 +; GCN-IR-NEXT: s_cmp_lg_u32 s6, 0 +; GCN-IR-NEXT: s_addc_u32 s6, s9, 0 +; GCN-IR-NEXT: s_cselect_b64 s[6:7], -1, 0 ; GCN-IR-NEXT: s_sub_i32 s8, 63, s8 ; GCN-IR-NEXT: s_andn2_b64 vcc, exec, s[6:7] ; GCN-IR-NEXT: s_lshl_b64 s[6:7], 24, s8 ; GCN-IR-NEXT: s_cbranch_vccz .LBB8_4 ; GCN-IR-NEXT: ; %bb.2: ; %udiv-preheader ; GCN-IR-NEXT: s_lshr_b64 s[10:11], 24, s10 -; GCN-IR-NEXT: s_add_u32 s14, s2, -1 -; GCN-IR-NEXT: s_addc_u32 s15, s3, -1 -; GCN-IR-NEXT: s_sub_u32 s8, 58, s12 -; GCN-IR-NEXT: s_subb_u32 s9, 0, 0 -; GCN-IR-NEXT: s_mov_b64 s[12:13], 0 +; GCN-IR-NEXT: s_add_u32 s12, s2, -1 +; GCN-IR-NEXT: s_addc_u32 s13, s3, -1 +; GCN-IR-NEXT: s_sub_u32 s14, 58, s14 +; GCN-IR-NEXT: s_subb_u32 s15, 0, 0 +; GCN-IR-NEXT: s_mov_b64 s[8:9], 0 ; GCN-IR-NEXT: s_mov_b32 s5, 0 ; GCN-IR-NEXT: .LBB8_3: ; %udiv-do-while ; GCN-IR-NEXT: ; =>This Inner Loop Header: Depth=1 @@ -962,19 +970,22 @@ define amdgpu_kernel void @s_test_udiv_k_num_i64(ptr addrspace(1) %out, i64 %x) ; GCN-IR-NEXT: s_lshr_b32 s4, s7, 31 ; GCN-IR-NEXT: s_lshl_b64 s[6:7], s[6:7], 1 ; GCN-IR-NEXT: s_or_b64 s[10:11], s[10:11], s[4:5] -; GCN-IR-NEXT: s_or_b64 s[6:7], s[12:13], s[6:7] -; GCN-IR-NEXT: s_sub_u32 s4, s14, s10 -; GCN-IR-NEXT: s_subb_u32 s4, s15, s11 -; GCN-IR-NEXT: s_ashr_i32 s12, s4, 31 -; GCN-IR-NEXT: s_mov_b32 s13, s12 -; GCN-IR-NEXT: s_and_b32 s4, s12, 1 -; GCN-IR-NEXT: s_and_b64 s[12:13], s[12:13], s[2:3] -; GCN-IR-NEXT: s_sub_u32 s10, s10, s12 -; GCN-IR-NEXT: s_subb_u32 s11, s11, s13 -; GCN-IR-NEXT: s_add_u32 s8, s8, 1 -; GCN-IR-NEXT: s_addc_u32 s9, s9, 0 -; GCN-IR-NEXT: v_cmp_eq_u64_e64 s[16:17], s[8:9], 0 -; GCN-IR-NEXT: s_mov_b64 s[12:13], s[4:5] +; GCN-IR-NEXT: s_or_b64 s[6:7], s[8:9], s[6:7] +; GCN-IR-NEXT: s_sub_u32 s4, s12, s10 +; GCN-IR-NEXT: s_subb_u32 s4, s13, s11 +; GCN-IR-NEXT: s_ashr_i32 s8, s4, 31 +; GCN-IR-NEXT: s_mov_b32 s9, s8 +; GCN-IR-NEXT: s_and_b32 s4, s8, 1 +; GCN-IR-NEXT: s_and_b64 s[16:17], s[8:9], s[2:3] +; GCN-IR-NEXT: s_sub_u32 s10, s10, s16 +; GCN-IR-NEXT: s_subb_u32 s11, s11, s17 +; GCN-IR-NEXT: s_add_u32 s14, s14, 1 +; GCN-IR-NEXT: s_cselect_b64 s[16:17], -1, 0 +; GCN-IR-NEXT: s_or_b32 s16, s16, s17 +; GCN-IR-NEXT: s_cmp_lg_u32 s16, 0 +; GCN-IR-NEXT: s_addc_u32 s15, s15, 0 +; GCN-IR-NEXT: s_cselect_b64 s[16:17], -1, 0 +; GCN-IR-NEXT: s_mov_b64 s[8:9], s[4:5] ; GCN-IR-NEXT: s_and_b64 vcc, exec, s[16:17] ; GCN-IR-NEXT: s_cbranch_vccz .LBB8_3 ; GCN-IR-NEXT: .LBB8_4: ; %Flow6 @@ -1098,12 +1109,12 @@ define i64 @v_test_udiv_pow2_k_num_i64(i64 %x) { ; GCN-IR-NEXT: v_ffbh_u32_e32 v2, v0 ; GCN-IR-NEXT: v_add_i32_e32 v2, vcc, 32, v2 ; GCN-IR-NEXT: v_ffbh_u32_e32 v3, v1 -; GCN-IR-NEXT: v_min_u32_e32 v10, v2, v3 -; GCN-IR-NEXT: v_add_i32_e32 v6, vcc, 0xffffffd0, v10 -; GCN-IR-NEXT: v_addc_u32_e64 v7, s[6:7], 0, -1, vcc +; GCN-IR-NEXT: v_min_u32_e32 v8, v2, v3 +; GCN-IR-NEXT: v_add_i32_e32 v4, vcc, 0xffffffd0, v8 +; GCN-IR-NEXT: v_addc_u32_e64 v5, s[6:7], 0, -1, vcc ; GCN-IR-NEXT: v_cmp_eq_u64_e64 s[4:5], 0, v[0:1] -; GCN-IR-NEXT: v_cmp_lt_u64_e32 vcc, 63, v[6:7] -; GCN-IR-NEXT: v_cmp_ne_u64_e64 s[6:7], 63, v[6:7] +; GCN-IR-NEXT: v_cmp_lt_u64_e32 vcc, 63, v[4:5] +; GCN-IR-NEXT: v_cmp_ne_u64_e64 s[6:7], 63, v[4:5] ; GCN-IR-NEXT: v_mov_b32_e32 v3, 0x8000 ; GCN-IR-NEXT: s_or_b64 s[4:5], s[4:5], vcc ; GCN-IR-NEXT: v_cndmask_b32_e64 v3, v3, 0, s[4:5] @@ -1113,55 +1124,54 @@ define i64 @v_test_udiv_pow2_k_num_i64(i64 %x) { ; GCN-IR-NEXT: s_and_saveexec_b64 s[6:7], s[4:5] ; GCN-IR-NEXT: s_cbranch_execz .LBB9_6 ; GCN-IR-NEXT: ; %bb.1: ; %udiv-bb1 -; GCN-IR-NEXT: v_add_i32_e32 v8, vcc, 1, v6 -; GCN-IR-NEXT: v_sub_i32_e64 v2, s[4:5], 63, v6 -; GCN-IR-NEXT: v_addc_u32_e32 v9, vcc, 0, v7, vcc -; GCN-IR-NEXT: s_mov_b64 s[4:5], 0x8000 +; GCN-IR-NEXT: v_add_i32_e32 v6, vcc, 1, v4 +; GCN-IR-NEXT: v_addc_u32_e32 v2, vcc, 0, v5, vcc +; GCN-IR-NEXT: v_sub_i32_e64 v2, s[4:5], 63, v4 +; GCN-IR-NEXT: s_mov_b64 s[8:9], 0x8000 +; GCN-IR-NEXT: v_lshl_b64 v[2:3], s[8:9], v2 ; GCN-IR-NEXT: v_mov_b32_e32 v4, 0 -; GCN-IR-NEXT: v_cmp_ne_u64_e32 vcc, 0, v[8:9] -; GCN-IR-NEXT: v_lshl_b64 v[2:3], s[4:5], v2 ; GCN-IR-NEXT: v_mov_b32_e32 v5, 0 -; GCN-IR-NEXT: s_and_saveexec_b64 s[8:9], vcc -; GCN-IR-NEXT: s_xor_b64 s[8:9], exec, s[8:9] +; GCN-IR-NEXT: s_xor_b64 s[4:5], vcc, -1 +; GCN-IR-NEXT: s_and_saveexec_b64 s[10:11], s[4:5] +; GCN-IR-NEXT: s_xor_b64 s[4:5], exec, s[10:11] ; GCN-IR-NEXT: s_cbranch_execz .LBB9_5 ; GCN-IR-NEXT: ; %bb.2: ; %udiv-preheader -; GCN-IR-NEXT: v_add_i32_e32 v12, vcc, -1, v0 -; GCN-IR-NEXT: v_addc_u32_e32 v13, vcc, -1, v1, vcc -; GCN-IR-NEXT: v_lshr_b64 v[8:9], s[4:5], v8 -; GCN-IR-NEXT: v_sub_i32_e32 v6, vcc, 47, v10 -; GCN-IR-NEXT: v_mov_b32_e32 v10, 0 -; GCN-IR-NEXT: v_subb_u32_e64 v7, s[4:5], 0, 0, vcc -; GCN-IR-NEXT: s_mov_b64 s[10:11], 0 -; GCN-IR-NEXT: v_mov_b32_e32 v11, 0 +; GCN-IR-NEXT: v_add_i32_e32 v10, vcc, -1, v0 +; GCN-IR-NEXT: v_addc_u32_e32 v11, vcc, -1, v1, vcc +; GCN-IR-NEXT: v_sub_i32_e32 v12, vcc, 47, v8 +; GCN-IR-NEXT: v_lshr_b64 v[6:7], s[8:9], v6 +; GCN-IR-NEXT: v_subb_u32_e64 v13, s[8:9], 0, 0, vcc +; GCN-IR-NEXT: v_mov_b32_e32 v8, 0 +; GCN-IR-NEXT: s_mov_b64 s[8:9], 0 +; GCN-IR-NEXT: v_mov_b32_e32 v9, 0 ; GCN-IR-NEXT: v_mov_b32_e32 v5, 0 ; GCN-IR-NEXT: .LBB9_3: ; %udiv-do-while ; GCN-IR-NEXT: ; =>This Inner Loop Header: Depth=1 -; GCN-IR-NEXT: v_lshl_b64 v[8:9], v[8:9], 1 +; GCN-IR-NEXT: v_lshl_b64 v[6:7], v[6:7], 1 ; GCN-IR-NEXT: v_lshrrev_b32_e32 v4, 31, v3 -; GCN-IR-NEXT: v_or_b32_e32 v8, v8, v4 +; GCN-IR-NEXT: v_or_b32_e32 v6, v6, v4 ; GCN-IR-NEXT: v_lshl_b64 v[2:3], v[2:3], 1 -; GCN-IR-NEXT: v_sub_i32_e32 v4, vcc, v12, v8 -; GCN-IR-NEXT: v_subb_u32_e32 v4, vcc, v13, v9, vcc -; GCN-IR-NEXT: v_or_b32_e32 v2, v10, v2 -; GCN-IR-NEXT: v_ashrrev_i32_e32 v10, 31, v4 -; GCN-IR-NEXT: v_add_i32_e32 v6, vcc, 1, v6 -; GCN-IR-NEXT: v_or_b32_e32 v3, v11, v3 -; GCN-IR-NEXT: v_and_b32_e32 v4, 1, v10 -; GCN-IR-NEXT: v_and_b32_e32 v11, v10, v1 -; GCN-IR-NEXT: v_and_b32_e32 v10, v10, v0 -; GCN-IR-NEXT: v_addc_u32_e32 v7, vcc, 0, v7, vcc -; GCN-IR-NEXT: v_cmp_eq_u64_e32 vcc, 0, v[6:7] -; GCN-IR-NEXT: v_sub_i32_e64 v8, s[4:5], v8, v10 -; GCN-IR-NEXT: v_subb_u32_e64 v9, s[4:5], v9, v11, s[4:5] -; GCN-IR-NEXT: v_mov_b32_e32 v11, v5 -; GCN-IR-NEXT: s_or_b64 s[10:11], vcc, s[10:11] -; GCN-IR-NEXT: v_mov_b32_e32 v10, v4 -; GCN-IR-NEXT: s_andn2_b64 exec, exec, s[10:11] +; GCN-IR-NEXT: v_sub_i32_e32 v4, vcc, v10, v6 +; GCN-IR-NEXT: v_subb_u32_e32 v4, vcc, v11, v7, vcc +; GCN-IR-NEXT: v_or_b32_e32 v2, v8, v2 +; GCN-IR-NEXT: v_ashrrev_i32_e32 v8, 31, v4 +; GCN-IR-NEXT: v_or_b32_e32 v3, v9, v3 +; GCN-IR-NEXT: v_and_b32_e32 v4, 1, v8 +; GCN-IR-NEXT: v_and_b32_e32 v9, v8, v1 +; GCN-IR-NEXT: v_and_b32_e32 v8, v8, v0 +; GCN-IR-NEXT: v_sub_i32_e32 v6, vcc, v6, v8 +; GCN-IR-NEXT: v_subb_u32_e32 v7, vcc, v7, v9, vcc +; GCN-IR-NEXT: v_add_i32_e32 v12, vcc, 1, v12 +; GCN-IR-NEXT: v_addc_u32_e32 v13, vcc, 0, v13, vcc +; GCN-IR-NEXT: v_mov_b32_e32 v9, v5 +; GCN-IR-NEXT: s_or_b64 s[8:9], vcc, s[8:9] +; GCN-IR-NEXT: v_mov_b32_e32 v8, v4 +; GCN-IR-NEXT: s_andn2_b64 exec, exec, s[8:9] ; GCN-IR-NEXT: s_cbranch_execnz .LBB9_3 ; GCN-IR-NEXT: ; %bb.4: ; %Flow -; GCN-IR-NEXT: s_or_b64 exec, exec, s[10:11] -; GCN-IR-NEXT: .LBB9_5: ; %Flow4 ; GCN-IR-NEXT: s_or_b64 exec, exec, s[8:9] +; GCN-IR-NEXT: .LBB9_5: ; %Flow4 +; GCN-IR-NEXT: s_or_b64 exec, exec, s[4:5] ; GCN-IR-NEXT: v_lshl_b64 v[0:1], v[2:3], 1 ; GCN-IR-NEXT: v_or_b32_e32 v2, v5, v1 ; GCN-IR-NEXT: v_or_b32_e32 v3, v4, v0 @@ -1188,13 +1198,13 @@ define i64 @v_test_udiv_pow2_k_den_i64(i64 %x) { ; GCN-IR-NEXT: v_ffbh_u32_e32 v2, v0 ; GCN-IR-NEXT: v_add_i32_e64 v2, s[4:5], 32, v2 ; GCN-IR-NEXT: v_ffbh_u32_e32 v3, v1 -; GCN-IR-NEXT: v_min_u32_e32 v10, v2, v3 -; GCN-IR-NEXT: v_sub_i32_e64 v6, s[4:5], 48, v10 -; GCN-IR-NEXT: v_subb_u32_e64 v7, s[4:5], 0, 0, s[4:5] +; GCN-IR-NEXT: v_min_u32_e32 v6, v2, v3 +; GCN-IR-NEXT: v_sub_i32_e64 v4, s[4:5], 48, v6 +; GCN-IR-NEXT: v_subb_u32_e64 v5, s[4:5], 0, 0, s[4:5] ; GCN-IR-NEXT: v_cmp_eq_u64_e32 vcc, 0, v[0:1] -; GCN-IR-NEXT: v_cmp_lt_u64_e64 s[4:5], 63, v[6:7] +; GCN-IR-NEXT: v_cmp_lt_u64_e64 s[4:5], 63, v[4:5] ; GCN-IR-NEXT: s_or_b64 s[4:5], vcc, s[4:5] -; GCN-IR-NEXT: v_cmp_ne_u64_e32 vcc, 63, v[6:7] +; GCN-IR-NEXT: v_cmp_ne_u64_e32 vcc, 63, v[4:5] ; GCN-IR-NEXT: s_xor_b64 s[6:7], s[4:5], -1 ; GCN-IR-NEXT: v_cndmask_b32_e64 v2, v1, 0, s[4:5] ; GCN-IR-NEXT: v_cndmask_b32_e64 v3, v0, 0, s[4:5] @@ -1202,52 +1212,51 @@ define i64 @v_test_udiv_pow2_k_den_i64(i64 %x) { ; GCN-IR-NEXT: s_and_saveexec_b64 s[6:7], s[4:5] ; GCN-IR-NEXT: s_cbranch_execz .LBB10_6 ; GCN-IR-NEXT: ; %bb.1: ; %udiv-bb1 -; GCN-IR-NEXT: v_add_i32_e32 v8, vcc, 1, v6 -; GCN-IR-NEXT: v_addc_u32_e32 v9, vcc, 0, v7, vcc -; GCN-IR-NEXT: v_sub_i32_e64 v2, s[4:5], 63, v6 -; GCN-IR-NEXT: v_mov_b32_e32 v4, 0 -; GCN-IR-NEXT: v_cmp_ne_u64_e32 vcc, 0, v[8:9] +; GCN-IR-NEXT: v_add_i32_e32 v7, vcc, 1, v4 +; GCN-IR-NEXT: v_addc_u32_e32 v2, vcc, 0, v5, vcc +; GCN-IR-NEXT: v_sub_i32_e64 v2, s[4:5], 63, v4 ; GCN-IR-NEXT: v_lshl_b64 v[2:3], v[0:1], v2 +; GCN-IR-NEXT: v_mov_b32_e32 v4, 0 ; GCN-IR-NEXT: v_mov_b32_e32 v5, 0 -; GCN-IR-NEXT: s_and_saveexec_b64 s[4:5], vcc -; GCN-IR-NEXT: s_xor_b64 s[8:9], exec, s[4:5] +; GCN-IR-NEXT: s_xor_b64 s[4:5], vcc, -1 +; GCN-IR-NEXT: s_and_saveexec_b64 s[8:9], s[4:5] +; GCN-IR-NEXT: s_xor_b64 s[4:5], exec, s[8:9] ; GCN-IR-NEXT: s_cbranch_execz .LBB10_5 ; GCN-IR-NEXT: ; %bb.2: ; %udiv-preheader -; GCN-IR-NEXT: v_lshr_b64 v[6:7], v[0:1], v8 -; GCN-IR-NEXT: v_add_i32_e32 v0, vcc, 0xffffffcf, v10 -; GCN-IR-NEXT: v_mov_b32_e32 v8, 0 -; GCN-IR-NEXT: v_addc_u32_e64 v1, s[4:5], 0, -1, vcc -; GCN-IR-NEXT: s_mov_b64 s[10:11], 0 -; GCN-IR-NEXT: v_mov_b32_e32 v9, 0 +; GCN-IR-NEXT: v_add_i32_e32 v8, vcc, 0xffffffcf, v6 +; GCN-IR-NEXT: v_lshr_b64 v[0:1], v[0:1], v7 +; GCN-IR-NEXT: v_addc_u32_e64 v9, s[8:9], 0, -1, vcc +; GCN-IR-NEXT: v_mov_b32_e32 v6, 0 +; GCN-IR-NEXT: s_mov_b64 s[8:9], 0 +; GCN-IR-NEXT: v_mov_b32_e32 v7, 0 ; GCN-IR-NEXT: v_mov_b32_e32 v5, 0 -; GCN-IR-NEXT: s_movk_i32 s12, 0x7fff +; GCN-IR-NEXT: s_movk_i32 s10, 0x7fff ; GCN-IR-NEXT: .LBB10_3: ; %udiv-do-while ; GCN-IR-NEXT: ; =>This Inner Loop Header: Depth=1 -; GCN-IR-NEXT: v_lshl_b64 v[6:7], v[6:7], 1 +; GCN-IR-NEXT: v_lshl_b64 v[0:1], v[0:1], 1 ; GCN-IR-NEXT: v_lshrrev_b32_e32 v4, 31, v3 -; GCN-IR-NEXT: v_or_b32_e32 v6, v6, v4 -; GCN-IR-NEXT: v_sub_i32_e32 v4, vcc, s12, v6 +; GCN-IR-NEXT: v_or_b32_e32 v0, v0, v4 ; GCN-IR-NEXT: v_lshl_b64 v[2:3], v[2:3], 1 -; GCN-IR-NEXT: v_subb_u32_e32 v4, vcc, 0, v7, vcc -; GCN-IR-NEXT: v_add_i32_e32 v0, vcc, 1, v0 -; GCN-IR-NEXT: v_or_b32_e32 v2, v8, v2 -; GCN-IR-NEXT: v_ashrrev_i32_e32 v8, 31, v4 -; GCN-IR-NEXT: v_addc_u32_e32 v1, vcc, 0, v1, vcc -; GCN-IR-NEXT: v_and_b32_e32 v4, 1, v8 -; GCN-IR-NEXT: v_and_b32_e32 v8, 0x8000, v8 -; GCN-IR-NEXT: v_cmp_eq_u64_e32 vcc, 0, v[0:1] -; GCN-IR-NEXT: v_or_b32_e32 v3, v9, v3 -; GCN-IR-NEXT: v_sub_i32_e64 v6, s[4:5], v6, v8 -; GCN-IR-NEXT: v_mov_b32_e32 v9, v5 -; GCN-IR-NEXT: v_subbrev_u32_e64 v7, s[4:5], 0, v7, s[4:5] -; GCN-IR-NEXT: s_or_b64 s[10:11], vcc, s[10:11] -; GCN-IR-NEXT: v_mov_b32_e32 v8, v4 -; GCN-IR-NEXT: s_andn2_b64 exec, exec, s[10:11] +; GCN-IR-NEXT: v_sub_i32_e32 v4, vcc, s10, v0 +; GCN-IR-NEXT: v_subb_u32_e32 v4, vcc, 0, v1, vcc +; GCN-IR-NEXT: v_or_b32_e32 v2, v6, v2 +; GCN-IR-NEXT: v_ashrrev_i32_e32 v6, 31, v4 +; GCN-IR-NEXT: v_and_b32_e32 v4, 1, v6 +; GCN-IR-NEXT: v_and_b32_e32 v6, 0x8000, v6 +; GCN-IR-NEXT: v_sub_i32_e32 v0, vcc, v0, v6 +; GCN-IR-NEXT: v_subbrev_u32_e32 v1, vcc, 0, v1, vcc +; GCN-IR-NEXT: v_add_i32_e32 v8, vcc, 1, v8 +; GCN-IR-NEXT: v_or_b32_e32 v3, v7, v3 +; GCN-IR-NEXT: v_addc_u32_e32 v9, vcc, 0, v9, vcc +; GCN-IR-NEXT: v_mov_b32_e32 v7, v5 +; GCN-IR-NEXT: s_or_b64 s[8:9], vcc, s[8:9] +; GCN-IR-NEXT: v_mov_b32_e32 v6, v4 +; GCN-IR-NEXT: s_andn2_b64 exec, exec, s[8:9] ; GCN-IR-NEXT: s_cbranch_execnz .LBB10_3 ; GCN-IR-NEXT: ; %bb.4: ; %Flow -; GCN-IR-NEXT: s_or_b64 exec, exec, s[10:11] -; GCN-IR-NEXT: .LBB10_5: ; %Flow4 ; GCN-IR-NEXT: s_or_b64 exec, exec, s[8:9] +; GCN-IR-NEXT: .LBB10_5: ; %Flow4 +; GCN-IR-NEXT: s_or_b64 exec, exec, s[4:5] ; GCN-IR-NEXT: v_lshl_b64 v[0:1], v[2:3], 1 ; GCN-IR-NEXT: v_or_b32_e32 v2, v5, v1 ; GCN-IR-NEXT: v_or_b32_e32 v3, v4, v0 @@ -1294,52 +1303,58 @@ define amdgpu_kernel void @s_test_udiv_k_den_i64(ptr addrspace(1) %out, i64 %x) ; GCN-IR: ; %bb.0: ; %_udiv-special-cases ; GCN-IR-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x9 ; GCN-IR-NEXT: s_waitcnt lgkmcnt(0) -; GCN-IR-NEXT: s_flbit_i32_b64 s12, s[2:3] -; GCN-IR-NEXT: s_sub_u32 s8, 59, s12 +; GCN-IR-NEXT: s_flbit_i32_b64 s10, s[2:3] +; GCN-IR-NEXT: s_sub_u32 s8, 59, s10 ; GCN-IR-NEXT: s_subb_u32 s9, 0, 0 ; GCN-IR-NEXT: v_cmp_eq_u64_e64 s[4:5], s[2:3], 0 ; GCN-IR-NEXT: v_cmp_gt_u64_e64 s[6:7], s[8:9], 63 -; GCN-IR-NEXT: v_cmp_eq_u64_e64 s[10:11], s[8:9], 63 +; GCN-IR-NEXT: v_cmp_eq_u64_e64 s[12:13], s[8:9], 63 ; GCN-IR-NEXT: s_or_b64 s[4:5], s[4:5], s[6:7] ; GCN-IR-NEXT: s_and_b64 s[6:7], s[4:5], exec ; GCN-IR-NEXT: s_cselect_b32 s7, 0, s3 ; GCN-IR-NEXT: s_cselect_b32 s6, 0, s2 -; GCN-IR-NEXT: s_or_b64 s[4:5], s[4:5], s[10:11] +; GCN-IR-NEXT: s_or_b64 s[4:5], s[4:5], s[12:13] ; GCN-IR-NEXT: s_andn2_b64 vcc, exec, s[4:5] ; GCN-IR-NEXT: s_mov_b64 s[4:5], 0 ; GCN-IR-NEXT: s_cbranch_vccz .LBB11_5 ; GCN-IR-NEXT: ; %bb.1: ; %udiv-bb1 -; GCN-IR-NEXT: s_add_u32 s10, s8, 1 -; GCN-IR-NEXT: s_addc_u32 s11, s9, 0 -; GCN-IR-NEXT: v_cmp_eq_u64_e64 s[6:7], s[10:11], 0 +; GCN-IR-NEXT: s_add_u32 s11, s8, 1 +; GCN-IR-NEXT: s_cselect_b64 s[6:7], -1, 0 +; GCN-IR-NEXT: s_or_b32 s6, s6, s7 +; GCN-IR-NEXT: s_cmp_lg_u32 s6, 0 +; GCN-IR-NEXT: s_addc_u32 s6, s9, 0 +; GCN-IR-NEXT: s_cselect_b64 s[6:7], -1, 0 ; GCN-IR-NEXT: s_sub_i32 s8, 63, s8 ; GCN-IR-NEXT: s_andn2_b64 vcc, exec, s[6:7] ; GCN-IR-NEXT: s_lshl_b64 s[6:7], s[2:3], s8 ; GCN-IR-NEXT: s_cbranch_vccz .LBB11_4 ; GCN-IR-NEXT: ; %bb.2: ; %udiv-preheader -; GCN-IR-NEXT: s_lshr_b64 s[8:9], s[2:3], s10 -; GCN-IR-NEXT: s_add_u32 s2, s12, 0xffffffc4 -; GCN-IR-NEXT: s_addc_u32 s3, 0, -1 -; GCN-IR-NEXT: s_mov_b64 s[10:11], 0 +; GCN-IR-NEXT: s_lshr_b64 s[2:3], s[2:3], s11 +; GCN-IR-NEXT: s_add_u32 s10, s10, 0xffffffc4 +; GCN-IR-NEXT: s_addc_u32 s11, 0, -1 +; GCN-IR-NEXT: s_mov_b64 s[8:9], 0 ; GCN-IR-NEXT: s_mov_b32 s5, 0 ; GCN-IR-NEXT: .LBB11_3: ; %udiv-do-while ; GCN-IR-NEXT: ; =>This Inner Loop Header: Depth=1 -; GCN-IR-NEXT: s_lshl_b64 s[8:9], s[8:9], 1 +; GCN-IR-NEXT: s_lshl_b64 s[2:3], s[2:3], 1 ; GCN-IR-NEXT: s_lshr_b32 s4, s7, 31 ; GCN-IR-NEXT: s_lshl_b64 s[6:7], s[6:7], 1 -; GCN-IR-NEXT: s_or_b64 s[8:9], s[8:9], s[4:5] -; GCN-IR-NEXT: s_or_b64 s[6:7], s[10:11], s[6:7] -; GCN-IR-NEXT: s_sub_u32 s4, 23, s8 -; GCN-IR-NEXT: s_subb_u32 s4, 0, s9 -; GCN-IR-NEXT: s_ashr_i32 s10, s4, 31 -; GCN-IR-NEXT: s_and_b32 s4, s10, 1 -; GCN-IR-NEXT: s_and_b32 s10, s10, 24 -; GCN-IR-NEXT: s_sub_u32 s8, s8, s10 -; GCN-IR-NEXT: s_subb_u32 s9, s9, 0 -; GCN-IR-NEXT: s_add_u32 s2, s2, 1 -; GCN-IR-NEXT: s_addc_u32 s3, s3, 0 -; GCN-IR-NEXT: v_cmp_eq_u64_e64 s[12:13], s[2:3], 0 -; GCN-IR-NEXT: s_mov_b64 s[10:11], s[4:5] +; GCN-IR-NEXT: s_or_b64 s[2:3], s[2:3], s[4:5] +; GCN-IR-NEXT: s_or_b64 s[6:7], s[8:9], s[6:7] +; GCN-IR-NEXT: s_sub_u32 s4, 23, s2 +; GCN-IR-NEXT: s_subb_u32 s4, 0, s3 +; GCN-IR-NEXT: s_ashr_i32 s8, s4, 31 +; GCN-IR-NEXT: s_and_b32 s4, s8, 1 +; GCN-IR-NEXT: s_and_b32 s8, s8, 24 +; GCN-IR-NEXT: s_sub_u32 s2, s2, s8 +; GCN-IR-NEXT: s_subb_u32 s3, s3, 0 +; GCN-IR-NEXT: s_add_u32 s10, s10, 1 +; GCN-IR-NEXT: s_cselect_b64 s[12:13], -1, 0 +; GCN-IR-NEXT: s_or_b32 s12, s12, s13 +; GCN-IR-NEXT: s_cmp_lg_u32 s12, 0 +; GCN-IR-NEXT: s_addc_u32 s11, s11, 0 +; GCN-IR-NEXT: s_cselect_b64 s[12:13], -1, 0 +; GCN-IR-NEXT: s_mov_b64 s[8:9], s[4:5] ; GCN-IR-NEXT: s_and_b64 vcc, exec, s[12:13] ; GCN-IR-NEXT: s_cbranch_vccz .LBB11_3 ; GCN-IR-NEXT: .LBB11_4: ; %Flow6 @@ -1388,13 +1403,13 @@ define i64 @v_test_udiv_k_den_i64(i64 %x) { ; GCN-IR-NEXT: v_ffbh_u32_e32 v2, v0 ; GCN-IR-NEXT: v_add_i32_e64 v2, s[4:5], 32, v2 ; GCN-IR-NEXT: v_ffbh_u32_e32 v3, v1 -; GCN-IR-NEXT: v_min_u32_e32 v10, v2, v3 -; GCN-IR-NEXT: v_sub_i32_e64 v6, s[4:5], 59, v10 -; GCN-IR-NEXT: v_subb_u32_e64 v7, s[4:5], 0, 0, s[4:5] +; GCN-IR-NEXT: v_min_u32_e32 v6, v2, v3 +; GCN-IR-NEXT: v_sub_i32_e64 v4, s[4:5], 59, v6 +; GCN-IR-NEXT: v_subb_u32_e64 v5, s[4:5], 0, 0, s[4:5] ; GCN-IR-NEXT: v_cmp_eq_u64_e32 vcc, 0, v[0:1] -; GCN-IR-NEXT: v_cmp_lt_u64_e64 s[4:5], 63, v[6:7] +; GCN-IR-NEXT: v_cmp_lt_u64_e64 s[4:5], 63, v[4:5] ; GCN-IR-NEXT: s_or_b64 s[4:5], vcc, s[4:5] -; GCN-IR-NEXT: v_cmp_ne_u64_e32 vcc, 63, v[6:7] +; GCN-IR-NEXT: v_cmp_ne_u64_e32 vcc, 63, v[4:5] ; GCN-IR-NEXT: s_xor_b64 s[6:7], s[4:5], -1 ; GCN-IR-NEXT: v_cndmask_b32_e64 v2, v1, 0, s[4:5] ; GCN-IR-NEXT: v_cndmask_b32_e64 v3, v0, 0, s[4:5] @@ -1402,51 +1417,50 @@ define i64 @v_test_udiv_k_den_i64(i64 %x) { ; GCN-IR-NEXT: s_and_saveexec_b64 s[6:7], s[4:5] ; GCN-IR-NEXT: s_cbranch_execz .LBB12_6 ; GCN-IR-NEXT: ; %bb.1: ; %udiv-bb1 -; GCN-IR-NEXT: v_add_i32_e32 v8, vcc, 1, v6 -; GCN-IR-NEXT: v_addc_u32_e32 v9, vcc, 0, v7, vcc -; GCN-IR-NEXT: v_sub_i32_e64 v2, s[4:5], 63, v6 -; GCN-IR-NEXT: v_mov_b32_e32 v4, 0 -; GCN-IR-NEXT: v_cmp_ne_u64_e32 vcc, 0, v[8:9] +; GCN-IR-NEXT: v_add_i32_e32 v7, vcc, 1, v4 +; GCN-IR-NEXT: v_addc_u32_e32 v2, vcc, 0, v5, vcc +; GCN-IR-NEXT: v_sub_i32_e64 v2, s[4:5], 63, v4 ; GCN-IR-NEXT: v_lshl_b64 v[2:3], v[0:1], v2 +; GCN-IR-NEXT: v_mov_b32_e32 v4, 0 ; GCN-IR-NEXT: v_mov_b32_e32 v5, 0 -; GCN-IR-NEXT: s_and_saveexec_b64 s[4:5], vcc -; GCN-IR-NEXT: s_xor_b64 s[8:9], exec, s[4:5] +; GCN-IR-NEXT: s_xor_b64 s[4:5], vcc, -1 +; GCN-IR-NEXT: s_and_saveexec_b64 s[8:9], s[4:5] +; GCN-IR-NEXT: s_xor_b64 s[4:5], exec, s[8:9] ; GCN-IR-NEXT: s_cbranch_execz .LBB12_5 ; GCN-IR-NEXT: ; %bb.2: ; %udiv-preheader -; GCN-IR-NEXT: v_lshr_b64 v[6:7], v[0:1], v8 -; GCN-IR-NEXT: v_add_i32_e32 v0, vcc, 0xffffffc4, v10 -; GCN-IR-NEXT: v_mov_b32_e32 v8, 0 -; GCN-IR-NEXT: v_addc_u32_e64 v1, s[4:5], 0, -1, vcc -; GCN-IR-NEXT: s_mov_b64 s[10:11], 0 -; GCN-IR-NEXT: v_mov_b32_e32 v9, 0 +; GCN-IR-NEXT: v_add_i32_e32 v8, vcc, 0xffffffc4, v6 +; GCN-IR-NEXT: v_lshr_b64 v[0:1], v[0:1], v7 +; GCN-IR-NEXT: v_addc_u32_e64 v9, s[8:9], 0, -1, vcc +; GCN-IR-NEXT: v_mov_b32_e32 v6, 0 +; GCN-IR-NEXT: s_mov_b64 s[8:9], 0 +; GCN-IR-NEXT: v_mov_b32_e32 v7, 0 ; GCN-IR-NEXT: v_mov_b32_e32 v5, 0 ; GCN-IR-NEXT: .LBB12_3: ; %udiv-do-while ; GCN-IR-NEXT: ; =>This Inner Loop Header: Depth=1 -; GCN-IR-NEXT: v_lshl_b64 v[6:7], v[6:7], 1 +; GCN-IR-NEXT: v_lshl_b64 v[0:1], v[0:1], 1 ; GCN-IR-NEXT: v_lshrrev_b32_e32 v4, 31, v3 -; GCN-IR-NEXT: v_or_b32_e32 v6, v6, v4 -; GCN-IR-NEXT: v_sub_i32_e32 v4, vcc, 23, v6 +; GCN-IR-NEXT: v_or_b32_e32 v0, v0, v4 ; GCN-IR-NEXT: v_lshl_b64 v[2:3], v[2:3], 1 -; GCN-IR-NEXT: v_subb_u32_e32 v4, vcc, 0, v7, vcc -; GCN-IR-NEXT: v_add_i32_e32 v0, vcc, 1, v0 -; GCN-IR-NEXT: v_or_b32_e32 v2, v8, v2 -; GCN-IR-NEXT: v_ashrrev_i32_e32 v8, 31, v4 -; GCN-IR-NEXT: v_addc_u32_e32 v1, vcc, 0, v1, vcc -; GCN-IR-NEXT: v_and_b32_e32 v4, 1, v8 -; GCN-IR-NEXT: v_and_b32_e32 v8, 24, v8 -; GCN-IR-NEXT: v_cmp_eq_u64_e32 vcc, 0, v[0:1] -; GCN-IR-NEXT: v_or_b32_e32 v3, v9, v3 -; GCN-IR-NEXT: v_sub_i32_e64 v6, s[4:5], v6, v8 -; GCN-IR-NEXT: v_mov_b32_e32 v9, v5 -; GCN-IR-NEXT: v_subbrev_u32_e64 v7, s[4:5], 0, v7, s[4:5] -; GCN-IR-NEXT: s_or_b64 s[10:11], vcc, s[10:11] -; GCN-IR-NEXT: v_mov_b32_e32 v8, v4 -; GCN-IR-NEXT: s_andn2_b64 exec, exec, s[10:11] +; GCN-IR-NEXT: v_sub_i32_e32 v4, vcc, 23, v0 +; GCN-IR-NEXT: v_subb_u32_e32 v4, vcc, 0, v1, vcc +; GCN-IR-NEXT: v_or_b32_e32 v2, v6, v2 +; GCN-IR-NEXT: v_ashrrev_i32_e32 v6, 31, v4 +; GCN-IR-NEXT: v_and_b32_e32 v4, 1, v6 +; GCN-IR-NEXT: v_and_b32_e32 v6, 24, v6 +; GCN-IR-NEXT: v_sub_i32_e32 v0, vcc, v0, v6 +; GCN-IR-NEXT: v_subbrev_u32_e32 v1, vcc, 0, v1, vcc +; GCN-IR-NEXT: v_add_i32_e32 v8, vcc, 1, v8 +; GCN-IR-NEXT: v_or_b32_e32 v3, v7, v3 +; GCN-IR-NEXT: v_addc_u32_e32 v9, vcc, 0, v9, vcc +; GCN-IR-NEXT: v_mov_b32_e32 v7, v5 +; GCN-IR-NEXT: s_or_b64 s[8:9], vcc, s[8:9] +; GCN-IR-NEXT: v_mov_b32_e32 v6, v4 +; GCN-IR-NEXT: s_andn2_b64 exec, exec, s[8:9] ; GCN-IR-NEXT: s_cbranch_execnz .LBB12_3 ; GCN-IR-NEXT: ; %bb.4: ; %Flow -; GCN-IR-NEXT: s_or_b64 exec, exec, s[10:11] -; GCN-IR-NEXT: .LBB12_5: ; %Flow4 ; GCN-IR-NEXT: s_or_b64 exec, exec, s[8:9] +; GCN-IR-NEXT: .LBB12_5: ; %Flow4 +; GCN-IR-NEXT: s_or_b64 exec, exec, s[4:5] ; GCN-IR-NEXT: v_lshl_b64 v[0:1], v[2:3], 1 ; GCN-IR-NEXT: v_or_b32_e32 v2, v5, v1 ; GCN-IR-NEXT: v_or_b32_e32 v3, v4, v0 diff --git a/llvm/test/CodeGen/AMDGPU/urem64.ll b/llvm/test/CodeGen/AMDGPU/urem64.ll index dc25caadb99a9..a0dc07745ccb2 100644 --- a/llvm/test/CodeGen/AMDGPU/urem64.ll +++ b/llvm/test/CodeGen/AMDGPU/urem64.ll @@ -170,35 +170,38 @@ define amdgpu_kernel void @s_test_urem_i64(ptr addrspace(1) %out, i64 %x, i64 %y ; GCN-IR-NEXT: v_cmp_eq_u64_e64 s[8:9], s[6:7], 0 ; GCN-IR-NEXT: v_cmp_eq_u64_e64 s[12:13], s[2:3], 0 ; GCN-IR-NEXT: s_flbit_i32_b64 s10, s[6:7] -; GCN-IR-NEXT: s_flbit_i32_b64 s18, s[2:3] +; GCN-IR-NEXT: s_flbit_i32_b64 s16, s[2:3] ; GCN-IR-NEXT: s_or_b64 s[8:9], s[8:9], s[12:13] -; GCN-IR-NEXT: s_sub_u32 s12, s10, s18 +; GCN-IR-NEXT: s_sub_u32 s12, s10, s16 ; GCN-IR-NEXT: s_subb_u32 s13, 0, 0 ; GCN-IR-NEXT: v_cmp_gt_u64_e64 s[14:15], s[12:13], 63 -; GCN-IR-NEXT: v_cmp_eq_u64_e64 s[16:17], s[12:13], 63 +; GCN-IR-NEXT: v_cmp_eq_u64_e64 s[18:19], s[12:13], 63 ; GCN-IR-NEXT: s_or_b64 s[14:15], s[8:9], s[14:15] ; GCN-IR-NEXT: s_and_b64 s[8:9], s[14:15], exec ; GCN-IR-NEXT: s_cselect_b32 s9, 0, s3 ; GCN-IR-NEXT: s_cselect_b32 s8, 0, s2 -; GCN-IR-NEXT: s_or_b64 s[14:15], s[14:15], s[16:17] +; GCN-IR-NEXT: s_or_b64 s[14:15], s[14:15], s[18:19] ; GCN-IR-NEXT: s_andn2_b64 vcc, exec, s[14:15] ; GCN-IR-NEXT: s_cbranch_vccz .LBB0_5 ; GCN-IR-NEXT: ; %bb.1: ; %udiv-bb1 ; GCN-IR-NEXT: s_add_u32 s14, s12, 1 -; GCN-IR-NEXT: s_addc_u32 s15, s13, 0 -; GCN-IR-NEXT: v_cmp_eq_u64_e64 s[8:9], s[14:15], 0 +; GCN-IR-NEXT: s_cselect_b64 s[8:9], -1, 0 +; GCN-IR-NEXT: s_or_b32 s8, s8, s9 +; GCN-IR-NEXT: s_cmp_lg_u32 s8, 0 +; GCN-IR-NEXT: s_addc_u32 s8, s13, 0 +; GCN-IR-NEXT: s_cselect_b64 s[8:9], -1, 0 ; GCN-IR-NEXT: s_sub_i32 s12, 63, s12 ; GCN-IR-NEXT: s_andn2_b64 vcc, exec, s[8:9] ; GCN-IR-NEXT: s_lshl_b64 s[8:9], s[2:3], s12 ; GCN-IR-NEXT: s_cbranch_vccz .LBB0_4 ; GCN-IR-NEXT: ; %bb.2: ; %udiv-preheader ; GCN-IR-NEXT: s_lshr_b64 s[12:13], s[2:3], s14 -; GCN-IR-NEXT: s_add_u32 s16, s6, -1 -; GCN-IR-NEXT: s_addc_u32 s17, s7, -1 +; GCN-IR-NEXT: s_add_u32 s14, s6, -1 +; GCN-IR-NEXT: s_addc_u32 s15, s7, -1 ; GCN-IR-NEXT: s_not_b64 s[4:5], s[10:11] -; GCN-IR-NEXT: s_add_u32 s10, s4, s18 -; GCN-IR-NEXT: s_addc_u32 s11, s5, 0 -; GCN-IR-NEXT: s_mov_b64 s[14:15], 0 +; GCN-IR-NEXT: s_add_u32 s16, s4, s16 +; GCN-IR-NEXT: s_addc_u32 s17, s5, 0 +; GCN-IR-NEXT: s_mov_b64 s[10:11], 0 ; GCN-IR-NEXT: s_mov_b32 s5, 0 ; GCN-IR-NEXT: .LBB0_3: ; %udiv-do-while ; GCN-IR-NEXT: ; =>This Inner Loop Header: Depth=1 @@ -206,19 +209,22 @@ define amdgpu_kernel void @s_test_urem_i64(ptr addrspace(1) %out, i64 %x, i64 %y ; GCN-IR-NEXT: s_lshr_b32 s4, s9, 31 ; GCN-IR-NEXT: s_lshl_b64 s[8:9], s[8:9], 1 ; GCN-IR-NEXT: s_or_b64 s[12:13], s[12:13], s[4:5] -; GCN-IR-NEXT: s_or_b64 s[8:9], s[14:15], s[8:9] -; GCN-IR-NEXT: s_sub_u32 s4, s16, s12 -; GCN-IR-NEXT: s_subb_u32 s4, s17, s13 -; GCN-IR-NEXT: s_ashr_i32 s14, s4, 31 -; GCN-IR-NEXT: s_mov_b32 s15, s14 -; GCN-IR-NEXT: s_and_b32 s4, s14, 1 -; GCN-IR-NEXT: s_and_b64 s[14:15], s[14:15], s[6:7] -; GCN-IR-NEXT: s_sub_u32 s12, s12, s14 -; GCN-IR-NEXT: s_subb_u32 s13, s13, s15 -; GCN-IR-NEXT: s_add_u32 s10, s10, 1 -; GCN-IR-NEXT: s_addc_u32 s11, s11, 0 -; GCN-IR-NEXT: v_cmp_eq_u64_e64 s[18:19], s[10:11], 0 -; GCN-IR-NEXT: s_mov_b64 s[14:15], s[4:5] +; GCN-IR-NEXT: s_or_b64 s[8:9], s[10:11], s[8:9] +; GCN-IR-NEXT: s_sub_u32 s4, s14, s12 +; GCN-IR-NEXT: s_subb_u32 s4, s15, s13 +; GCN-IR-NEXT: s_ashr_i32 s10, s4, 31 +; GCN-IR-NEXT: s_mov_b32 s11, s10 +; GCN-IR-NEXT: s_and_b32 s4, s10, 1 +; GCN-IR-NEXT: s_and_b64 s[18:19], s[10:11], s[6:7] +; GCN-IR-NEXT: s_sub_u32 s12, s12, s18 +; GCN-IR-NEXT: s_subb_u32 s13, s13, s19 +; GCN-IR-NEXT: s_add_u32 s16, s16, 1 +; GCN-IR-NEXT: s_cselect_b64 s[18:19], -1, 0 +; GCN-IR-NEXT: s_or_b32 s18, s18, s19 +; GCN-IR-NEXT: s_cmp_lg_u32 s18, 0 +; GCN-IR-NEXT: s_addc_u32 s17, s17, 0 +; GCN-IR-NEXT: s_cselect_b64 s[18:19], -1, 0 +; GCN-IR-NEXT: s_mov_b64 s[10:11], s[4:5] ; GCN-IR-NEXT: s_and_b64 vcc, exec, s[18:19] ; GCN-IR-NEXT: s_cbranch_vccz .LBB0_3 ; GCN-IR-NEXT: .LBB0_4: ; %Flow7 @@ -362,12 +368,12 @@ define i64 @v_test_urem_i64(i64 %x, i64 %y) { ; GCN-IR-NEXT: v_ffbh_u32_e32 v4, v2 ; GCN-IR-NEXT: v_add_i32_e64 v4, s[6:7], 32, v4 ; GCN-IR-NEXT: v_ffbh_u32_e32 v5, v3 -; GCN-IR-NEXT: v_min_u32_e32 v12, v4, v5 +; GCN-IR-NEXT: v_min_u32_e32 v10, v4, v5 ; GCN-IR-NEXT: v_ffbh_u32_e32 v4, v0 ; GCN-IR-NEXT: v_add_i32_e64 v4, s[6:7], 32, v4 ; GCN-IR-NEXT: v_ffbh_u32_e32 v5, v1 -; GCN-IR-NEXT: v_min_u32_e32 v13, v4, v5 -; GCN-IR-NEXT: v_sub_i32_e64 v4, s[6:7], v12, v13 +; GCN-IR-NEXT: v_min_u32_e32 v11, v4, v5 +; GCN-IR-NEXT: v_sub_i32_e64 v4, s[6:7], v10, v11 ; GCN-IR-NEXT: v_cmp_eq_u64_e32 vcc, 0, v[2:3] ; GCN-IR-NEXT: v_cmp_eq_u64_e64 s[4:5], 0, v[0:1] ; GCN-IR-NEXT: v_subb_u32_e64 v5, s[6:7], 0, 0, s[6:7] @@ -383,54 +389,53 @@ define i64 @v_test_urem_i64(i64 %x, i64 %y) { ; GCN-IR-NEXT: s_cbranch_execz .LBB1_6 ; GCN-IR-NEXT: ; %bb.1: ; %udiv-bb1 ; GCN-IR-NEXT: v_add_i32_e32 v8, vcc, 1, v4 -; GCN-IR-NEXT: v_addc_u32_e32 v9, vcc, 0, v5, vcc +; GCN-IR-NEXT: v_addc_u32_e32 v5, vcc, 0, v5, vcc ; GCN-IR-NEXT: v_sub_i32_e64 v4, s[4:5], 63, v4 -; GCN-IR-NEXT: v_mov_b32_e32 v6, 0 -; GCN-IR-NEXT: v_cmp_ne_u64_e32 vcc, 0, v[8:9] ; GCN-IR-NEXT: v_lshl_b64 v[4:5], v[0:1], v4 +; GCN-IR-NEXT: v_mov_b32_e32 v6, 0 ; GCN-IR-NEXT: v_mov_b32_e32 v7, 0 -; GCN-IR-NEXT: s_and_saveexec_b64 s[4:5], vcc -; GCN-IR-NEXT: s_xor_b64 s[8:9], exec, s[4:5] +; GCN-IR-NEXT: s_xor_b64 s[4:5], vcc, -1 +; GCN-IR-NEXT: s_and_saveexec_b64 s[8:9], s[4:5] +; GCN-IR-NEXT: s_xor_b64 s[4:5], exec, s[8:9] ; GCN-IR-NEXT: s_cbranch_execz .LBB1_5 ; GCN-IR-NEXT: ; %bb.2: ; %udiv-preheader -; GCN-IR-NEXT: v_add_i32_e32 v14, vcc, -1, v2 -; GCN-IR-NEXT: v_addc_u32_e32 v15, vcc, -1, v3, vcc -; GCN-IR-NEXT: v_not_b32_e32 v6, v12 -; GCN-IR-NEXT: v_lshr_b64 v[10:11], v[0:1], v8 -; GCN-IR-NEXT: v_add_i32_e32 v8, vcc, v6, v13 -; GCN-IR-NEXT: v_mov_b32_e32 v12, 0 -; GCN-IR-NEXT: v_addc_u32_e64 v9, s[4:5], -1, 0, vcc -; GCN-IR-NEXT: s_mov_b64 s[10:11], 0 -; GCN-IR-NEXT: v_mov_b32_e32 v13, 0 +; GCN-IR-NEXT: v_add_i32_e32 v12, vcc, -1, v2 +; GCN-IR-NEXT: v_addc_u32_e32 v13, vcc, -1, v3, vcc +; GCN-IR-NEXT: v_not_b32_e32 v6, v10 +; GCN-IR-NEXT: v_add_i32_e32 v14, vcc, v6, v11 +; GCN-IR-NEXT: v_lshr_b64 v[8:9], v[0:1], v8 +; GCN-IR-NEXT: v_addc_u32_e64 v15, s[8:9], -1, 0, vcc +; GCN-IR-NEXT: v_mov_b32_e32 v10, 0 +; GCN-IR-NEXT: s_mov_b64 s[8:9], 0 +; GCN-IR-NEXT: v_mov_b32_e32 v11, 0 ; GCN-IR-NEXT: v_mov_b32_e32 v7, 0 ; GCN-IR-NEXT: .LBB1_3: ; %udiv-do-while ; GCN-IR-NEXT: ; =>This Inner Loop Header: Depth=1 -; GCN-IR-NEXT: v_lshl_b64 v[10:11], v[10:11], 1 +; GCN-IR-NEXT: v_lshl_b64 v[8:9], v[8:9], 1 ; GCN-IR-NEXT: v_lshrrev_b32_e32 v6, 31, v5 -; GCN-IR-NEXT: v_or_b32_e32 v10, v10, v6 +; GCN-IR-NEXT: v_or_b32_e32 v8, v8, v6 ; GCN-IR-NEXT: v_lshl_b64 v[4:5], v[4:5], 1 -; GCN-IR-NEXT: v_sub_i32_e32 v6, vcc, v14, v10 -; GCN-IR-NEXT: v_subb_u32_e32 v6, vcc, v15, v11, vcc -; GCN-IR-NEXT: v_or_b32_e32 v4, v12, v4 -; GCN-IR-NEXT: v_ashrrev_i32_e32 v12, 31, v6 -; GCN-IR-NEXT: v_add_i32_e32 v8, vcc, 1, v8 -; GCN-IR-NEXT: v_or_b32_e32 v5, v13, v5 -; GCN-IR-NEXT: v_and_b32_e32 v6, 1, v12 -; GCN-IR-NEXT: v_and_b32_e32 v13, v12, v3 -; GCN-IR-NEXT: v_and_b32_e32 v12, v12, v2 -; GCN-IR-NEXT: v_addc_u32_e32 v9, vcc, 0, v9, vcc -; GCN-IR-NEXT: v_cmp_eq_u64_e32 vcc, 0, v[8:9] -; GCN-IR-NEXT: v_sub_i32_e64 v10, s[4:5], v10, v12 -; GCN-IR-NEXT: v_subb_u32_e64 v11, s[4:5], v11, v13, s[4:5] -; GCN-IR-NEXT: v_mov_b32_e32 v13, v7 -; GCN-IR-NEXT: s_or_b64 s[10:11], vcc, s[10:11] -; GCN-IR-NEXT: v_mov_b32_e32 v12, v6 -; GCN-IR-NEXT: s_andn2_b64 exec, exec, s[10:11] +; GCN-IR-NEXT: v_sub_i32_e32 v6, vcc, v12, v8 +; GCN-IR-NEXT: v_subb_u32_e32 v6, vcc, v13, v9, vcc +; GCN-IR-NEXT: v_or_b32_e32 v4, v10, v4 +; GCN-IR-NEXT: v_ashrrev_i32_e32 v10, 31, v6 +; GCN-IR-NEXT: v_or_b32_e32 v5, v11, v5 +; GCN-IR-NEXT: v_and_b32_e32 v6, 1, v10 +; GCN-IR-NEXT: v_and_b32_e32 v11, v10, v3 +; GCN-IR-NEXT: v_and_b32_e32 v10, v10, v2 +; GCN-IR-NEXT: v_sub_i32_e32 v8, vcc, v8, v10 +; GCN-IR-NEXT: v_subb_u32_e32 v9, vcc, v9, v11, vcc +; GCN-IR-NEXT: v_add_i32_e32 v14, vcc, 1, v14 +; GCN-IR-NEXT: v_addc_u32_e32 v15, vcc, 0, v15, vcc +; GCN-IR-NEXT: v_mov_b32_e32 v11, v7 +; GCN-IR-NEXT: s_or_b64 s[8:9], vcc, s[8:9] +; GCN-IR-NEXT: v_mov_b32_e32 v10, v6 +; GCN-IR-NEXT: s_andn2_b64 exec, exec, s[8:9] ; GCN-IR-NEXT: s_cbranch_execnz .LBB1_3 ; GCN-IR-NEXT: ; %bb.4: ; %Flow -; GCN-IR-NEXT: s_or_b64 exec, exec, s[10:11] -; GCN-IR-NEXT: .LBB1_5: ; %Flow4 ; GCN-IR-NEXT: s_or_b64 exec, exec, s[8:9] +; GCN-IR-NEXT: .LBB1_5: ; %Flow4 +; GCN-IR-NEXT: s_or_b64 exec, exec, s[4:5] ; GCN-IR-NEXT: v_lshl_b64 v[4:5], v[4:5], 1 ; GCN-IR-NEXT: v_or_b32_e32 v7, v7, v5 ; GCN-IR-NEXT: v_or_b32_e32 v6, v6, v4 @@ -948,34 +953,37 @@ define amdgpu_kernel void @s_test_urem_k_num_i64(ptr addrspace(1) %out, i64 %x) ; GCN-IR-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x9 ; GCN-IR-NEXT: s_mov_b64 s[4:5], 0 ; GCN-IR-NEXT: s_waitcnt lgkmcnt(0) -; GCN-IR-NEXT: s_flbit_i32_b64 s12, s[2:3] -; GCN-IR-NEXT: s_add_u32 s8, s12, 0xffffffc5 +; GCN-IR-NEXT: s_flbit_i32_b64 s14, s[2:3] +; GCN-IR-NEXT: s_add_u32 s8, s14, 0xffffffc5 ; GCN-IR-NEXT: s_addc_u32 s9, 0, -1 ; GCN-IR-NEXT: v_cmp_eq_u64_e64 s[6:7], s[2:3], 0 ; GCN-IR-NEXT: v_cmp_gt_u64_e64 s[10:11], s[8:9], 63 -; GCN-IR-NEXT: v_cmp_eq_u64_e64 s[14:15], s[8:9], 63 +; GCN-IR-NEXT: v_cmp_eq_u64_e64 s[12:13], s[8:9], 63 ; GCN-IR-NEXT: s_or_b64 s[10:11], s[6:7], s[10:11] ; GCN-IR-NEXT: s_and_b64 s[6:7], s[10:11], exec ; GCN-IR-NEXT: s_cselect_b32 s6, 0, 24 -; GCN-IR-NEXT: s_or_b64 s[10:11], s[10:11], s[14:15] +; GCN-IR-NEXT: s_or_b64 s[10:11], s[10:11], s[12:13] ; GCN-IR-NEXT: s_andn2_b64 vcc, exec, s[10:11] ; GCN-IR-NEXT: s_mov_b32 s7, 0 ; GCN-IR-NEXT: s_cbranch_vccz .LBB6_5 ; GCN-IR-NEXT: ; %bb.1: ; %udiv-bb1 ; GCN-IR-NEXT: s_add_u32 s10, s8, 1 -; GCN-IR-NEXT: s_addc_u32 s11, s9, 0 -; GCN-IR-NEXT: v_cmp_eq_u64_e64 s[6:7], s[10:11], 0 +; GCN-IR-NEXT: s_cselect_b64 s[6:7], -1, 0 +; GCN-IR-NEXT: s_or_b32 s6, s6, s7 +; GCN-IR-NEXT: s_cmp_lg_u32 s6, 0 +; GCN-IR-NEXT: s_addc_u32 s6, s9, 0 +; GCN-IR-NEXT: s_cselect_b64 s[6:7], -1, 0 ; GCN-IR-NEXT: s_sub_i32 s8, 63, s8 ; GCN-IR-NEXT: s_andn2_b64 vcc, exec, s[6:7] ; GCN-IR-NEXT: s_lshl_b64 s[6:7], 24, s8 ; GCN-IR-NEXT: s_cbranch_vccz .LBB6_4 ; GCN-IR-NEXT: ; %bb.2: ; %udiv-preheader ; GCN-IR-NEXT: s_lshr_b64 s[10:11], 24, s10 -; GCN-IR-NEXT: s_add_u32 s14, s2, -1 -; GCN-IR-NEXT: s_addc_u32 s15, s3, -1 -; GCN-IR-NEXT: s_sub_u32 s8, 58, s12 -; GCN-IR-NEXT: s_subb_u32 s9, 0, 0 -; GCN-IR-NEXT: s_mov_b64 s[12:13], 0 +; GCN-IR-NEXT: s_add_u32 s12, s2, -1 +; GCN-IR-NEXT: s_addc_u32 s13, s3, -1 +; GCN-IR-NEXT: s_sub_u32 s14, 58, s14 +; GCN-IR-NEXT: s_subb_u32 s15, 0, 0 +; GCN-IR-NEXT: s_mov_b64 s[8:9], 0 ; GCN-IR-NEXT: s_mov_b32 s5, 0 ; GCN-IR-NEXT: .LBB6_3: ; %udiv-do-while ; GCN-IR-NEXT: ; =>This Inner Loop Header: Depth=1 @@ -983,19 +991,22 @@ define amdgpu_kernel void @s_test_urem_k_num_i64(ptr addrspace(1) %out, i64 %x) ; GCN-IR-NEXT: s_lshr_b32 s4, s7, 31 ; GCN-IR-NEXT: s_lshl_b64 s[6:7], s[6:7], 1 ; GCN-IR-NEXT: s_or_b64 s[10:11], s[10:11], s[4:5] -; GCN-IR-NEXT: s_or_b64 s[6:7], s[12:13], s[6:7] -; GCN-IR-NEXT: s_sub_u32 s4, s14, s10 -; GCN-IR-NEXT: s_subb_u32 s4, s15, s11 -; GCN-IR-NEXT: s_ashr_i32 s12, s4, 31 -; GCN-IR-NEXT: s_mov_b32 s13, s12 -; GCN-IR-NEXT: s_and_b32 s4, s12, 1 -; GCN-IR-NEXT: s_and_b64 s[12:13], s[12:13], s[2:3] -; GCN-IR-NEXT: s_sub_u32 s10, s10, s12 -; GCN-IR-NEXT: s_subb_u32 s11, s11, s13 -; GCN-IR-NEXT: s_add_u32 s8, s8, 1 -; GCN-IR-NEXT: s_addc_u32 s9, s9, 0 -; GCN-IR-NEXT: v_cmp_eq_u64_e64 s[16:17], s[8:9], 0 -; GCN-IR-NEXT: s_mov_b64 s[12:13], s[4:5] +; GCN-IR-NEXT: s_or_b64 s[6:7], s[8:9], s[6:7] +; GCN-IR-NEXT: s_sub_u32 s4, s12, s10 +; GCN-IR-NEXT: s_subb_u32 s4, s13, s11 +; GCN-IR-NEXT: s_ashr_i32 s8, s4, 31 +; GCN-IR-NEXT: s_mov_b32 s9, s8 +; GCN-IR-NEXT: s_and_b32 s4, s8, 1 +; GCN-IR-NEXT: s_and_b64 s[16:17], s[8:9], s[2:3] +; GCN-IR-NEXT: s_sub_u32 s10, s10, s16 +; GCN-IR-NEXT: s_subb_u32 s11, s11, s17 +; GCN-IR-NEXT: s_add_u32 s14, s14, 1 +; GCN-IR-NEXT: s_cselect_b64 s[16:17], -1, 0 +; GCN-IR-NEXT: s_or_b32 s16, s16, s17 +; GCN-IR-NEXT: s_cmp_lg_u32 s16, 0 +; GCN-IR-NEXT: s_addc_u32 s15, s15, 0 +; GCN-IR-NEXT: s_cselect_b64 s[16:17], -1, 0 +; GCN-IR-NEXT: s_mov_b64 s[8:9], s[4:5] ; GCN-IR-NEXT: s_and_b64 vcc, exec, s[16:17] ; GCN-IR-NEXT: s_cbranch_vccz .LBB6_3 ; GCN-IR-NEXT: .LBB6_4: ; %Flow6 @@ -1064,52 +1075,58 @@ define amdgpu_kernel void @s_test_urem_k_den_i64(ptr addrspace(1) %out, i64 %x) ; GCN-IR: ; %bb.0: ; %_udiv-special-cases ; GCN-IR-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x9 ; GCN-IR-NEXT: s_waitcnt lgkmcnt(0) -; GCN-IR-NEXT: s_flbit_i32_b64 s12, s[2:3] -; GCN-IR-NEXT: s_sub_u32 s8, 59, s12 +; GCN-IR-NEXT: s_flbit_i32_b64 s10, s[2:3] +; GCN-IR-NEXT: s_sub_u32 s8, 59, s10 ; GCN-IR-NEXT: s_subb_u32 s9, 0, 0 ; GCN-IR-NEXT: v_cmp_eq_u64_e64 s[4:5], s[2:3], 0 ; GCN-IR-NEXT: v_cmp_gt_u64_e64 s[6:7], s[8:9], 63 -; GCN-IR-NEXT: v_cmp_eq_u64_e64 s[10:11], s[8:9], 63 +; GCN-IR-NEXT: v_cmp_eq_u64_e64 s[12:13], s[8:9], 63 ; GCN-IR-NEXT: s_or_b64 s[4:5], s[4:5], s[6:7] ; GCN-IR-NEXT: s_and_b64 s[6:7], s[4:5], exec ; GCN-IR-NEXT: s_cselect_b32 s7, 0, s3 ; GCN-IR-NEXT: s_cselect_b32 s6, 0, s2 -; GCN-IR-NEXT: s_or_b64 s[4:5], s[4:5], s[10:11] +; GCN-IR-NEXT: s_or_b64 s[4:5], s[4:5], s[12:13] ; GCN-IR-NEXT: s_andn2_b64 vcc, exec, s[4:5] ; GCN-IR-NEXT: s_mov_b64 s[4:5], 0 ; GCN-IR-NEXT: s_cbranch_vccz .LBB7_5 ; GCN-IR-NEXT: ; %bb.1: ; %udiv-bb1 -; GCN-IR-NEXT: s_add_u32 s10, s8, 1 -; GCN-IR-NEXT: s_addc_u32 s11, s9, 0 -; GCN-IR-NEXT: v_cmp_eq_u64_e64 s[6:7], s[10:11], 0 +; GCN-IR-NEXT: s_add_u32 s11, s8, 1 +; GCN-IR-NEXT: s_cselect_b64 s[6:7], -1, 0 +; GCN-IR-NEXT: s_or_b32 s6, s6, s7 +; GCN-IR-NEXT: s_cmp_lg_u32 s6, 0 +; GCN-IR-NEXT: s_addc_u32 s6, s9, 0 +; GCN-IR-NEXT: s_cselect_b64 s[6:7], -1, 0 ; GCN-IR-NEXT: s_sub_i32 s8, 63, s8 ; GCN-IR-NEXT: s_andn2_b64 vcc, exec, s[6:7] ; GCN-IR-NEXT: s_lshl_b64 s[6:7], s[2:3], s8 ; GCN-IR-NEXT: s_cbranch_vccz .LBB7_4 ; GCN-IR-NEXT: ; %bb.2: ; %udiv-preheader -; GCN-IR-NEXT: s_lshr_b64 s[10:11], s[2:3], s10 -; GCN-IR-NEXT: s_add_u32 s8, s12, 0xffffffc4 -; GCN-IR-NEXT: s_addc_u32 s9, 0, -1 -; GCN-IR-NEXT: s_mov_b64 s[12:13], 0 +; GCN-IR-NEXT: s_lshr_b64 s[8:9], s[2:3], s11 +; GCN-IR-NEXT: s_add_u32 s12, s10, 0xffffffc4 +; GCN-IR-NEXT: s_addc_u32 s13, 0, -1 +; GCN-IR-NEXT: s_mov_b64 s[10:11], 0 ; GCN-IR-NEXT: s_mov_b32 s5, 0 ; GCN-IR-NEXT: .LBB7_3: ; %udiv-do-while ; GCN-IR-NEXT: ; =>This Inner Loop Header: Depth=1 -; GCN-IR-NEXT: s_lshl_b64 s[10:11], s[10:11], 1 +; GCN-IR-NEXT: s_lshl_b64 s[8:9], s[8:9], 1 ; GCN-IR-NEXT: s_lshr_b32 s4, s7, 31 ; GCN-IR-NEXT: s_lshl_b64 s[6:7], s[6:7], 1 -; GCN-IR-NEXT: s_or_b64 s[10:11], s[10:11], s[4:5] -; GCN-IR-NEXT: s_or_b64 s[6:7], s[12:13], s[6:7] -; GCN-IR-NEXT: s_sub_u32 s4, 23, s10 -; GCN-IR-NEXT: s_subb_u32 s4, 0, s11 -; GCN-IR-NEXT: s_ashr_i32 s12, s4, 31 -; GCN-IR-NEXT: s_and_b32 s4, s12, 1 -; GCN-IR-NEXT: s_and_b32 s12, s12, 24 -; GCN-IR-NEXT: s_sub_u32 s10, s10, s12 -; GCN-IR-NEXT: s_subb_u32 s11, s11, 0 -; GCN-IR-NEXT: s_add_u32 s8, s8, 1 -; GCN-IR-NEXT: s_addc_u32 s9, s9, 0 -; GCN-IR-NEXT: v_cmp_eq_u64_e64 s[14:15], s[8:9], 0 -; GCN-IR-NEXT: s_mov_b64 s[12:13], s[4:5] +; GCN-IR-NEXT: s_or_b64 s[8:9], s[8:9], s[4:5] +; GCN-IR-NEXT: s_or_b64 s[6:7], s[10:11], s[6:7] +; GCN-IR-NEXT: s_sub_u32 s4, 23, s8 +; GCN-IR-NEXT: s_subb_u32 s4, 0, s9 +; GCN-IR-NEXT: s_ashr_i32 s10, s4, 31 +; GCN-IR-NEXT: s_and_b32 s4, s10, 1 +; GCN-IR-NEXT: s_and_b32 s10, s10, 24 +; GCN-IR-NEXT: s_sub_u32 s8, s8, s10 +; GCN-IR-NEXT: s_subb_u32 s9, s9, 0 +; GCN-IR-NEXT: s_add_u32 s12, s12, 1 +; GCN-IR-NEXT: s_cselect_b64 s[14:15], -1, 0 +; GCN-IR-NEXT: s_or_b32 s14, s14, s15 +; GCN-IR-NEXT: s_cmp_lg_u32 s14, 0 +; GCN-IR-NEXT: s_addc_u32 s13, s13, 0 +; GCN-IR-NEXT: s_cselect_b64 s[14:15], -1, 0 +; GCN-IR-NEXT: s_mov_b64 s[10:11], s[4:5] ; GCN-IR-NEXT: s_and_b64 vcc, exec, s[14:15] ; GCN-IR-NEXT: s_cbranch_vccz .LBB7_3 ; GCN-IR-NEXT: .LBB7_4: ; %Flow6 @@ -1241,8 +1258,8 @@ define i64 @v_test_urem_pow2_k_num_i64(i64 %x) { ; GCN-IR-NEXT: v_ffbh_u32_e32 v2, v0 ; GCN-IR-NEXT: v_add_i32_e32 v2, vcc, 32, v2 ; GCN-IR-NEXT: v_ffbh_u32_e32 v3, v1 -; GCN-IR-NEXT: v_min_u32_e32 v10, v2, v3 -; GCN-IR-NEXT: v_add_i32_e32 v2, vcc, 0xffffffd0, v10 +; GCN-IR-NEXT: v_min_u32_e32 v8, v2, v3 +; GCN-IR-NEXT: v_add_i32_e32 v2, vcc, 0xffffffd0, v8 ; GCN-IR-NEXT: v_addc_u32_e64 v3, s[6:7], 0, -1, vcc ; GCN-IR-NEXT: v_cmp_eq_u64_e64 s[4:5], 0, v[0:1] ; GCN-IR-NEXT: v_cmp_lt_u64_e32 vcc, 63, v[2:3] @@ -1257,54 +1274,53 @@ define i64 @v_test_urem_pow2_k_num_i64(i64 %x) { ; GCN-IR-NEXT: s_cbranch_execz .LBB8_6 ; GCN-IR-NEXT: ; %bb.1: ; %udiv-bb1 ; GCN-IR-NEXT: v_add_i32_e32 v6, vcc, 1, v2 +; GCN-IR-NEXT: v_addc_u32_e32 v3, vcc, 0, v3, vcc ; GCN-IR-NEXT: v_sub_i32_e64 v2, s[4:5], 63, v2 -; GCN-IR-NEXT: v_addc_u32_e32 v7, vcc, 0, v3, vcc -; GCN-IR-NEXT: s_mov_b64 s[4:5], 0x8000 +; GCN-IR-NEXT: s_mov_b64 s[8:9], 0x8000 +; GCN-IR-NEXT: v_lshl_b64 v[2:3], s[8:9], v2 ; GCN-IR-NEXT: v_mov_b32_e32 v4, 0 -; GCN-IR-NEXT: v_cmp_ne_u64_e32 vcc, 0, v[6:7] -; GCN-IR-NEXT: v_lshl_b64 v[2:3], s[4:5], v2 ; GCN-IR-NEXT: v_mov_b32_e32 v5, 0 -; GCN-IR-NEXT: s_and_saveexec_b64 s[8:9], vcc -; GCN-IR-NEXT: s_xor_b64 s[8:9], exec, s[8:9] +; GCN-IR-NEXT: s_xor_b64 s[4:5], vcc, -1 +; GCN-IR-NEXT: s_and_saveexec_b64 s[10:11], s[4:5] +; GCN-IR-NEXT: s_xor_b64 s[4:5], exec, s[10:11] ; GCN-IR-NEXT: s_cbranch_execz .LBB8_5 ; GCN-IR-NEXT: ; %bb.2: ; %udiv-preheader -; GCN-IR-NEXT: v_add_i32_e32 v12, vcc, -1, v0 -; GCN-IR-NEXT: v_addc_u32_e32 v13, vcc, -1, v1, vcc -; GCN-IR-NEXT: v_lshr_b64 v[8:9], s[4:5], v6 -; GCN-IR-NEXT: v_sub_i32_e32 v6, vcc, 47, v10 -; GCN-IR-NEXT: v_mov_b32_e32 v10, 0 -; GCN-IR-NEXT: v_subb_u32_e64 v7, s[4:5], 0, 0, vcc -; GCN-IR-NEXT: s_mov_b64 s[10:11], 0 -; GCN-IR-NEXT: v_mov_b32_e32 v11, 0 +; GCN-IR-NEXT: v_add_i32_e32 v10, vcc, -1, v0 +; GCN-IR-NEXT: v_addc_u32_e32 v11, vcc, -1, v1, vcc +; GCN-IR-NEXT: v_sub_i32_e32 v12, vcc, 47, v8 +; GCN-IR-NEXT: v_lshr_b64 v[6:7], s[8:9], v6 +; GCN-IR-NEXT: v_subb_u32_e64 v13, s[8:9], 0, 0, vcc +; GCN-IR-NEXT: v_mov_b32_e32 v8, 0 +; GCN-IR-NEXT: s_mov_b64 s[8:9], 0 +; GCN-IR-NEXT: v_mov_b32_e32 v9, 0 ; GCN-IR-NEXT: v_mov_b32_e32 v5, 0 ; GCN-IR-NEXT: .LBB8_3: ; %udiv-do-while ; GCN-IR-NEXT: ; =>This Inner Loop Header: Depth=1 -; GCN-IR-NEXT: v_lshl_b64 v[8:9], v[8:9], 1 +; GCN-IR-NEXT: v_lshl_b64 v[6:7], v[6:7], 1 ; GCN-IR-NEXT: v_lshrrev_b32_e32 v4, 31, v3 -; GCN-IR-NEXT: v_or_b32_e32 v8, v8, v4 +; GCN-IR-NEXT: v_or_b32_e32 v6, v6, v4 ; GCN-IR-NEXT: v_lshl_b64 v[2:3], v[2:3], 1 -; GCN-IR-NEXT: v_sub_i32_e32 v4, vcc, v12, v8 -; GCN-IR-NEXT: v_subb_u32_e32 v4, vcc, v13, v9, vcc -; GCN-IR-NEXT: v_or_b32_e32 v2, v10, v2 -; GCN-IR-NEXT: v_ashrrev_i32_e32 v10, 31, v4 -; GCN-IR-NEXT: v_add_i32_e32 v6, vcc, 1, v6 -; GCN-IR-NEXT: v_or_b32_e32 v3, v11, v3 -; GCN-IR-NEXT: v_and_b32_e32 v4, 1, v10 -; GCN-IR-NEXT: v_and_b32_e32 v11, v10, v1 -; GCN-IR-NEXT: v_and_b32_e32 v10, v10, v0 -; GCN-IR-NEXT: v_addc_u32_e32 v7, vcc, 0, v7, vcc -; GCN-IR-NEXT: v_cmp_eq_u64_e32 vcc, 0, v[6:7] -; GCN-IR-NEXT: v_sub_i32_e64 v8, s[4:5], v8, v10 -; GCN-IR-NEXT: v_subb_u32_e64 v9, s[4:5], v9, v11, s[4:5] -; GCN-IR-NEXT: v_mov_b32_e32 v11, v5 -; GCN-IR-NEXT: s_or_b64 s[10:11], vcc, s[10:11] -; GCN-IR-NEXT: v_mov_b32_e32 v10, v4 -; GCN-IR-NEXT: s_andn2_b64 exec, exec, s[10:11] +; GCN-IR-NEXT: v_sub_i32_e32 v4, vcc, v10, v6 +; GCN-IR-NEXT: v_subb_u32_e32 v4, vcc, v11, v7, vcc +; GCN-IR-NEXT: v_or_b32_e32 v2, v8, v2 +; GCN-IR-NEXT: v_ashrrev_i32_e32 v8, 31, v4 +; GCN-IR-NEXT: v_or_b32_e32 v3, v9, v3 +; GCN-IR-NEXT: v_and_b32_e32 v4, 1, v8 +; GCN-IR-NEXT: v_and_b32_e32 v9, v8, v1 +; GCN-IR-NEXT: v_and_b32_e32 v8, v8, v0 +; GCN-IR-NEXT: v_sub_i32_e32 v6, vcc, v6, v8 +; GCN-IR-NEXT: v_subb_u32_e32 v7, vcc, v7, v9, vcc +; GCN-IR-NEXT: v_add_i32_e32 v12, vcc, 1, v12 +; GCN-IR-NEXT: v_addc_u32_e32 v13, vcc, 0, v13, vcc +; GCN-IR-NEXT: v_mov_b32_e32 v9, v5 +; GCN-IR-NEXT: s_or_b64 s[8:9], vcc, s[8:9] +; GCN-IR-NEXT: v_mov_b32_e32 v8, v4 +; GCN-IR-NEXT: s_andn2_b64 exec, exec, s[8:9] ; GCN-IR-NEXT: s_cbranch_execnz .LBB8_3 ; GCN-IR-NEXT: ; %bb.4: ; %Flow -; GCN-IR-NEXT: s_or_b64 exec, exec, s[10:11] -; GCN-IR-NEXT: .LBB8_5: ; %Flow4 ; GCN-IR-NEXT: s_or_b64 exec, exec, s[8:9] +; GCN-IR-NEXT: .LBB8_5: ; %Flow4 +; GCN-IR-NEXT: s_or_b64 exec, exec, s[4:5] ; GCN-IR-NEXT: v_lshl_b64 v[2:3], v[2:3], 1 ; GCN-IR-NEXT: v_or_b32_e32 v5, v5, v3 ; GCN-IR-NEXT: v_or_b32_e32 v4, v4, v2 @@ -1337,8 +1353,8 @@ define i64 @v_test_urem_pow2_k_den_i64(i64 %x) { ; GCN-IR-NEXT: v_ffbh_u32_e32 v2, v0 ; GCN-IR-NEXT: v_add_i32_e64 v2, s[4:5], 32, v2 ; GCN-IR-NEXT: v_ffbh_u32_e32 v3, v1 -; GCN-IR-NEXT: v_min_u32_e32 v10, v2, v3 -; GCN-IR-NEXT: v_sub_i32_e64 v2, s[4:5], 48, v10 +; GCN-IR-NEXT: v_min_u32_e32 v8, v2, v3 +; GCN-IR-NEXT: v_sub_i32_e64 v2, s[4:5], 48, v8 ; GCN-IR-NEXT: v_subb_u32_e64 v3, s[4:5], 0, 0, s[4:5] ; GCN-IR-NEXT: v_cmp_eq_u64_e32 vcc, 0, v[0:1] ; GCN-IR-NEXT: v_cmp_lt_u64_e64 s[4:5], 63, v[2:3] @@ -1352,51 +1368,50 @@ define i64 @v_test_urem_pow2_k_den_i64(i64 %x) { ; GCN-IR-NEXT: s_cbranch_execz .LBB9_6 ; GCN-IR-NEXT: ; %bb.1: ; %udiv-bb1 ; GCN-IR-NEXT: v_add_i32_e32 v6, vcc, 1, v2 -; GCN-IR-NEXT: v_addc_u32_e32 v7, vcc, 0, v3, vcc +; GCN-IR-NEXT: v_addc_u32_e32 v3, vcc, 0, v3, vcc ; GCN-IR-NEXT: v_sub_i32_e64 v2, s[4:5], 63, v2 -; GCN-IR-NEXT: v_mov_b32_e32 v4, 0 -; GCN-IR-NEXT: v_cmp_ne_u64_e32 vcc, 0, v[6:7] ; GCN-IR-NEXT: v_lshl_b64 v[2:3], v[0:1], v2 +; GCN-IR-NEXT: v_mov_b32_e32 v4, 0 ; GCN-IR-NEXT: v_mov_b32_e32 v5, 0 -; GCN-IR-NEXT: s_and_saveexec_b64 s[4:5], vcc -; GCN-IR-NEXT: s_xor_b64 s[8:9], exec, s[4:5] +; GCN-IR-NEXT: s_xor_b64 s[4:5], vcc, -1 +; GCN-IR-NEXT: s_and_saveexec_b64 s[8:9], s[4:5] +; GCN-IR-NEXT: s_xor_b64 s[4:5], exec, s[8:9] ; GCN-IR-NEXT: s_cbranch_execz .LBB9_5 ; GCN-IR-NEXT: ; %bb.2: ; %udiv-preheader -; GCN-IR-NEXT: v_lshr_b64 v[8:9], v[0:1], v6 -; GCN-IR-NEXT: v_add_i32_e32 v6, vcc, 0xffffffcf, v10 -; GCN-IR-NEXT: v_mov_b32_e32 v10, 0 -; GCN-IR-NEXT: v_addc_u32_e64 v7, s[4:5], 0, -1, vcc -; GCN-IR-NEXT: s_mov_b64 s[10:11], 0 -; GCN-IR-NEXT: v_mov_b32_e32 v11, 0 +; GCN-IR-NEXT: v_add_i32_e32 v10, vcc, 0xffffffcf, v8 +; GCN-IR-NEXT: v_lshr_b64 v[6:7], v[0:1], v6 +; GCN-IR-NEXT: v_addc_u32_e64 v11, s[8:9], 0, -1, vcc +; GCN-IR-NEXT: v_mov_b32_e32 v8, 0 +; GCN-IR-NEXT: s_mov_b64 s[8:9], 0 +; GCN-IR-NEXT: v_mov_b32_e32 v9, 0 ; GCN-IR-NEXT: v_mov_b32_e32 v5, 0 -; GCN-IR-NEXT: s_movk_i32 s12, 0x7fff +; GCN-IR-NEXT: s_movk_i32 s10, 0x7fff ; GCN-IR-NEXT: .LBB9_3: ; %udiv-do-while ; GCN-IR-NEXT: ; =>This Inner Loop Header: Depth=1 -; GCN-IR-NEXT: v_lshl_b64 v[8:9], v[8:9], 1 +; GCN-IR-NEXT: v_lshl_b64 v[6:7], v[6:7], 1 ; GCN-IR-NEXT: v_lshrrev_b32_e32 v4, 31, v3 -; GCN-IR-NEXT: v_or_b32_e32 v8, v8, v4 -; GCN-IR-NEXT: v_sub_i32_e32 v4, vcc, s12, v8 +; GCN-IR-NEXT: v_or_b32_e32 v6, v6, v4 ; GCN-IR-NEXT: v_lshl_b64 v[2:3], v[2:3], 1 -; GCN-IR-NEXT: v_subb_u32_e32 v4, vcc, 0, v9, vcc -; GCN-IR-NEXT: v_add_i32_e32 v6, vcc, 1, v6 -; GCN-IR-NEXT: v_or_b32_e32 v2, v10, v2 -; GCN-IR-NEXT: v_ashrrev_i32_e32 v10, 31, v4 -; GCN-IR-NEXT: v_addc_u32_e32 v7, vcc, 0, v7, vcc -; GCN-IR-NEXT: v_and_b32_e32 v4, 1, v10 -; GCN-IR-NEXT: v_and_b32_e32 v10, 0x8000, v10 -; GCN-IR-NEXT: v_cmp_eq_u64_e32 vcc, 0, v[6:7] -; GCN-IR-NEXT: v_or_b32_e32 v3, v11, v3 -; GCN-IR-NEXT: v_sub_i32_e64 v8, s[4:5], v8, v10 -; GCN-IR-NEXT: v_mov_b32_e32 v11, v5 -; GCN-IR-NEXT: v_subbrev_u32_e64 v9, s[4:5], 0, v9, s[4:5] -; GCN-IR-NEXT: s_or_b64 s[10:11], vcc, s[10:11] -; GCN-IR-NEXT: v_mov_b32_e32 v10, v4 -; GCN-IR-NEXT: s_andn2_b64 exec, exec, s[10:11] +; GCN-IR-NEXT: v_sub_i32_e32 v4, vcc, s10, v6 +; GCN-IR-NEXT: v_subb_u32_e32 v4, vcc, 0, v7, vcc +; GCN-IR-NEXT: v_or_b32_e32 v2, v8, v2 +; GCN-IR-NEXT: v_ashrrev_i32_e32 v8, 31, v4 +; GCN-IR-NEXT: v_and_b32_e32 v4, 1, v8 +; GCN-IR-NEXT: v_and_b32_e32 v8, 0x8000, v8 +; GCN-IR-NEXT: v_sub_i32_e32 v6, vcc, v6, v8 +; GCN-IR-NEXT: v_subbrev_u32_e32 v7, vcc, 0, v7, vcc +; GCN-IR-NEXT: v_add_i32_e32 v10, vcc, 1, v10 +; GCN-IR-NEXT: v_or_b32_e32 v3, v9, v3 +; GCN-IR-NEXT: v_addc_u32_e32 v11, vcc, 0, v11, vcc +; GCN-IR-NEXT: v_mov_b32_e32 v9, v5 +; GCN-IR-NEXT: s_or_b64 s[8:9], vcc, s[8:9] +; GCN-IR-NEXT: v_mov_b32_e32 v8, v4 +; GCN-IR-NEXT: s_andn2_b64 exec, exec, s[8:9] ; GCN-IR-NEXT: s_cbranch_execnz .LBB9_3 ; GCN-IR-NEXT: ; %bb.4: ; %Flow -; GCN-IR-NEXT: s_or_b64 exec, exec, s[10:11] -; GCN-IR-NEXT: .LBB9_5: ; %Flow4 ; GCN-IR-NEXT: s_or_b64 exec, exec, s[8:9] +; GCN-IR-NEXT: .LBB9_5: ; %Flow4 +; GCN-IR-NEXT: s_or_b64 exec, exec, s[4:5] ; GCN-IR-NEXT: v_lshl_b64 v[2:3], v[2:3], 1 ; GCN-IR-NEXT: v_or_b32_e32 v5, v5, v3 ; GCN-IR-NEXT: v_or_b32_e32 v4, v4, v2 diff --git a/llvm/test/CodeGen/AMDGPU/usubo.ll b/llvm/test/CodeGen/AMDGPU/usubo.ll index 0289dab4588a2..d67a7b151948e 100644 --- a/llvm/test/CodeGen/AMDGPU/usubo.ll +++ b/llvm/test/CodeGen/AMDGPU/usubo.ll @@ -14,15 +14,16 @@ define amdgpu_kernel void @s_usubo_i64_zext(ptr addrspace(1) %out, i64 %a, i64 % ; SI-NEXT: s_mov_b32 s6, -1 ; SI-NEXT: s_waitcnt lgkmcnt(0) ; SI-NEXT: s_mov_b32 s4, s0 -; SI-NEXT: s_sub_u32 s0, s2, s8 -; SI-NEXT: v_mov_b32_e32 v0, s2 +; SI-NEXT: s_sub_u32 s2, s2, s8 ; SI-NEXT: s_mov_b32 s5, s1 -; SI-NEXT: s_subb_u32 s1, s3, s9 +; SI-NEXT: s_cselect_b64 s[0:1], -1, 0 +; SI-NEXT: s_or_b32 s0, s0, s1 +; SI-NEXT: s_cmp_lg_u32 s0, 0 +; SI-NEXT: s_subb_u32 s3, s3, s9 +; SI-NEXT: s_cselect_b64 s[0:1], -1, 0 +; SI-NEXT: v_cndmask_b32_e64 v0, 0, 1, s[0:1] ; SI-NEXT: v_mov_b32_e32 v1, s3 -; SI-NEXT: v_cmp_gt_u64_e32 vcc, s[0:1], v[0:1] -; SI-NEXT: v_mov_b32_e32 v1, s1 -; SI-NEXT: v_cndmask_b32_e64 v0, 0, 1, vcc -; SI-NEXT: v_add_i32_e32 v0, vcc, s0, v0 +; SI-NEXT: v_add_i32_e32 v0, vcc, s2, v0 ; SI-NEXT: v_addc_u32_e32 v1, vcc, 0, v1, vcc ; SI-NEXT: buffer_store_dwordx2 v[0:1], off, s[4:7], 0 ; SI-NEXT: s_endpgm @@ -33,15 +34,15 @@ define amdgpu_kernel void @s_usubo_i64_zext(ptr addrspace(1) %out, i64 %a, i64 % ; VI-NEXT: s_load_dwordx2 s[4:5], s[4:5], 0x34 ; VI-NEXT: s_waitcnt lgkmcnt(0) ; VI-NEXT: v_mov_b32_e32 v0, s0 -; VI-NEXT: s_sub_u32 s0, s2, s4 -; VI-NEXT: v_mov_b32_e32 v2, s2 +; VI-NEXT: s_sub_u32 s2, s2, s4 ; VI-NEXT: v_mov_b32_e32 v1, s1 +; VI-NEXT: s_cselect_b64 s[0:1], -1, 0 +; VI-NEXT: s_cmp_lg_u64 s[0:1], 0 +; VI-NEXT: s_subb_u32 s3, s3, s5 +; VI-NEXT: s_cselect_b64 s[0:1], -1, 0 +; VI-NEXT: v_cndmask_b32_e64 v2, 0, 1, s[0:1] ; VI-NEXT: v_mov_b32_e32 v3, s3 -; VI-NEXT: s_subb_u32 s1, s3, s5 -; VI-NEXT: v_cmp_gt_u64_e32 vcc, s[0:1], v[2:3] -; VI-NEXT: v_mov_b32_e32 v3, s1 -; VI-NEXT: v_cndmask_b32_e64 v2, 0, 1, vcc -; VI-NEXT: v_add_u32_e32 v2, vcc, s0, v2 +; VI-NEXT: v_add_u32_e32 v2, vcc, s2, v2 ; VI-NEXT: v_addc_u32_e32 v3, vcc, 0, v3, vcc ; VI-NEXT: flat_store_dwordx2 v[0:1], v[2:3] ; VI-NEXT: s_endpgm @@ -52,14 +53,14 @@ define amdgpu_kernel void @s_usubo_i64_zext(ptr addrspace(1) %out, i64 %a, i64 % ; GFX9-NEXT: s_load_dwordx2 s[6:7], s[4:5], 0x34 ; GFX9-NEXT: v_mov_b32_e32 v2, 0 ; GFX9-NEXT: s_waitcnt lgkmcnt(0) -; GFX9-NEXT: v_mov_b32_e32 v0, s2 -; GFX9-NEXT: s_sub_u32 s4, s2, s6 -; GFX9-NEXT: v_mov_b32_e32 v1, s3 -; GFX9-NEXT: s_subb_u32 s5, s3, s7 -; GFX9-NEXT: v_cmp_gt_u64_e32 vcc, s[4:5], v[0:1] -; GFX9-NEXT: v_mov_b32_e32 v1, s5 -; GFX9-NEXT: v_cndmask_b32_e64 v0, 0, 1, vcc -; GFX9-NEXT: v_add_co_u32_e32 v0, vcc, s4, v0 +; GFX9-NEXT: s_sub_u32 s6, s2, s6 +; GFX9-NEXT: s_cselect_b64 s[4:5], -1, 0 +; GFX9-NEXT: s_cmp_lg_u64 s[4:5], 0 +; GFX9-NEXT: s_subb_u32 s4, s3, s7 +; GFX9-NEXT: s_cselect_b64 s[2:3], -1, 0 +; GFX9-NEXT: v_cndmask_b32_e64 v0, 0, 1, s[2:3] +; GFX9-NEXT: v_mov_b32_e32 v1, s4 +; GFX9-NEXT: v_add_co_u32_e32 v0, vcc, s6, v0 ; GFX9-NEXT: v_addc_co_u32_e32 v1, vcc, 0, v1, vcc ; GFX9-NEXT: global_store_dwordx2 v2, v[0:1], s[0:1] ; GFX9-NEXT: s_endpgm @@ -71,12 +72,14 @@ define amdgpu_kernel void @s_usubo_i64_zext(ptr addrspace(1) %out, i64 %a, i64 % ; GFX10-NEXT: s_load_dwordx2 s[6:7], s[4:5], 0x34 ; GFX10-NEXT: v_mov_b32_e32 v2, 0 ; GFX10-NEXT: s_waitcnt lgkmcnt(0) -; GFX10-NEXT: s_sub_u32 s4, s2, s6 -; GFX10-NEXT: s_subb_u32 s5, s3, s7 -; GFX10-NEXT: v_cmp_gt_u64_e64 s2, s[4:5], s[2:3] -; GFX10-NEXT: v_cndmask_b32_e64 v0, 0, 1, s2 -; GFX10-NEXT: v_add_co_u32 v0, s2, s4, v0 -; GFX10-NEXT: v_add_co_ci_u32_e64 v1, s2, s5, 0, s2 +; GFX10-NEXT: s_sub_u32 s2, s2, s6 +; GFX10-NEXT: s_cselect_b32 s4, -1, 0 +; GFX10-NEXT: s_cmp_lg_u32 s4, 0 +; GFX10-NEXT: s_subb_u32 s3, s3, s7 +; GFX10-NEXT: s_cselect_b32 s4, -1, 0 +; GFX10-NEXT: v_cndmask_b32_e64 v0, 0, 1, s4 +; GFX10-NEXT: v_add_co_u32 v0, s2, s2, v0 +; GFX10-NEXT: v_add_co_ci_u32_e64 v1, s2, s3, 0, s2 ; GFX10-NEXT: global_store_dwordx2 v2, v[0:1], s[0:1] ; GFX10-NEXT: s_endpgm ; @@ -87,14 +90,16 @@ define amdgpu_kernel void @s_usubo_i64_zext(ptr addrspace(1) %out, i64 %a, i64 % ; GFX11-NEXT: s_load_b64 s[4:5], s[4:5], 0x34 ; GFX11-NEXT: v_mov_b32_e32 v2, 0 ; GFX11-NEXT: s_waitcnt lgkmcnt(0) -; GFX11-NEXT: s_sub_u32 s4, s2, s4 -; GFX11-NEXT: s_subb_u32 s5, s3, s5 -; GFX11-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(VALU_DEP_1) -; GFX11-NEXT: v_cmp_gt_u64_e64 s2, s[4:5], s[2:3] -; GFX11-NEXT: v_cndmask_b32_e64 v0, 0, 1, s2 +; GFX11-NEXT: s_sub_u32 s2, s2, s4 +; GFX11-NEXT: s_cselect_b32 s4, -1, 0 +; GFX11-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_2) | instid1(SALU_CYCLE_1) +; GFX11-NEXT: s_cmp_lg_u32 s4, 0 +; GFX11-NEXT: s_subb_u32 s3, s3, s5 +; GFX11-NEXT: s_cselect_b32 s4, -1, 0 +; GFX11-NEXT: v_cndmask_b32_e64 v0, 0, 1, s4 ; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1) -; GFX11-NEXT: v_add_co_u32 v0, s2, s4, v0 -; GFX11-NEXT: v_add_co_ci_u32_e64 v1, null, s5, 0, s2 +; GFX11-NEXT: v_add_co_u32 v0, s2, s2, v0 +; GFX11-NEXT: v_add_co_ci_u32_e64 v1, null, s3, 0, s2 ; GFX11-NEXT: global_store_b64 v2, v[0:1], s[0:1] ; GFX11-NEXT: s_endpgm %usub = call { i64, i1 } @llvm.usub.with.overflow.i64(i64 %a, i64 %b) #0 @@ -435,21 +440,23 @@ define amdgpu_kernel void @s_usubo_i64(ptr addrspace(1) %out, ptr addrspace(1) % ; SI-NEXT: s_mov_b32 s11, 0xf000 ; SI-NEXT: s_mov_b32 s10, -1 ; SI-NEXT: s_waitcnt lgkmcnt(0) -; SI-NEXT: s_sub_u32 s6, s4, s6 -; SI-NEXT: v_mov_b32_e32 v0, s4 -; SI-NEXT: s_subb_u32 s7, s5, s7 -; SI-NEXT: v_mov_b32_e32 v1, s5 -; SI-NEXT: v_cmp_gt_u64_e32 vcc, s[6:7], v[0:1] -; SI-NEXT: v_mov_b32_e32 v2, s6 +; SI-NEXT: s_sub_u32 s4, s4, s6 +; SI-NEXT: s_cselect_b64 s[12:13], -1, 0 +; SI-NEXT: s_or_b32 s6, s12, s13 +; SI-NEXT: s_cmp_lg_u32 s6, 0 +; SI-NEXT: s_subb_u32 s5, s5, s7 ; SI-NEXT: s_mov_b32 s8, s0 ; SI-NEXT: s_mov_b32 s9, s1 +; SI-NEXT: v_mov_b32_e32 v0, s4 +; SI-NEXT: v_mov_b32_e32 v1, s5 +; SI-NEXT: s_cselect_b64 s[4:5], -1, 0 ; SI-NEXT: s_mov_b32 s0, s2 ; SI-NEXT: s_mov_b32 s1, s3 ; SI-NEXT: s_mov_b32 s2, s10 ; SI-NEXT: s_mov_b32 s3, s11 -; SI-NEXT: v_mov_b32_e32 v3, s7 -; SI-NEXT: v_cndmask_b32_e64 v0, 0, 1, vcc -; SI-NEXT: buffer_store_dwordx2 v[2:3], off, s[8:11], 0 +; SI-NEXT: buffer_store_dwordx2 v[0:1], off, s[8:11], 0 +; SI-NEXT: s_waitcnt expcnt(0) +; SI-NEXT: v_cndmask_b32_e64 v0, 0, 1, s[4:5] ; SI-NEXT: buffer_store_byte v0, off, s[0:3], 0 ; SI-NEXT: s_endpgm ; @@ -457,37 +464,37 @@ define amdgpu_kernel void @s_usubo_i64(ptr addrspace(1) %out, ptr addrspace(1) % ; VI: ; %bb.0: ; VI-NEXT: s_load_dwordx8 s[0:7], s[4:5], 0x24 ; VI-NEXT: s_waitcnt lgkmcnt(0) +; VI-NEXT: v_mov_b32_e32 v2, s2 +; VI-NEXT: s_sub_u32 s2, s4, s6 ; VI-NEXT: v_mov_b32_e32 v0, s0 -; VI-NEXT: s_sub_u32 s0, s4, s6 -; VI-NEXT: v_mov_b32_e32 v4, s4 ; VI-NEXT: v_mov_b32_e32 v1, s1 -; VI-NEXT: s_subb_u32 s1, s5, s7 -; VI-NEXT: v_mov_b32_e32 v5, s5 -; VI-NEXT: v_mov_b32_e32 v7, s1 -; VI-NEXT: v_cmp_gt_u64_e32 vcc, s[0:1], v[4:5] -; VI-NEXT: v_mov_b32_e32 v6, s0 -; VI-NEXT: v_mov_b32_e32 v2, s2 +; VI-NEXT: s_cselect_b64 s[0:1], -1, 0 +; VI-NEXT: s_cmp_lg_u64 s[0:1], 0 +; VI-NEXT: s_subb_u32 s0, s5, s7 +; VI-NEXT: v_mov_b32_e32 v4, s2 +; VI-NEXT: v_mov_b32_e32 v5, s0 +; VI-NEXT: s_cselect_b64 s[0:1], -1, 0 ; VI-NEXT: v_mov_b32_e32 v3, s3 -; VI-NEXT: flat_store_dwordx2 v[0:1], v[6:7] -; VI-NEXT: v_cndmask_b32_e64 v0, 0, 1, vcc +; VI-NEXT: flat_store_dwordx2 v[0:1], v[4:5] +; VI-NEXT: v_cndmask_b32_e64 v0, 0, 1, s[0:1] ; VI-NEXT: flat_store_byte v[2:3], v0 ; VI-NEXT: s_endpgm ; ; GFX9-LABEL: s_usubo_i64: ; GFX9: ; %bb.0: ; GFX9-NEXT: s_load_dwordx8 s[8:15], s[4:5], 0x24 -; GFX9-NEXT: v_mov_b32_e32 v4, 0 +; GFX9-NEXT: v_mov_b32_e32 v2, 0 ; GFX9-NEXT: s_waitcnt lgkmcnt(0) -; GFX9-NEXT: s_sub_u32 s0, s12, s14 -; GFX9-NEXT: v_mov_b32_e32 v0, s12 -; GFX9-NEXT: v_mov_b32_e32 v1, s13 -; GFX9-NEXT: s_subb_u32 s1, s13, s15 -; GFX9-NEXT: v_mov_b32_e32 v3, s1 -; GFX9-NEXT: v_cmp_gt_u64_e32 vcc, s[0:1], v[0:1] -; GFX9-NEXT: v_mov_b32_e32 v2, s0 -; GFX9-NEXT: v_cndmask_b32_e64 v0, 0, 1, vcc -; GFX9-NEXT: global_store_dwordx2 v4, v[2:3], s[8:9] -; GFX9-NEXT: global_store_byte v4, v0, s[10:11] +; GFX9-NEXT: s_sub_u32 s2, s12, s14 +; GFX9-NEXT: s_cselect_b64 s[0:1], -1, 0 +; GFX9-NEXT: s_cmp_lg_u64 s[0:1], 0 +; GFX9-NEXT: s_subb_u32 s0, s13, s15 +; GFX9-NEXT: v_mov_b32_e32 v0, s2 +; GFX9-NEXT: v_mov_b32_e32 v1, s0 +; GFX9-NEXT: s_cselect_b64 s[0:1], -1, 0 +; GFX9-NEXT: v_cndmask_b32_e64 v3, 0, 1, s[0:1] +; GFX9-NEXT: global_store_dwordx2 v2, v[0:1], s[8:9] +; GFX9-NEXT: global_store_byte v2, v3, s[10:11] ; GFX9-NEXT: s_endpgm ; ; GFX10-LABEL: s_usubo_i64: @@ -496,10 +503,12 @@ define amdgpu_kernel void @s_usubo_i64(ptr addrspace(1) %out, ptr addrspace(1) % ; GFX10-NEXT: v_mov_b32_e32 v2, 0 ; GFX10-NEXT: s_waitcnt lgkmcnt(0) ; GFX10-NEXT: s_sub_u32 s0, s12, s14 -; GFX10-NEXT: s_subb_u32 s1, s13, s15 +; GFX10-NEXT: s_cselect_b32 s1, -1, 0 ; GFX10-NEXT: v_mov_b32_e32 v0, s0 +; GFX10-NEXT: s_cmp_lg_u32 s1, 0 +; GFX10-NEXT: s_subb_u32 s1, s13, s15 +; GFX10-NEXT: s_cselect_b32 s0, -1, 0 ; GFX10-NEXT: v_mov_b32_e32 v1, s1 -; GFX10-NEXT: v_cmp_gt_u64_e64 s0, s[0:1], s[12:13] ; GFX10-NEXT: v_cndmask_b32_e64 v3, 0, 1, s0 ; GFX10-NEXT: global_store_dwordx2 v2, v[0:1], s[8:9] ; GFX10-NEXT: global_store_byte v2, v3, s[10:11] @@ -509,12 +518,13 @@ define amdgpu_kernel void @s_usubo_i64(ptr addrspace(1) %out, ptr addrspace(1) % ; GFX11: ; %bb.0: ; GFX11-NEXT: s_load_b256 s[0:7], s[4:5], 0x24 ; GFX11-NEXT: s_waitcnt lgkmcnt(0) -; GFX11-NEXT: s_sub_u32 s6, s4, s6 -; GFX11-NEXT: s_subb_u32 s7, s5, s7 -; GFX11-NEXT: v_mov_b32_e32 v0, s6 -; GFX11-NEXT: v_cmp_gt_u64_e64 s4, s[6:7], s[4:5] -; GFX11-NEXT: v_dual_mov_b32 v2, 0 :: v_dual_mov_b32 v1, s7 -; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_2) +; GFX11-NEXT: s_sub_u32 s4, s4, s6 +; GFX11-NEXT: s_cselect_b32 s6, -1, 0 +; GFX11-NEXT: v_mov_b32_e32 v0, s4 +; GFX11-NEXT: s_cmp_lg_u32 s6, 0 +; GFX11-NEXT: s_subb_u32 s5, s5, s7 +; GFX11-NEXT: s_cselect_b32 s4, -1, 0 +; GFX11-NEXT: v_dual_mov_b32 v2, 0 :: v_dual_mov_b32 v1, s5 ; GFX11-NEXT: v_cndmask_b32_e64 v3, 0, 1, s4 ; GFX11-NEXT: s_clause 0x1 ; GFX11-NEXT: global_store_b64 v2, v[0:1], s[0:1] @@ -550,10 +560,10 @@ define amdgpu_kernel void @v_usubo_i64(ptr addrspace(1) %out, ptr addrspace(1) % ; SI-NEXT: s_mov_b32 s4, s2 ; SI-NEXT: s_mov_b32 s5, s3 ; SI-NEXT: s_waitcnt vmcnt(0) -; SI-NEXT: v_sub_i32_e32 v2, vcc, v0, v2 -; SI-NEXT: v_subb_u32_e32 v3, vcc, v1, v3, vcc -; SI-NEXT: v_cmp_gt_u64_e32 vcc, v[2:3], v[0:1] -; SI-NEXT: buffer_store_dwordx2 v[2:3], off, s[8:11], 0 +; SI-NEXT: v_sub_i32_e32 v0, vcc, v0, v2 +; SI-NEXT: v_subb_u32_e32 v1, vcc, v1, v3, vcc +; SI-NEXT: buffer_store_dwordx2 v[0:1], off, s[8:11], 0 +; SI-NEXT: s_waitcnt expcnt(0) ; SI-NEXT: v_cndmask_b32_e64 v0, 0, 1, vcc ; SI-NEXT: buffer_store_byte v0, off, s[4:7], 0 ; SI-NEXT: s_endpgm @@ -573,10 +583,9 @@ define amdgpu_kernel void @v_usubo_i64(ptr addrspace(1) %out, ptr addrspace(1) % ; VI-NEXT: v_mov_b32_e32 v6, s2 ; VI-NEXT: v_mov_b32_e32 v7, s3 ; VI-NEXT: s_waitcnt vmcnt(0) -; VI-NEXT: v_sub_u32_e32 v2, vcc, v0, v2 -; VI-NEXT: v_subb_u32_e32 v3, vcc, v1, v3, vcc -; VI-NEXT: v_cmp_gt_u64_e32 vcc, v[2:3], v[0:1] -; VI-NEXT: flat_store_dwordx2 v[4:5], v[2:3] +; VI-NEXT: v_sub_u32_e32 v0, vcc, v0, v2 +; VI-NEXT: v_subb_u32_e32 v1, vcc, v1, v3, vcc +; VI-NEXT: flat_store_dwordx2 v[4:5], v[0:1] ; VI-NEXT: v_cndmask_b32_e64 v0, 0, 1, vcc ; VI-NEXT: flat_store_byte v[6:7], v0 ; VI-NEXT: s_endpgm @@ -589,10 +598,9 @@ define amdgpu_kernel void @v_usubo_i64(ptr addrspace(1) %out, ptr addrspace(1) % ; GFX9-NEXT: global_load_dwordx2 v[0:1], v4, s[12:13] ; GFX9-NEXT: global_load_dwordx2 v[2:3], v4, s[14:15] ; GFX9-NEXT: s_waitcnt vmcnt(0) -; GFX9-NEXT: v_sub_co_u32_e32 v2, vcc, v0, v2 -; GFX9-NEXT: v_subb_co_u32_e32 v3, vcc, v1, v3, vcc -; GFX9-NEXT: v_cmp_gt_u64_e32 vcc, v[2:3], v[0:1] -; GFX9-NEXT: global_store_dwordx2 v4, v[2:3], s[8:9] +; GFX9-NEXT: v_sub_co_u32_e32 v0, vcc, v0, v2 +; GFX9-NEXT: v_subb_co_u32_e32 v1, vcc, v1, v3, vcc +; GFX9-NEXT: global_store_dwordx2 v4, v[0:1], s[8:9] ; GFX9-NEXT: v_cndmask_b32_e64 v0, 0, 1, vcc ; GFX9-NEXT: global_store_byte v4, v0, s[10:11] ; GFX9-NEXT: s_endpgm @@ -606,12 +614,11 @@ define amdgpu_kernel void @v_usubo_i64(ptr addrspace(1) %out, ptr addrspace(1) % ; GFX10-NEXT: global_load_dwordx2 v[0:1], v4, s[12:13] ; GFX10-NEXT: global_load_dwordx2 v[2:3], v4, s[14:15] ; GFX10-NEXT: s_waitcnt vmcnt(0) -; GFX10-NEXT: v_sub_co_u32 v2, vcc_lo, v0, v2 -; GFX10-NEXT: v_sub_co_ci_u32_e32 v3, vcc_lo, v1, v3, vcc_lo -; GFX10-NEXT: v_cmp_gt_u64_e32 vcc_lo, v[2:3], v[0:1] -; GFX10-NEXT: v_cndmask_b32_e64 v0, 0, 1, vcc_lo -; GFX10-NEXT: global_store_dwordx2 v4, v[2:3], s[8:9] -; GFX10-NEXT: global_store_byte v4, v0, s[10:11] +; GFX10-NEXT: v_sub_co_u32 v0, vcc_lo, v0, v2 +; GFX10-NEXT: v_sub_co_ci_u32_e32 v1, vcc_lo, v1, v3, vcc_lo +; GFX10-NEXT: v_cndmask_b32_e64 v2, 0, 1, vcc_lo +; GFX10-NEXT: global_store_dwordx2 v4, v[0:1], s[8:9] +; GFX10-NEXT: global_store_byte v4, v2, s[10:11] ; GFX10-NEXT: s_endpgm ; ; GFX11-LABEL: v_usubo_i64: @@ -623,14 +630,12 @@ define amdgpu_kernel void @v_usubo_i64(ptr addrspace(1) %out, ptr addrspace(1) % ; GFX11-NEXT: global_load_b64 v[0:1], v4, s[4:5] ; GFX11-NEXT: global_load_b64 v[2:3], v4, s[6:7] ; GFX11-NEXT: s_waitcnt vmcnt(0) -; GFX11-NEXT: v_sub_co_u32 v2, vcc_lo, v0, v2 -; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1) -; GFX11-NEXT: v_sub_co_ci_u32_e64 v3, null, v1, v3, vcc_lo -; GFX11-NEXT: v_cmp_gt_u64_e32 vcc_lo, v[2:3], v[0:1] -; GFX11-NEXT: v_cndmask_b32_e64 v0, 0, 1, vcc_lo +; GFX11-NEXT: v_sub_co_u32 v0, vcc_lo, v0, v2 +; GFX11-NEXT: v_sub_co_ci_u32_e32 v1, vcc_lo, v1, v3, vcc_lo +; GFX11-NEXT: v_cndmask_b32_e64 v2, 0, 1, vcc_lo ; GFX11-NEXT: s_clause 0x1 -; GFX11-NEXT: global_store_b64 v4, v[2:3], s[0:1] -; GFX11-NEXT: global_store_b8 v4, v0, s[2:3] +; GFX11-NEXT: global_store_b64 v4, v[0:1], s[0:1] +; GFX11-NEXT: global_store_b8 v4, v2, s[2:3] ; GFX11-NEXT: s_endpgm %tid = call i32 @llvm.amdgcn.workitem.id.x() %tid.ext = sext i32 %tid to i64 diff --git a/llvm/test/CodeGen/AMDGPU/usubsat.ll b/llvm/test/CodeGen/AMDGPU/usubsat.ll index 90491a07289a0..3ddb2f02c48fe 100644 --- a/llvm/test/CodeGen/AMDGPU/usubsat.ll +++ b/llvm/test/CodeGen/AMDGPU/usubsat.ll @@ -730,52 +730,38 @@ define i64 @v_usubsat_i64(i64 %lhs, i64 %rhs) { ; GFX6-LABEL: v_usubsat_i64: ; GFX6: ; %bb.0: ; GFX6-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) -; GFX6-NEXT: v_sub_i32_e32 v2, vcc, v0, v2 -; GFX6-NEXT: v_subb_u32_e32 v3, vcc, v1, v3, vcc -; GFX6-NEXT: v_cmp_gt_u64_e32 vcc, v[2:3], v[0:1] -; GFX6-NEXT: v_cndmask_b32_e64 v0, v2, 0, vcc -; GFX6-NEXT: v_cndmask_b32_e64 v1, v3, 0, vcc +; GFX6-NEXT: v_sub_i32_e32 v0, vcc, v0, v2 +; GFX6-NEXT: v_subb_u32_e32 v1, vcc, v1, v3, vcc +; GFX6-NEXT: v_cndmask_b32_e64 v0, v0, 0, vcc +; GFX6-NEXT: v_cndmask_b32_e64 v1, v1, 0, vcc ; GFX6-NEXT: s_setpc_b64 s[30:31] ; ; GFX8-LABEL: v_usubsat_i64: ; GFX8: ; %bb.0: ; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) -; GFX8-NEXT: v_sub_u32_e32 v2, vcc, v0, v2 -; GFX8-NEXT: v_subb_u32_e32 v3, vcc, v1, v3, vcc -; GFX8-NEXT: v_cmp_gt_u64_e32 vcc, v[2:3], v[0:1] -; GFX8-NEXT: v_cndmask_b32_e64 v0, v2, 0, vcc -; GFX8-NEXT: v_cndmask_b32_e64 v1, v3, 0, vcc +; GFX8-NEXT: v_sub_u32_e32 v0, vcc, v0, v2 +; GFX8-NEXT: v_subb_u32_e32 v1, vcc, v1, v3, vcc +; GFX8-NEXT: v_cndmask_b32_e64 v0, v0, 0, vcc +; GFX8-NEXT: v_cndmask_b32_e64 v1, v1, 0, vcc ; GFX8-NEXT: s_setpc_b64 s[30:31] ; ; GFX9-LABEL: v_usubsat_i64: ; GFX9: ; %bb.0: ; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) -; GFX9-NEXT: v_sub_co_u32_e32 v2, vcc, v0, v2 -; GFX9-NEXT: v_subb_co_u32_e32 v3, vcc, v1, v3, vcc -; GFX9-NEXT: v_cmp_gt_u64_e32 vcc, v[2:3], v[0:1] -; GFX9-NEXT: v_cndmask_b32_e64 v0, v2, 0, vcc -; GFX9-NEXT: v_cndmask_b32_e64 v1, v3, 0, vcc +; GFX9-NEXT: v_sub_co_u32_e32 v0, vcc, v0, v2 +; GFX9-NEXT: v_subb_co_u32_e32 v1, vcc, v1, v3, vcc +; GFX9-NEXT: v_cndmask_b32_e64 v0, v0, 0, vcc +; GFX9-NEXT: v_cndmask_b32_e64 v1, v1, 0, vcc ; GFX9-NEXT: s_setpc_b64 s[30:31] ; -; GFX10-LABEL: v_usubsat_i64: -; GFX10: ; %bb.0: -; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) -; GFX10-NEXT: v_sub_co_u32 v2, vcc_lo, v0, v2 -; GFX10-NEXT: v_sub_co_ci_u32_e32 v3, vcc_lo, v1, v3, vcc_lo -; GFX10-NEXT: v_cmp_gt_u64_e32 vcc_lo, v[2:3], v[0:1] -; GFX10-NEXT: v_cndmask_b32_e64 v0, v2, 0, vcc_lo -; GFX10-NEXT: v_cndmask_b32_e64 v1, v3, 0, vcc_lo -; GFX10-NEXT: s_setpc_b64 s[30:31] -; -; GFX11-LABEL: v_usubsat_i64: -; GFX11: ; %bb.0: -; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) -; GFX11-NEXT: v_sub_co_u32 v2, vcc_lo, v0, v2 -; GFX11-NEXT: v_sub_co_ci_u32_e64 v3, null, v1, v3, vcc_lo -; GFX11-NEXT: v_cmp_gt_u64_e32 vcc_lo, v[2:3], v[0:1] -; GFX11-NEXT: v_cndmask_b32_e64 v0, v2, 0, vcc_lo -; GFX11-NEXT: v_cndmask_b32_e64 v1, v3, 0, vcc_lo -; GFX11-NEXT: s_setpc_b64 s[30:31] +; GFX10PLUS-LABEL: v_usubsat_i64: +; GFX10PLUS: ; %bb.0: +; GFX10PLUS-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX10PLUS-NEXT: v_sub_co_u32 v0, vcc_lo, v0, v2 +; GFX10PLUS-NEXT: v_sub_co_ci_u32_e32 v1, vcc_lo, v1, v3, vcc_lo +; GFX10PLUS-NEXT: v_cndmask_b32_e64 v0, v0, 0, vcc_lo +; GFX10PLUS-NEXT: v_cndmask_b32_e64 v1, v1, 0, vcc_lo +; GFX10PLUS-NEXT: s_setpc_b64 s[30:31] %result = call i64 @llvm.usub.sat.i64(i64 %lhs, i64 %rhs) ret i64 %result }