# RUN: llc -mtriple=amdgcn--amdhsa -mcpu=gfx900 -verify-machineinstrs -run-pass si-fold-operands,dead-mi-elimination -o - %s | FileCheck -enable-var-scope -check-prefix=GCN %s ... # GCN-LABEL: name: s_fold_and_imm_regimm_32{{$}} # GCN: %10:vgpr_32 = V_MOV_B32_e32 1543, implicit $exec # GCN: BUFFER_STORE_DWORD_OFFSET killed %10, name: s_fold_and_imm_regimm_32 tracksRegLiveness: true liveins: - { reg: '$sgpr0_sgpr1', virtual-reg: '%0' } frameInfo: isFrameAddressTaken: false isReturnAddressTaken: false hasStackMap: false hasPatchPoint: false stackSize: 0 offsetAdjustment: 0 maxAlignment: 0 adjustsStack: false hasCalls: false maxCallFrameSize: 0 hasOpaqueSPAdjustment: false hasVAStart: false hasMustTailInVarArgFunc: false body: | bb.0: liveins: $sgpr0_sgpr1 %0:sgpr_64 = COPY $sgpr0_sgpr1 %1:sreg_64_xexec = S_LOAD_DWORDX2_IMM %0, 36, 0 %2:sreg_32_xm0 = COPY %1.sub1 %3:sreg_32_xm0 = COPY %1.sub0 %4:sreg_32_xm0 = S_MOV_B32 61440 %5:sreg_32_xm0 = S_MOV_B32 -1 %6:sgpr_128 = REG_SEQUENCE killed %2, %subreg.hi16, killed %3, %subreg.lo16, killed %4, %subreg.sub0, killed %5, %subreg.sub0_sub1 %7:sreg_32_xm0 = S_MOV_B32 1234567 %8:sreg_32_xm0 = S_MOV_B32 9999 %9:sreg_32_xm0 = S_AND_B32 killed %7, killed %8, implicit-def dead $scc %10:vgpr_32 = COPY %9 BUFFER_STORE_DWORD_OFFSET killed %10, killed %6, 0, 0, 0, 0, 0, implicit $exec S_ENDPGM 0 ... --- # GCN-LABEL: name: v_fold_and_imm_regimm_32{{$}} # GCN: [[VAL0:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 646, implicit $exec # GCN: FLAT_STORE_DWORD %10, [[VAL0]], # GCN: [[VAL1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 646, implicit $exec # GCN: FLAT_STORE_DWORD %10, [[VAL1]], # GCN: [[VAL2:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 646, implicit $exec # GCN: FLAT_STORE_DWORD %10, [[VAL2]], # GCN: [[VAL3:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 1234567, implicit $exec # GCN: FLAT_STORE_DWORD %10, [[VAL3]], # GCN: [[VAL4:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 63, implicit $exec # GCN: FLAT_STORE_DWORD %10, [[VAL4]], name: v_fold_and_imm_regimm_32 tracksRegLiveness: true liveins: - { reg: '$sgpr0_sgpr1', virtual-reg: '%0' } - { reg: '$vgpr0', virtual-reg: '%3' } body: | bb.0: liveins: $sgpr0_sgpr1, $vgpr0 %3:vgpr_32 = COPY $vgpr0 %0:sgpr_64 = COPY $sgpr0_sgpr1 %4:sreg_64_xexec = S_LOAD_DWORDX2_IMM %0, 36, 0 %13:vgpr_32 = V_ASHRREV_I32_e64 31, %3, implicit $exec %14:vreg_64 = REG_SEQUENCE %3, %subreg.hi16, %13, %subreg.lo16 %15:vreg_64 = V_LSHLREV_B64_e64 2, killed %14, implicit $exec %5:sreg_32_xm0 = COPY %4.sub1 %20:vgpr_32 = V_ADD_CO_U32_e32 %4.sub0, %15.sub0, implicit-def $vcc, implicit $exec %18:vgpr_32 = COPY killed %5 %17:vgpr_32 = V_ADDC_U32_e32 %15.sub1, %18, implicit-def $vcc, implicit $vcc, implicit $exec %19:vreg_64 = REG_SEQUENCE %20, %subreg.hi16, killed %17, %subreg.lo16 %6:vgpr_32 = V_MOV_B32_e32 982, implicit $exec %8:sreg_32_xm0 = S_MOV_B32 1234567 %16:vgpr_32 = V_MOV_B32_e32 63, implicit $exec %9:vgpr_32 = V_AND_B32_e64 %8, %6, implicit $exec FLAT_STORE_DWORD %19, %9, 0, 0, implicit $exec, implicit $flat_scr %10:vgpr_32 = V_AND_B32_e64 %6, %8, implicit $exec FLAT_STORE_DWORD %19, %10, 0, 0, implicit $exec, implicit $flat_scr %11:vgpr_32 = V_AND_B32_e32 %8, %6, implicit $exec FLAT_STORE_DWORD %19, %11, 0, 0, implicit $exec, implicit $flat_scr %12:vgpr_32 = V_AND_B32_e64 %8, %8, implicit $exec FLAT_STORE_DWORD %19, %12, 0, 0, implicit $exec, implicit $flat_scr %13:vgpr_32 = V_AND_B32_e64 %16, %16, implicit $exec FLAT_STORE_DWORD %19, %13, 0, 0, implicit $exec, implicit $flat_scr S_ENDPGM 0 ... --- # GCN-LABEL: name: s_fold_shl_imm_regimm_32{{$}} # GCN: [[VAL:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 4096, implicit $exec # GCN: BUFFER_STORE_DWORD_OFFSET killed [[VAL]], name: s_fold_shl_imm_regimm_32 tracksRegLiveness: true liveins: - { reg: '$sgpr0_sgpr1', virtual-reg: '%0' } body: | bb.0: liveins: $sgpr0_sgpr1 %0:sgpr_64 = COPY $sgpr0_sgpr1 %4:sreg_64_xexec = S_LOAD_DWORDX2_IMM %0, 36, 0 %5:sreg_32_xm0_xexec = S_MOV_B32 1 %6:sreg_32_xm0 = COPY %4.sub1 %7:sreg_32_xm0 = COPY %4.sub0 %8:sreg_32_xm0 = S_MOV_B32 61440 %9:sreg_32_xm0 = S_MOV_B32 -1 %10:sgpr_128 = REG_SEQUENCE killed %7, %subreg.hi16, killed %6, %subreg.lo16, killed %9, %subreg.sub0, killed %8, %subreg.sub0_sub1 %12:sreg_32_xm0 = S_LSHL_B32 killed %5, 12, implicit-def dead $scc %13:vgpr_32 = COPY %12 BUFFER_STORE_DWORD_OFFSET killed %13, killed %10, 0, 0, 0, 0, 0, implicit $exec S_ENDPGM 0 ... --- # GCN-LABEL: name: s_fold_ashr_imm_regimm_32{{$}} # GCN: [[VAL:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 243, implicit $exec # GCN: BUFFER_STORE_DWORD_OFFSET killed [[VAL]], killed %7, name: s_fold_ashr_imm_regimm_32 tracksRegLiveness: true liveins: - { reg: '$sgpr0_sgpr1', virtual-reg: '%0' } body: | bb.0: liveins: $sgpr0_sgpr1 %0:sgpr_64 = COPY $sgpr0_sgpr1 %2:sreg_64_xexec = S_LOAD_DWORDX2_IMM %0, 36, 0 %3:sreg_32_xm0_xexec = S_MOV_B32 999123 %4:sreg_32_xm0 = COPY %2.sub1 %5:sreg_32_xm0 = COPY %2.sub0 %6:sreg_32_xm0 = S_MOV_B32 61440 %7:sreg_32_xm0 = S_MOV_B32 -1 %8:sgpr_128 = REG_SEQUENCE killed %5, %subreg.hi16, killed %4, %subreg.lo16, killed %7, %subreg.sub0, killed %6, %subreg.sub0_sub1 %10:sreg_32_xm0 = S_ASHR_I32 killed %3, 12, implicit-def dead $scc %11:vgpr_32 = COPY %10 BUFFER_STORE_DWORD_OFFSET killed %11, killed %8, 0, 0, 0, 0, 0, implicit $exec S_ENDPGM 0 ... # GCN-LABEL: name: v_fold_ashr_imm_regimm_32{{$}} # GCN: [[VAL0:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 3903258, implicit $exec # GCN: FLAT_STORE_DWORD %10, [[VAL0]], # GCN: [[VAL1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 62452139, implicit $exec # GCN: FLAT_STORE_DWORD %10, [[VAL1]], # GCN: [[VAL2:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 1678031, implicit $exec # GCN: FLAT_STORE_DWORD %10, [[VAL2]], # GCN: [[VAL3:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 3, implicit $exec # GCN: FLAT_STORE_DWORD %10, [[VAL3]], # GCN: [[VAL4:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 -1, implicit $exec # GCN: FLAT_STORE_DWORD %10, [[VAL4]], # GCN: [[VAL5:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 62500, implicit $exec # GCN: FLAT_STORE_DWORD %10, [[VAL5]], # GCN: [[VAL6:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 500000, implicit $exec # GCN: FLAT_STORE_DWORD %10, [[VAL6]], # GCN: [[VAL7:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 1920, implicit $exec # GCN: FLAT_STORE_DWORD %10, [[VAL7]], # GCN: [[VAL8:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 487907, implicit $exec # GCN: FLAT_STORE_DWORD %10, [[VAL8]], # GCN: [[VAL9:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 -1, implicit $exec # GCN: FLAT_STORE_DWORD %10, [[VAL9]], name: v_fold_ashr_imm_regimm_32 tracksRegLiveness: true liveins: - { reg: '$sgpr0_sgpr1', virtual-reg: '%0' } - { reg: '$vgpr0', virtual-reg: '%2' } body: | bb.0: liveins: $sgpr0_sgpr1, $vgpr0 %2:vgpr_32 = COPY $vgpr0 %0:sgpr_64 = COPY $sgpr0_sgpr1 %3:sreg_64_xexec = S_LOAD_DWORDX2_IMM %0, 36, 0 %15:vgpr_32 = V_ASHRREV_I32_e64 31, %2, implicit $exec %16:vreg_64 = REG_SEQUENCE %2, %subreg.hi16, %15, %subreg.lo16 %17:vreg_64 = V_LSHLREV_B64_e64 2, killed %16, implicit $exec %9:sreg_32_xm0 = COPY %3.sub1 %21:vgpr_32 = V_ADD_CO_U32_e32 %3.sub0, %17.sub0, implicit-def $vcc, implicit $exec %19:vgpr_32 = COPY killed %9 %18:vgpr_32 = V_ADDC_U32_e32 %17.sub1, %19, implicit-def $vcc, implicit $vcc, implicit $exec %20:vreg_64 = REG_SEQUENCE %21, %subreg.hi16, killed %18, %subreg.lo16 %10:vgpr_32 = V_MOV_B32_e32 999234234, implicit $exec %24:vgpr_32 = V_MOV_B32_e32 3871, implicit $exec %6:vgpr_32 = V_MOV_B32_e32 1000000, implicit $exec %7:sreg_32_xm0 = S_MOV_B32 13424252 %8:sreg_32_xm0 = S_MOV_B32 4 %27:sreg_32_xm0 = S_MOV_B32 -4 %29:sreg_32_xm0 = S_MOV_B32 1 %30:sreg_32_xm0 = S_MOV_B32 3841 %31:vgpr_32 = V_MOV_B32_e32 3841, implicit $exec %32:vgpr_32 = V_MOV_B32_e32 2, implicit $exec %11:vgpr_32 = V_ASHRREV_I32_e64 8, %10, implicit $exec FLAT_STORE_DWORD %20, %11, 0, 0, implicit $exec, implicit $flat_scr %12:vgpr_32 = V_ASHRREV_I32_e64 %8, %10, implicit $exec FLAT_STORE_DWORD %20, %12, 0, 0, implicit $exec, implicit $flat_scr %13:vgpr_32 = V_ASHR_I32_e64 %7, 3, implicit $exec FLAT_STORE_DWORD %20, %13, 0, 0, implicit $exec, implicit $flat_scr %14:vgpr_32 = V_ASHR_I32_e64 7, %29, implicit $exec FLAT_STORE_DWORD %20, %14, 0, 0, implicit $exec, implicit $flat_scr %15:vgpr_32 = V_ASHR_I32_e64 %27, %24, implicit $exec FLAT_STORE_DWORD %20, %15, 0, 0, implicit $exec, implicit $flat_scr %22:vgpr_32 = V_ASHR_I32_e64 %6, 4, implicit $exec FLAT_STORE_DWORD %20, %22, 0, 0, implicit $exec, implicit $flat_scr %23:vgpr_32 = V_ASHR_I32_e64 %6, %30, implicit $exec FLAT_STORE_DWORD %20, %23, 0, 0, implicit $exec, implicit $flat_scr %25:vgpr_32 = V_ASHR_I32_e32 %31, %31, implicit $exec FLAT_STORE_DWORD %20, %25, 0, 0, implicit $exec, implicit $flat_scr %26:vgpr_32 = V_ASHRREV_I32_e32 11, %10, implicit $exec FLAT_STORE_DWORD %20, %26, 0, 0, implicit $exec, implicit $flat_scr %28:vgpr_32 = V_ASHR_I32_e32 %27, %32, implicit $exec FLAT_STORE_DWORD %20, %28, 0, 0, implicit $exec, implicit $flat_scr S_ENDPGM 0 ... --- # GCN-LABEL: name: s_fold_lshr_imm_regimm_32{{$}} # GCN: [[VAL:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 1048332, implicit $exec # GCN: BUFFER_STORE_DWORD_OFFSET killed [[VAL]], killed %7, name: s_fold_lshr_imm_regimm_32 tracksRegLiveness: true liveins: - { reg: '$sgpr0_sgpr1', virtual-reg: '%0' } body: | bb.0: liveins: $sgpr0_sgpr1 %0:sgpr_64 = COPY $sgpr0_sgpr1 %2:sreg_64_xexec = S_LOAD_DWORDX2_IMM %0, 36, 0 %3:sreg_32_xm0_xexec = S_MOV_B32 -999123 %4:sreg_32_xm0 = COPY %2.sub1 %5:sreg_32_xm0 = COPY %2.sub0 %6:sreg_32_xm0 = S_MOV_B32 61440 %7:sreg_32_xm0 = S_MOV_B32 -1 %8:sgpr_128 = REG_SEQUENCE killed %5, %subreg.hi16, killed %4, %subreg.lo16, killed %7, %subreg.sub0, killed %6, %subreg.sub0_sub1 %10:sreg_32_xm0 = S_LSHR_B32 killed %3, 12, implicit-def dead $scc %11:vgpr_32 = COPY %10 BUFFER_STORE_DWORD_OFFSET killed %11, killed %8, 0, 0, 0, 0, 0, implicit $exec S_ENDPGM 0 ... --- # GCN-LABEL: name: v_fold_lshr_imm_regimm_32{{$}} # GCN: [[VAL0:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 3903258, implicit $exec # GCN: FLAT_STORE_DWORD %10, [[VAL0]], # GCN: [[VAL1:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 62452139, implicit $exec # GCN: FLAT_STORE_DWORD %10, [[VAL1]], # GCN: [[VAL2:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 1678031, implicit $exec # GCN: FLAT_STORE_DWORD %10, [[VAL2]], # GCN: [[VAL3:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 3, implicit $exec # GCN: FLAT_STORE_DWORD %10, [[VAL3]], # GCN: [[VAL4:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 1, implicit $exec # GCN: FLAT_STORE_DWORD %10, [[VAL4]], # GCN: [[VAL5:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 62500, implicit $exec # GCN: FLAT_STORE_DWORD %10, [[VAL5]], # GCN: [[VAL6:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 500000, implicit $exec # GCN: FLAT_STORE_DWORD %10, [[VAL6]], # GCN: [[VAL7:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 1920, implicit $exec # GCN: FLAT_STORE_DWORD %10, [[VAL7]], # GCN: [[VAL8:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 487907, implicit $exec # GCN: FLAT_STORE_DWORD %10, [[VAL8]], # GCN: [[VAL9:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 1073741823, implicit $exec # GCN: FLAT_STORE_DWORD %10, [[VAL9]], name: v_fold_lshr_imm_regimm_32 tracksRegLiveness: true liveins: - { reg: '$sgpr0_sgpr1', virtual-reg: '%0' } - { reg: '$vgpr0', virtual-reg: '%2' } body: | bb.0: liveins: $sgpr0_sgpr1, $vgpr0 %2:vgpr_32 = COPY $vgpr0 %0:sgpr_64 = COPY $sgpr0_sgpr1 %3:sreg_64_xexec = S_LOAD_DWORDX2_IMM %0, 36, 0 %15:vgpr_32 = V_ASHRREV_I32_e64 31, %2, implicit $exec %16:vreg_64 = REG_SEQUENCE %2, %subreg.hi16, %15, %subreg.lo16 %17:vreg_64 = V_LSHLREV_B64_e64 2, killed %16, implicit $exec %9:sreg_32_xm0 = COPY %3.sub1 %21:vgpr_32 = V_ADD_CO_U32_e32 %3.sub0, %17.sub0, implicit-def $vcc, implicit $exec %19:vgpr_32 = COPY killed %9 %18:vgpr_32 = V_ADDC_U32_e32 %17.sub1, %19, implicit-def $vcc, implicit $vcc, implicit $exec %20:vreg_64 = REG_SEQUENCE %21, %subreg.hi16, killed %18, %subreg.lo16 %10:vgpr_32 = V_MOV_B32_e32 999234234, implicit $exec %24:vgpr_32 = V_MOV_B32_e32 3871, implicit $exec %6:vgpr_32 = V_MOV_B32_e32 1000000, implicit $exec %7:sreg_32_xm0 = S_MOV_B32 13424252 %8:sreg_32_xm0 = S_MOV_B32 4 %27:sreg_32_xm0 = S_MOV_B32 -4 %29:sreg_32_xm0 = S_MOV_B32 1 %30:sreg_32_xm0 = S_MOV_B32 3841 %31:vgpr_32 = V_MOV_B32_e32 3841, implicit $exec %32:vgpr_32 = V_MOV_B32_e32 2, implicit $exec %11:vgpr_32 = V_LSHRREV_B32_e64 8, %10, implicit $exec FLAT_STORE_DWORD %20, %11, 0, 0, implicit $exec, implicit $flat_scr %12:vgpr_32 = V_LSHRREV_B32_e64 %8, %10, implicit $exec FLAT_STORE_DWORD %20, %12, 0, 0, implicit $exec, implicit $flat_scr %13:vgpr_32 = V_LSHR_B32_e64 %7, 3, implicit $exec FLAT_STORE_DWORD %20, %13, 0, 0, implicit $exec, implicit $flat_scr %14:vgpr_32 = V_LSHR_B32_e64 7, %29, implicit $exec FLAT_STORE_DWORD %20, %14, 0, 0, implicit $exec, implicit $flat_scr %15:vgpr_32 = V_LSHR_B32_e64 %27, %24, implicit $exec FLAT_STORE_DWORD %20, %15, 0, 0, implicit $exec, implicit $flat_scr %22:vgpr_32 = V_LSHR_B32_e64 %6, 4, implicit $exec FLAT_STORE_DWORD %20, %22, 0, 0, implicit $exec, implicit $flat_scr %23:vgpr_32 = V_LSHR_B32_e64 %6, %30, implicit $exec FLAT_STORE_DWORD %20, %23, 0, 0, implicit $exec, implicit $flat_scr %25:vgpr_32 = V_LSHR_B32_e32 %31, %31, implicit $exec FLAT_STORE_DWORD %20, %25, 0, 0, implicit $exec, implicit $flat_scr %26:vgpr_32 = V_LSHRREV_B32_e32 11, %10, implicit $exec FLAT_STORE_DWORD %20, %26, 0, 0, implicit $exec, implicit $flat_scr %28:vgpr_32 = V_LSHR_B32_e32 %27, %32, implicit $exec FLAT_STORE_DWORD %20, %28, 0, 0, implicit $exec, implicit $flat_scr S_ENDPGM 0 ... --- # There is only an undef use operand for %1, so there is no # corresponding defining instruction # GCN-LABEL: name: undefined_vreg_operand{{$}} # GCN: bb.0 # GCN-NEXT: FLAT_STORE_DWORD undef %3:vreg_64, undef %2:vgpr_32, # GCN-NEXT: S_ENDPGM 0 name: undefined_vreg_operand tracksRegLiveness: true body: | bb.0: %0:vgpr_32 = V_MOV_B32_e32 0, implicit $exec %2:vgpr_32 = V_XOR_B32_e64 killed %0, undef %1:vgpr_32, implicit $exec FLAT_STORE_DWORD undef %3:vreg_64, %2, 0, 0, implicit $exec, implicit $flat_scr S_ENDPGM 0 ... --- # Make sure there is no crash if one of the operands is a physical register # GCN-LABEL: name: constant_fold_physreg_op{{$}} # GCN: %3:sreg_64 = S_AND_B64 $exec, 0, implicit-def dead $scc name: constant_fold_physreg_op tracksRegLiveness: true body: | bb.0: successors: %bb.1, %bb.3 liveins: $vgpr0, $sgpr4_sgpr5 %19:sreg_64 = IMPLICIT_DEF %0:sreg_64 = SI_IF killed %19, %bb.3, implicit-def dead $exec, implicit-def dead $scc, implicit $exec S_BRANCH %bb.1 bb.1: %6:sreg_64 = S_MOV_B64 0 %7:sreg_64 = S_AND_B64 $exec, killed %6, implicit-def dead $scc $vcc = COPY %7 bb.3: liveins: $vcc SI_END_CF %0, implicit-def dead $exec, implicit-def dead $scc, implicit $exec S_ENDPGM 0, implicit $vcc ... --- # GCN-LABEL: name: s_fold_andn2_imm_regimm_32{{$}} # GCN: [[VAL:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 1233024, implicit $exec # GCN: BUFFER_STORE_DWORD_OFFSET killed [[VAL]], name: s_fold_andn2_imm_regimm_32 tracksRegLiveness: true liveins: - { reg: '$sgpr0_sgpr1', virtual-reg: '%0' } body: | bb.0: liveins: $sgpr0_sgpr1 %0:sgpr_64 = COPY $sgpr0_sgpr1 %1:sreg_64_xexec = S_LOAD_DWORDX2_IMM %0, 36, 0 %2:sreg_32_xm0 = COPY %1.sub1 %3:sreg_32_xm0 = COPY %1.sub0 %4:sreg_32_xm0 = S_MOV_B32 61440 %5:sreg_32_xm0 = S_MOV_B32 -1 %6:sgpr_128 = REG_SEQUENCE killed %2, %subreg.sub0, killed %3, %subreg.sub1, killed %4, %subreg.sub2, killed %5, %subreg.sub3 %7:sreg_32_xm0 = S_MOV_B32 1234567 %8:sreg_32_xm0 = S_MOV_B32 9999 %9:sreg_32_xm0 = S_ANDN2_B32 killed %7, killed %8, implicit-def dead $scc %10:vgpr_32 = COPY %9 BUFFER_STORE_DWORD_OFFSET killed %10, killed %6, 0, 0, 0, 0, 0, implicit $exec S_ENDPGM 0 ... --- # GCN-LABEL: name: s_fold_or_imm_regimm_32{{$}} # GCN: %10:vgpr_32 = V_MOV_B32_e32 1243023, implicit $exec # GCN: BUFFER_STORE_DWORD_OFFSET killed %10, name: s_fold_or_imm_regimm_32 alignment: 0 exposesReturnsTwice: false legalized: false regBankSelected: false selected: false tracksRegLiveness: true liveins: - { reg: '$sgpr0_sgpr1', virtual-reg: '%0' } body: | bb.0: liveins: $sgpr0_sgpr1 %0:sgpr_64 = COPY $sgpr0_sgpr1 %1:sreg_64_xexec = S_LOAD_DWORDX2_IMM %0, 36, 0 %2:sreg_32_xm0 = COPY %1.sub1 %3:sreg_32_xm0 = COPY %1.sub0 %4:sreg_32_xm0 = S_MOV_B32 61440 %5:sreg_32_xm0 = S_MOV_B32 -1 %6:sgpr_128 = REG_SEQUENCE killed %2, %subreg.sub0, killed %3, %subreg.sub1, killed %4, %subreg.sub2, killed %5, %subreg.sub3 %7:sreg_32_xm0 = S_MOV_B32 1234567 %8:sreg_32_xm0 = S_MOV_B32 9999 %9:sreg_32_xm0 = S_OR_B32 killed %7, killed %8, implicit-def dead $scc %10:vgpr_32 = COPY %9 BUFFER_STORE_DWORD_OFFSET killed %10, killed %6, 0, 0, 0, 0, 0, implicit $exec S_ENDPGM 0 ... --- # GCN-LABEL: name: v_fold_or_imm_regimm_32{{$}} # GCN: %14:vgpr_32 = V_MOV_B32_e32 1234903, implicit $exec # GCN: FLAT_STORE_DWORD %10, %14, # GCN: %15:vgpr_32 = V_MOV_B32_e32 1234903, implicit $exec # GCN: FLAT_STORE_DWORD %10, %15 # GCN: %16:vgpr_32 = V_MOV_B32_e32 1234903, implicit $exec # GCN: FLAT_STORE_DWORD %10, %16, # GCN: %17:vgpr_32 = V_MOV_B32_e32 1234567, implicit $exec # GCN: FLAT_STORE_DWORD %10, %17, # GCN: %3:vgpr_32 = V_MOV_B32_e32 63, implicit $exec # GCN: FLAT_STORE_DWORD %10, %3, name: v_fold_or_imm_regimm_32 alignment: 0 exposesReturnsTwice: false legalized: false regBankSelected: false selected: false tracksRegLiveness: true liveins: - { reg: '$sgpr0_sgpr1', virtual-reg: '%0' } - { reg: '$vgpr0', virtual-reg: '%3' } body: | bb.0: liveins: $sgpr0_sgpr1, $vgpr0 %3:vgpr_32 = COPY $vgpr0 %0:sgpr_64 = COPY $sgpr0_sgpr1 %4:sreg_64_xexec = S_LOAD_DWORDX2_IMM %0, 36, 0 %13:vgpr_32 = V_ASHRREV_I32_e64 31, %3, implicit $exec %14:vreg_64 = REG_SEQUENCE %3, %subreg.sub0, %13, %subreg.sub1 %15:vreg_64 = V_LSHLREV_B64_e64 2, killed %14, implicit $exec %5:sreg_32_xm0 = COPY %4.sub1 %20:vgpr_32 = V_ADD_CO_U32_e32 %4.sub0, %15.sub0, implicit-def $vcc, implicit $exec %18:vgpr_32 = COPY killed %5 %17:vgpr_32 = V_ADDC_U32_e32 %15.sub1, %18, implicit-def $vcc, implicit $vcc, implicit $exec %19:vreg_64 = REG_SEQUENCE %20, %subreg.sub0, killed %17, %subreg.sub1 %6:vgpr_32 = V_MOV_B32_e32 982, implicit $exec %8:sreg_32_xm0 = S_MOV_B32 1234567 %16:vgpr_32 = V_MOV_B32_e32 63, implicit $exec %9:vgpr_32 = V_OR_B32_e64 %8, %6, implicit $exec FLAT_STORE_DWORD %19, %9, 0, 0, implicit $exec, implicit $flat_scr %10:vgpr_32 = V_OR_B32_e64 %6, %8, implicit $exec FLAT_STORE_DWORD %19, %10, 0, 0, implicit $exec, implicit $flat_scr %11:vgpr_32 = V_OR_B32_e32 %8, %6, implicit $exec FLAT_STORE_DWORD %19, %11, 0, 0, implicit $exec, implicit $flat_scr %12:vgpr_32 = V_OR_B32_e64 %8, %8, implicit $exec FLAT_STORE_DWORD %19, %12, 0, 0, implicit $exec, implicit $flat_scr %13:vgpr_32 = V_OR_B32_e64 %16, %16, implicit $exec FLAT_STORE_DWORD %19, %13, 0, 0, implicit $exec, implicit $flat_scr S_ENDPGM 0 ... --- # GCN-LABEL: name: s_fold_orn2_imm_regimm_32{{$}} # GCN: %10:vgpr_32 = V_MOV_B32_e32 -8457, implicit $exec # GCN: BUFFER_STORE_DWORD_OFFSET killed %10, name: s_fold_orn2_imm_regimm_32 alignment: 0 exposesReturnsTwice: false legalized: false regBankSelected: false selected: false tracksRegLiveness: true liveins: - { reg: '$sgpr0_sgpr1', virtual-reg: '%0' } body: | bb.0: liveins: $sgpr0_sgpr1 %0:sgpr_64 = COPY $sgpr0_sgpr1 %1:sreg_64_xexec = S_LOAD_DWORDX2_IMM %0, 36, 0 %2:sreg_32_xm0 = COPY %1.sub1 %3:sreg_32_xm0 = COPY %1.sub0 %4:sreg_32_xm0 = S_MOV_B32 61440 %5:sreg_32_xm0 = S_MOV_B32 -1 %6:sgpr_128 = REG_SEQUENCE killed %2, %subreg.sub0, killed %3, %subreg.sub1, killed %4, %subreg.sub2, killed %5, %subreg.sub3 %7:sreg_32_xm0 = S_MOV_B32 1234567 %8:sreg_32_xm0 = S_MOV_B32 9999 %9:sreg_32_xm0 = S_ORN2_B32 killed %7, killed %8, implicit-def dead $scc %10:vgpr_32 = COPY %9 BUFFER_STORE_DWORD_OFFSET killed %10, killed %6, 0, 0, 0, 0, 0, implicit $exec S_ENDPGM 0 ... --- # GCN-LABEL: name: s_fold_nand_imm_regimm_32{{$}} # GCN: %10:vgpr_32 = V_MOV_B32_e32 -1544, implicit $exec # GCN: BUFFER_STORE_DWORD_OFFSET killed %10, name: s_fold_nand_imm_regimm_32 alignment: 0 exposesReturnsTwice: false legalized: false regBankSelected: false selected: false tracksRegLiveness: true liveins: - { reg: '$sgpr0_sgpr1', virtual-reg: '%0' } body: | bb.0: liveins: $sgpr0_sgpr1 %0:sgpr_64 = COPY $sgpr0_sgpr1 %1:sreg_64_xexec = S_LOAD_DWORDX2_IMM %0, 36, 0 %2:sreg_32_xm0 = COPY %1.sub1 %3:sreg_32_xm0 = COPY %1.sub0 %4:sreg_32_xm0 = S_MOV_B32 61440 %5:sreg_32_xm0 = S_MOV_B32 -1 %6:sgpr_128 = REG_SEQUENCE killed %2, %subreg.sub0, killed %3, %subreg.sub1, killed %4, %subreg.sub2, killed %5, %subreg.sub3 %7:sreg_32_xm0 = S_MOV_B32 1234567 %8:sreg_32_xm0 = S_MOV_B32 9999 %9:sreg_32_xm0 = S_NAND_B32 killed %7, killed %8, implicit-def dead $scc %10:vgpr_32 = COPY %9 BUFFER_STORE_DWORD_OFFSET killed %10, killed %6, 0, 0, 0, 0, 0, implicit $exec S_ENDPGM 0 ... --- # GCN-LABEL: name: s_fold_nor_imm_regimm_32{{$}} # GCN: %10:vgpr_32 = V_MOV_B32_e32 -1243024, implicit $exec # GCN: BUFFER_STORE_DWORD_OFFSET killed %10, name: s_fold_nor_imm_regimm_32 alignment: 0 exposesReturnsTwice: false legalized: false regBankSelected: false selected: false tracksRegLiveness: true liveins: - { reg: '$sgpr0_sgpr1', virtual-reg: '%0' } body: | bb.0: liveins: $sgpr0_sgpr1 %0:sgpr_64 = COPY $sgpr0_sgpr1 %1:sreg_64_xexec = S_LOAD_DWORDX2_IMM %0, 36, 0 %2:sreg_32_xm0 = COPY %1.sub1 %3:sreg_32_xm0 = COPY %1.sub0 %4:sreg_32_xm0 = S_MOV_B32 61440 %5:sreg_32_xm0 = S_MOV_B32 -1 %6:sgpr_128 = REG_SEQUENCE killed %2, %subreg.sub0, killed %3, %subreg.sub1, killed %4, %subreg.sub2, killed %5, %subreg.sub3 %7:sreg_32_xm0 = S_MOV_B32 1234567 %8:sreg_32_xm0 = S_MOV_B32 9999 %9:sreg_32_xm0 = S_NOR_B32 killed %7, killed %8, implicit-def dead $scc %10:vgpr_32 = COPY %9 BUFFER_STORE_DWORD_OFFSET killed %10, killed %6, 0, 0, 0, 0, 0, implicit $exec S_ENDPGM 0 ... --- # GCN-LABEL: name: s_fold_xnor_imm_regimm_32{{$}} # GCN: %10:vgpr_32 = V_MOV_B32_e32 -1241481, implicit $exec # GCN: BUFFER_STORE_DWORD_OFFSET killed %10, name: s_fold_xnor_imm_regimm_32 alignment: 0 exposesReturnsTwice: false legalized: false regBankSelected: false selected: false tracksRegLiveness: true liveins: - { reg: '$sgpr0_sgpr1', virtual-reg: '%0' } body: | bb.0: liveins: $sgpr0_sgpr1 %0:sgpr_64 = COPY $sgpr0_sgpr1 %1:sreg_64_xexec = S_LOAD_DWORDX2_IMM %0, 36, 0 %2:sreg_32_xm0 = COPY %1.sub1 %3:sreg_32_xm0 = COPY %1.sub0 %4:sreg_32_xm0 = S_MOV_B32 61440 %5:sreg_32_xm0 = S_MOV_B32 -1 %6:sgpr_128 = REG_SEQUENCE killed %2, %subreg.sub0, killed %3, %subreg.sub1, killed %4, %subreg.sub2, killed %5, %subreg.sub3 %7:sreg_32_xm0 = S_MOV_B32 1234567 %8:sreg_32_xm0 = S_MOV_B32 9999 %9:sreg_32_xm0 = S_XNOR_B32 killed %7, killed %8, implicit-def dead $scc %10:vgpr_32 = COPY %9 BUFFER_STORE_DWORD_OFFSET killed %10, killed %6, 0, 0, 0, 0, 0, implicit $exec S_ENDPGM 0 ... --- # GCN-LABEL: name: v_fold_shl_imm_regimm_32{{$}} # GCN: %16:vgpr_32 = V_MOV_B32_e32 40955904, implicit $exec # GCN: FLAT_STORE_DWORD %10, %16, # GCN: %17:vgpr_32 = V_MOV_B32_e32 24, implicit $exec # GCN: FLAT_STORE_DWORD %10, %17, # GCN: %18:vgpr_32 = V_MOV_B32_e32 4096, implicit $exec # GCN: FLAT_STORE_DWORD %10, %18, # GCN: %19:vgpr_32 = V_MOV_B32_e32 24, implicit $exec # GCN: FLAT_STORE_DWORD %10, %19, # GCN: %3:vgpr_32 = V_MOV_B32_e32 0, implicit $exec # GCN: FLAT_STORE_DWORD %10, %3, # GCN: %20:vgpr_32 = V_MOV_B32_e32 4096, implicit $exec # GCN: FLAT_STORE_DWORD %10, %20, # GCN: %21:vgpr_32 = V_MOV_B32_e32 1, implicit $exec # GCN: FLAT_STORE_DWORD %10, %21, # GCN: %22:vgpr_32 = V_MOV_B32_e32 2, implicit $exec # GCN: FLAT_STORE_DWORD %10, %22, # GCN: %23:vgpr_32 = V_MOV_B32_e32 7927808, implicit $exec # GCN: FLAT_STORE_DWORD %10, %23, # GCN: %24:vgpr_32 = V_MOV_B32_e32 -8, implicit $exec # GCN: FLAT_STORE_DWORD %10, %24, name: v_fold_shl_imm_regimm_32 alignment: 0 exposesReturnsTwice: false legalized: false regBankSelected: false selected: false tracksRegLiveness: true liveins: - { reg: '$sgpr0_sgpr1', virtual-reg: '%0' } - { reg: '$vgpr0', virtual-reg: '%2' } body: | bb.0: liveins: $sgpr0_sgpr1, $vgpr0 %2:vgpr_32 = COPY $vgpr0 %0:sgpr_64 = COPY $sgpr0_sgpr1 %3:sreg_64_xexec = S_LOAD_DWORDX2_IMM %0, 36, 0 :: (non-temporal dereferenceable invariant load (s64) from `i64 addrspace(4)* undef`, addrspace 4) %15:vgpr_32 = V_ASHRREV_I32_e64 31, %2, implicit $exec %16:vreg_64 = REG_SEQUENCE %2, %subreg.sub0, %15, %subreg.sub1 %17:vreg_64 = V_LSHLREV_B64_e64 2, killed %16, implicit $exec %9:sreg_32_xm0 = COPY %3.sub1 %21:vgpr_32 = V_ADD_CO_U32_e32 %3.sub0, %17.sub0, implicit-def $vcc, implicit $exec %19:vgpr_32 = COPY killed %9 %18:vgpr_32 = V_ADDC_U32_e32 %17.sub1, %19, implicit-def $vcc, implicit $vcc, implicit $exec %20:vreg_64 = REG_SEQUENCE %21, %subreg.sub0, killed %18, %subreg.sub1 %10:vgpr_32 = V_MOV_B32_e32 9999, implicit $exec %24:vgpr_32 = V_MOV_B32_e32 3871, implicit $exec %6:vgpr_32 = V_MOV_B32_e32 1, implicit $exec %7:sreg_32_xm0 = S_MOV_B32 1 %27:sreg_32_xm0 = S_MOV_B32 -4 %11:vgpr_32 = V_LSHLREV_B32_e64 12, %10, implicit $exec FLAT_STORE_DWORD %20, %11, 0, 0, implicit $exec, implicit $flat_scr %12:vgpr_32 = V_LSHLREV_B32_e64 %7, 12, implicit $exec FLAT_STORE_DWORD %20, %12, 0, 0, implicit $exec, implicit $flat_scr %13:vgpr_32 = V_LSHL_B32_e64 %7, 12, implicit $exec FLAT_STORE_DWORD %20, %13, 0, 0, implicit $exec, implicit $flat_scr %14:vgpr_32 = V_LSHL_B32_e64 12, %7, implicit $exec FLAT_STORE_DWORD %20, %14, 0, 0, implicit $exec, implicit $flat_scr %15:vgpr_32 = V_LSHL_B32_e64 12, %24, implicit $exec FLAT_STORE_DWORD %20, %15, 0, 0, implicit $exec, implicit $flat_scr %22:vgpr_32 = V_LSHL_B32_e64 %6, 12, implicit $exec FLAT_STORE_DWORD %20, %22, 0, 0, implicit $exec, implicit $flat_scr %23:vgpr_32 = V_LSHL_B32_e64 %6, 32, implicit $exec FLAT_STORE_DWORD %20, %23, 0, 0, implicit $exec, implicit $flat_scr %25:vgpr_32 = V_LSHL_B32_e32 %6, %6, implicit $exec FLAT_STORE_DWORD %20, %25, 0, 0, implicit $exec, implicit $flat_scr %26:vgpr_32 = V_LSHLREV_B32_e32 11, %24, implicit $exec FLAT_STORE_DWORD %20, %26, 0, 0, implicit $exec, implicit $flat_scr %28:vgpr_32 = V_LSHL_B32_e32 %27, %6, implicit $exec FLAT_STORE_DWORD %20, %28, 0, 0, implicit $exec, implicit $flat_scr S_ENDPGM 0 ... # This used to incorrectly interpret V_MOV_B32_sdwa as being a move # immediate, and interpreting the src0_modifiers field as a # materialized immediate. --- # GCN-LABEL: name: no_fold_sdwa_mov_imm # GCN: %2:vgpr_32 = V_MOV_B32_sdwa 0, %0, 0, 5, 2, 4, implicit $exec, implicit %0(tied-def 0) # GCN-NEXT: [[SHIFT:%[0-9]+]]:vgpr_32 = V_LSHRREV_B32_e64 16, %2, implicit $exec # GCN-NEXT: S_ENDPGM 0, implicit [[SHIFT]] name: no_fold_sdwa_mov_imm alignment: 0 exposesReturnsTwice: false legalized: false regBankSelected: false selected: false tracksRegLiveness: true body: | bb.0: liveins: $vgpr0 %0:vgpr_32 = COPY $vgpr0 %1:vgpr_32 = V_MOV_B32_e32 16, implicit $exec %2:vgpr_32 = V_MOV_B32_sdwa 0, %0:vgpr_32, 0, 5, 2, 4, implicit $exec, implicit %0:vgpr_32(tied-def 0) %3:vgpr_32 = V_MOV_B32_e32 16, implicit $exec %4:vgpr_32 = V_LSHRREV_B32_e64 %3:vgpr_32, %2:vgpr_32, implicit $exec S_ENDPGM 0, implicit %4 ...