109 lines · plain
1# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py2# RUN: llc -mtriple=amdgcn -mcpu=gfx1100 -mattr=-real-true16 -run-pass=gcn-dpp-combine -verify-machineinstrs -o - %s | FileCheck %s -check-prefix=GCN3 4---5 6name: vopc7tracksRegLiveness: true8body: |9 bb.0:10 liveins: $vgpr0, $vgpr1, $vgpr211 12 ; GCN-LABEL: name: vopc13 ; GCN: liveins: $vgpr0, $vgpr1, $vgpr214 ; GCN-NEXT: {{ $}}15 ; GCN-NEXT: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr016 ; GCN-NEXT: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr117 ; GCN-NEXT: [[COPY2:%[0-9]+]]:vgpr_32 = COPY $vgpr218 ; GCN-NEXT: [[DEF:%[0-9]+]]:vgpr_32 = IMPLICIT_DEF19 ; GCN-NEXT: V_CMP_LT_F32_e32_dpp 0, [[COPY1]], 0, [[COPY]], 1, 15, 15, 1, implicit-def $vcc, implicit $mode, implicit $exec20 ; GCN-NEXT: [[V_MOV_B32_dpp:%[0-9]+]]:vgpr_32 = V_MOV_B32_dpp [[DEF]], [[COPY1]], 1, 15, 15, 1, implicit $exec21 ; GCN-NEXT: V_CMPX_EQ_I16_fake16_nosdst_e64 [[V_MOV_B32_dpp]], [[COPY]], implicit-def $exec, implicit-def $vcc_lo, implicit $mode, implicit $exec22 ; GCN-NEXT: [[V_CMP_CLASS_F16_fake16_e64_dpp:%[0-9]+]]:sgpr_32 = V_CMP_CLASS_F16_fake16_e64_dpp 0, [[COPY1]], 0, [[COPY]], 1, 15, 15, 1, implicit $exec23 ; GCN-NEXT: [[V_CMP_GE_F16_fake16_e64_dpp:%[0-9]+]]:sgpr_32 = V_CMP_GE_F16_fake16_e64_dpp 1, [[COPY1]], 0, [[COPY]], 1, 1, 15, 15, 1, implicit $mode, implicit $exec24 ; GCN-NEXT: [[V_MOV_B32_dpp1:%[0-9]+]]:vgpr_32 = V_MOV_B32_dpp [[DEF]], [[COPY1]], 1, 15, 15, 1, implicit $exec25 ; GCN-NEXT: V_CMPX_GT_U32_nosdst_e64 [[V_MOV_B32_dpp1]], [[COPY]], implicit-def $exec, implicit $mode, implicit $exec26 ; GCN-NEXT: V_CMP_CLASS_F32_e32_dpp 2, [[COPY1]], [[COPY]], 1, 15, 15, 1, implicit-def $vcc, implicit $exec27 ; GCN-NEXT: V_CMP_NGE_F32_e32_dpp 0, [[COPY1]], 0, [[COPY]], 1, 15, 15, 1, implicit-def $vcc, implicit $mode, implicit $exec28 ; GCN-NEXT: [[V_MOV_B32_dpp2:%[0-9]+]]:vgpr_32 = V_MOV_B32_dpp [[DEF]], [[COPY1]], 1, 15, 15, 1, implicit $exec29 ; GCN-NEXT: [[V_CMP_NGE_F16_fake16_e64_:%[0-9]+]]:sgpr_32 = V_CMP_NGE_F16_fake16_e64 0, [[V_CMP_NGE_F16_fake16_e64_]], 0, [[COPY]], 0, implicit $mode, implicit $exec30 ; GCN-NEXT: [[V_CMP_NGE_F32_e64_dpp:%[0-9]+]]:sgpr_32 = V_CMP_NGE_F32_e64_dpp 0, [[COPY1]], 0, [[COPY]], 0, 1, 15, 15, 1, implicit $mode, implicit $exec31 ; GCN-NEXT: [[S_AND_B32_:%[0-9]+]]:sgpr_32 = S_AND_B32 [[V_CMP_NGE_F32_e64_dpp]], 10101, implicit-def $scc32 ; GCN-NEXT: V_CMP_GT_I32_e32_dpp [[COPY1]], [[COPY]], 1, 15, 15, 1, implicit-def $vcc, implicit $exec33 %0:vgpr_32 = COPY $vgpr034 %1:vgpr_32 = COPY $vgpr135 %2:vgpr_32 = COPY $vgpr236 %3:vgpr_32 = IMPLICIT_DEF37 38 %4:vgpr_32 = V_MOV_B32_dpp %3, %1, 1, 15, 15, 1, implicit $exec39 V_CMP_LT_F32_e32 %4, %0, implicit-def $vcc, implicit $mode, implicit $exec40 41 ; unsafe to combine cmpx42 %5:vgpr_32 = V_MOV_B32_dpp %3, %1, 1, 15, 15, 1, implicit $exec43 V_CMPX_EQ_I16_fake16_nosdst_e64 %5, %0, implicit-def $exec, implicit-def $vcc, implicit $mode, implicit $exec44 45 %6:vgpr_32 = V_MOV_B32_dpp %3, %1, 1, 15, 15, 1, implicit $exec46 %7:sgpr_32 = V_CMP_CLASS_F16_fake16_e64 0, %6, 0, %0, implicit-def $vcc, implicit $mode, implicit $exec47 48 %8:vgpr_32 = V_MOV_B32_dpp %3, %1, 1, 15, 15, 1, implicit $exec49 %9:sgpr_32 = V_CMP_GE_F16_fake16_e64 1, %8, 0, %0, 1, implicit $mode, implicit $exec50 51 ; unsafe to combine cmpx52 %10:vgpr_32 = V_MOV_B32_dpp %3, %1, 1, 15, 15, 1, implicit $exec53 V_CMPX_GT_U32_nosdst_e64 %10, %0, implicit-def $exec, implicit $mode, implicit $exec54 55 %11:vgpr_32 = V_MOV_B32_dpp %3, %1, 1, 15, 15, 1, implicit $exec56 %12:sgpr_32 = V_CMP_CLASS_F32_e64 2, %11, %0, implicit $mode, implicit $exec57 58 ; shrink59 %13:vgpr_32 = V_MOV_B32_dpp %3, %1, 1, 15, 15, 1, implicit $exec60 %14:sgpr_32 = V_CMP_NGE_F32_e64 0, %13, 0, %0, 0, implicit $mode, implicit $exec61 62 ; do not shrink True16 instructions63 %15:vgpr_32 = V_MOV_B32_dpp %3, %1, 1, 15, 15, 1, implicit $exec64 %16:sgpr_32 = V_CMP_NGE_F16_fake16_e64 0, %16, 0, %0, 0, implicit $mode, implicit $exec65 66 ; do not shrink, sdst used67 %17:vgpr_32 = V_MOV_B32_dpp %3, %1, 1, 15, 15, 1, implicit $exec68 %18:sgpr_32 = V_CMP_NGE_F32_e64 0, %17, 0, %0, 0, implicit $mode, implicit $exec69 %19:sgpr_32 = S_AND_B32 %18, 10101, implicit-def $scc70 71 ; commute72 %20:vgpr_32 = V_MOV_B32_dpp %3, %1, 1, 15, 15, 1, implicit $exec73 V_CMP_LT_I32_e32 %0, %20, implicit-def $vcc, implicit $exec74 75...76---77 78name: mask_not_full79tracksRegLiveness: true80body: |81 bb.0:82 liveins: $vgpr0, $vgpr1, $vgpr283 84 ; GCN-LABEL: name: mask_not_full85 ; GCN: liveins: $vgpr0, $vgpr1, $vgpr286 ; GCN-NEXT: {{ $}}87 ; GCN-NEXT: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr088 ; GCN-NEXT: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr189 ; GCN-NEXT: [[DEF:%[0-9]+]]:vgpr_32 = IMPLICIT_DEF90 ; GCN-NEXT: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 0, implicit $exec91 ; GCN-NEXT: [[V_MOV_B32_dpp:%[0-9]+]]:vgpr_32 = V_MOV_B32_dpp [[DEF]], [[COPY1]], 1, 15, 14, 1, implicit $exec92 ; GCN-NEXT: [[V_CMP_CLASS_F16_fake16_e64_:%[0-9]+]]:sgpr_32 = V_CMP_CLASS_F16_fake16_e64 0, [[V_MOV_B32_dpp]], 0, [[COPY]], implicit-def $vcc_lo, implicit $mode, implicit $exec93 ; GCN-NEXT: [[V_MOV_B32_dpp1:%[0-9]+]]:vgpr_32 = V_MOV_B32_dpp [[V_MOV_B32_e32_]], [[COPY1]], 1, 13, 15, 1, implicit $exec94 ; GCN-NEXT: [[V_CMP_GE_F32_e64_:%[0-9]+]]:sgpr_32 = V_CMP_GE_F32_e64 1, [[V_MOV_B32_dpp1]], 0, [[COPY]], 1, implicit $mode, implicit $exec95 %0:vgpr_32 = COPY $vgpr096 %1:vgpr_32 = COPY $vgpr197 %2:vgpr_32 = IMPLICIT_DEF98 %3:vgpr_32 = V_MOV_B32_e32 0, implicit $exec99 100 ; Do not combine VOPC when row_mask or bank_mask is not 0xf101 ; All cases are covered by generic rules for creating DPP instructions102 %4:vgpr_32 = V_MOV_B32_dpp %2, %1, 1, 15, 14, 1, implicit $exec103 %99:sgpr_32 = V_CMP_CLASS_F16_fake16_e64 0, %4, 0, %0, implicit-def $vcc, implicit $mode, implicit $exec104 105 %5:vgpr_32 = V_MOV_B32_dpp %3, %1, 1, 13, 15, 1, implicit $exec106 %6:sgpr_32 = V_CMP_GE_F32_e64 1, %5, 0, %0, 1, implicit $mode, implicit $exec107 108...109