Skip to content

Commit 3963d21

Browse files
jrbyrnesjrbyrnes
authored
[MachineSink] Fix typo in loop sinking (llvm#127133)
Failure to sink a candidate should not block us from attempting to sink other candidates. There are mechanisms in place to handle the case where the failed to be sunk instruction uses an instruction that gets sunk (we do not delete the original instruction corresponding with the sunk instruction if it still has uses).
1 parent bf9bf29 commit 3963d21

File tree

2 files changed

+83
-29
lines changed

2 files changed

+83
-29
lines changed

‎llvm/lib/CodeGen/MachineSink.cpp‎

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -919,7 +919,7 @@ bool MachineSinking::run(MachineFunction &MF) {
919919
continue;
920920

921921
if (!aggressivelySinkIntoCycle(Cycle, *I, SunkInstrs))
922-
break;
922+
continue;
923923
EverMadeChange = true;
924924
++NumCycleSunk;
925925
}

‎llvm/test/CodeGen/AMDGPU/machine-sink-cycle.mir‎

Lines changed: 82 additions & 28 deletions
Original file line numberDiff line numberDiff line change
@@ -1078,6 +1078,19 @@ body: |
10781078
; GFX9-SUNK-NEXT: [[DEF2:%[0-9]+]]:vreg_256_align2 = IMPLICIT_DEF
10791079
; GFX9-SUNK-NEXT: [[DEF3:%[0-9]+]]:vreg_256_align2 = IMPLICIT_DEF
10801080
; GFX9-SUNK-NEXT: [[DEF4:%[0-9]+]]:vreg_256_align2 = IMPLICIT_DEF
1081+
; GFX9-SUNK-NEXT: [[V_ADD_CO_U32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_CO_U32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADD_CO_U32_e64 [[DEF]].sub2, [[DEF1]].sub4, 0, implicit $exec
1082+
; GFX9-SUNK-NEXT: S_BRANCH %bb.1
1083+
; GFX9-SUNK-NEXT: {{ $}}
1084+
; GFX9-SUNK-NEXT: bb.1:
1085+
; GFX9-SUNK-NEXT: successors: %bb.2(0x40000000), %bb.3(0x40000000)
1086+
; GFX9-SUNK-NEXT: {{ $}}
1087+
; GFX9-SUNK-NEXT: INLINEASM &"", 1 /* sideeffect attdialect */, implicit [[V_ADD_CO_U32_e64_]], implicit [[V_ADD_CO_U32_e64_1]]
1088+
; GFX9-SUNK-NEXT: S_CBRANCH_EXECZ %bb.2, implicit $exec
1089+
; GFX9-SUNK-NEXT: S_BRANCH %bb.3
1090+
; GFX9-SUNK-NEXT: {{ $}}
1091+
; GFX9-SUNK-NEXT: bb.2:
1092+
; GFX9-SUNK-NEXT: successors: %bb.4(0x80000000)
1093+
; GFX9-SUNK-NEXT: {{ $}}
10811094
; GFX9-SUNK-NEXT: [[V_ADD_U32_e64_:%[0-9]+]]:vgpr_32 = V_ADD_U32_e64 [[DEF]].sub0, [[DEF1]].sub0, 0, implicit $exec
10821095
; GFX9-SUNK-NEXT: [[V_ADD_U32_e64_1:%[0-9]+]]:vgpr_32 = V_ADD_U32_e64 [[DEF]].sub1, [[DEF1]].sub0, 0, implicit $exec
10831096
; GFX9-SUNK-NEXT: [[V_ADD_U32_e64_2:%[0-9]+]]:vgpr_32 = V_ADD_U32_e64 [[DEF]].sub2, [[DEF1]].sub0, 0, implicit $exec
@@ -1105,28 +1118,42 @@ body: |
11051118
; GFX9-SUNK-NEXT: [[V_ADD_U32_e64_24:%[0-9]+]]:vgpr_32 = V_ADD_U32_e64 [[DEF]].sub0, [[DEF1]].sub3, 0, implicit $exec
11061119
; GFX9-SUNK-NEXT: [[V_ADD_U32_e64_25:%[0-9]+]]:vgpr_32 = V_ADD_U32_e64 [[DEF]].sub1, [[DEF1]].sub3, 0, implicit $exec
11071120
; GFX9-SUNK-NEXT: [[V_ADD_U32_e64_26:%[0-9]+]]:vgpr_32 = V_ADD_U32_e64 [[DEF]].sub2, [[DEF1]].sub3, 0, implicit $exec
1108-
; GFX9-SUNK-NEXT: [[V_ADD_CO_U32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_CO_U32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADD_CO_U32_e64 [[DEF]].sub2, [[DEF1]].sub4, 0, implicit $exec
1109-
; GFX9-SUNK-NEXT: S_BRANCH %bb.1
1110-
; GFX9-SUNK-NEXT: {{ $}}
1111-
; GFX9-SUNK-NEXT: bb.1:
1112-
; GFX9-SUNK-NEXT: successors: %bb.2(0x40000000), %bb.3(0x40000000)
1113-
; GFX9-SUNK-NEXT: {{ $}}
1114-
; GFX9-SUNK-NEXT: INLINEASM &"", 1 /* sideeffect attdialect */, implicit [[V_ADD_CO_U32_e64_]], implicit [[V_ADD_CO_U32_e64_1]]
1115-
; GFX9-SUNK-NEXT: S_CBRANCH_EXECZ %bb.2, implicit $exec
1116-
; GFX9-SUNK-NEXT: S_BRANCH %bb.3
1117-
; GFX9-SUNK-NEXT: {{ $}}
1118-
; GFX9-SUNK-NEXT: bb.2:
1119-
; GFX9-SUNK-NEXT: successors: %bb.4(0x80000000)
1120-
; GFX9-SUNK-NEXT: {{ $}}
11211121
; GFX9-SUNK-NEXT: INLINEASM &"", 1 /* sideeffect attdialect */, implicit [[DEF]], implicit [[DEF1]], implicit [[DEF2]], implicit [[DEF3]], implicit [[DEF4]]
11221122
; GFX9-SUNK-NEXT: INLINEASM &"", 1 /* sideeffect attdialect */, implicit [[V_ADD_U32_e64_]], implicit [[V_ADD_U32_e64_1]], implicit [[V_ADD_U32_e64_2]], implicit [[V_ADD_U32_e64_3]], implicit [[V_ADD_U32_e64_4]], implicit [[V_ADD_U32_e64_5]], implicit [[V_ADD_U32_e64_6]], implicit [[V_ADD_U32_e64_7]], implicit [[V_ADD_U32_e64_8]], implicit [[V_ADD_U32_e64_9]], implicit [[V_ADD_U32_e64_10]], implicit [[V_ADD_U32_e64_11]], implicit [[V_ADD_U32_e64_12]], implicit [[V_ADD_U32_e64_13]], implicit [[V_ADD_U32_e64_14]], implicit [[V_ADD_U32_e64_15]], implicit [[V_ADD_U32_e64_16]], implicit [[V_ADD_U32_e64_17]], implicit [[V_ADD_U32_e64_18]], implicit [[V_ADD_U32_e64_19]], implicit [[V_ADD_U32_e64_20]], implicit [[V_ADD_U32_e64_21]], implicit [[V_ADD_U32_e64_22]], implicit [[V_ADD_U32_e64_23]], implicit [[V_ADD_U32_e64_24]], implicit [[V_ADD_U32_e64_25]], implicit [[V_ADD_U32_e64_26]]
11231123
; GFX9-SUNK-NEXT: S_BRANCH %bb.4
11241124
; GFX9-SUNK-NEXT: {{ $}}
11251125
; GFX9-SUNK-NEXT: bb.3:
11261126
; GFX9-SUNK-NEXT: successors: %bb.4(0x80000000)
11271127
; GFX9-SUNK-NEXT: {{ $}}
1128+
; GFX9-SUNK-NEXT: [[V_ADD_U32_e64_27:%[0-9]+]]:vgpr_32 = V_ADD_U32_e64 [[DEF]].sub0, [[DEF1]].sub0, 0, implicit $exec
1129+
; GFX9-SUNK-NEXT: [[V_ADD_U32_e64_28:%[0-9]+]]:vgpr_32 = V_ADD_U32_e64 [[DEF]].sub1, [[DEF1]].sub0, 0, implicit $exec
1130+
; GFX9-SUNK-NEXT: [[V_ADD_U32_e64_29:%[0-9]+]]:vgpr_32 = V_ADD_U32_e64 [[DEF]].sub2, [[DEF1]].sub0, 0, implicit $exec
1131+
; GFX9-SUNK-NEXT: [[V_ADD_U32_e64_30:%[0-9]+]]:vgpr_32 = V_ADD_U32_e64 [[DEF]].sub3, [[DEF1]].sub0, 0, implicit $exec
1132+
; GFX9-SUNK-NEXT: [[V_ADD_U32_e64_31:%[0-9]+]]:vgpr_32 = V_ADD_U32_e64 [[DEF]].sub4, [[DEF1]].sub0, 0, implicit $exec
1133+
; GFX9-SUNK-NEXT: [[V_ADD_U32_e64_32:%[0-9]+]]:vgpr_32 = V_ADD_U32_e64 [[DEF]].sub5, [[DEF1]].sub0, 0, implicit $exec
1134+
; GFX9-SUNK-NEXT: [[V_ADD_U32_e64_33:%[0-9]+]]:vgpr_32 = V_ADD_U32_e64 [[DEF]].sub6, [[DEF1]].sub0, 0, implicit $exec
1135+
; GFX9-SUNK-NEXT: [[V_ADD_U32_e64_34:%[0-9]+]]:vgpr_32 = V_ADD_U32_e64 [[DEF]].sub7, [[DEF1]].sub0, 0, implicit $exec
1136+
; GFX9-SUNK-NEXT: [[V_ADD_U32_e64_35:%[0-9]+]]:vgpr_32 = V_ADD_U32_e64 [[DEF]].sub0, [[DEF1]].sub1, 0, implicit $exec
1137+
; GFX9-SUNK-NEXT: [[V_ADD_U32_e64_36:%[0-9]+]]:vgpr_32 = V_ADD_U32_e64 [[DEF]].sub1, [[DEF1]].sub1, 0, implicit $exec
1138+
; GFX9-SUNK-NEXT: [[V_ADD_U32_e64_37:%[0-9]+]]:vgpr_32 = V_ADD_U32_e64 [[DEF]].sub2, [[DEF1]].sub1, 0, implicit $exec
1139+
; GFX9-SUNK-NEXT: [[V_ADD_U32_e64_38:%[0-9]+]]:vgpr_32 = V_ADD_U32_e64 [[DEF]].sub3, [[DEF1]].sub1, 0, implicit $exec
1140+
; GFX9-SUNK-NEXT: [[V_ADD_U32_e64_39:%[0-9]+]]:vgpr_32 = V_ADD_U32_e64 [[DEF]].sub4, [[DEF1]].sub1, 0, implicit $exec
1141+
; GFX9-SUNK-NEXT: [[V_ADD_U32_e64_40:%[0-9]+]]:vgpr_32 = V_ADD_U32_e64 [[DEF]].sub5, [[DEF1]].sub1, 0, implicit $exec
1142+
; GFX9-SUNK-NEXT: [[V_ADD_U32_e64_41:%[0-9]+]]:vgpr_32 = V_ADD_U32_e64 [[DEF]].sub6, [[DEF1]].sub1, 0, implicit $exec
1143+
; GFX9-SUNK-NEXT: [[V_ADD_U32_e64_42:%[0-9]+]]:vgpr_32 = V_ADD_U32_e64 [[DEF]].sub7, [[DEF1]].sub1, 0, implicit $exec
1144+
; GFX9-SUNK-NEXT: [[V_ADD_U32_e64_43:%[0-9]+]]:vgpr_32 = V_ADD_U32_e64 [[DEF]].sub0, [[DEF1]].sub2, 0, implicit $exec
1145+
; GFX9-SUNK-NEXT: [[V_ADD_U32_e64_44:%[0-9]+]]:vgpr_32 = V_ADD_U32_e64 [[DEF]].sub1, [[DEF1]].sub2, 0, implicit $exec
1146+
; GFX9-SUNK-NEXT: [[V_ADD_U32_e64_45:%[0-9]+]]:vgpr_32 = V_ADD_U32_e64 [[DEF]].sub2, [[DEF1]].sub2, 0, implicit $exec
1147+
; GFX9-SUNK-NEXT: [[V_ADD_U32_e64_46:%[0-9]+]]:vgpr_32 = V_ADD_U32_e64 [[DEF]].sub3, [[DEF1]].sub2, 0, implicit $exec
1148+
; GFX9-SUNK-NEXT: [[V_ADD_U32_e64_47:%[0-9]+]]:vgpr_32 = V_ADD_U32_e64 [[DEF]].sub4, [[DEF1]].sub2, 0, implicit $exec
1149+
; GFX9-SUNK-NEXT: [[V_ADD_U32_e64_48:%[0-9]+]]:vgpr_32 = V_ADD_U32_e64 [[DEF]].sub5, [[DEF1]].sub2, 0, implicit $exec
1150+
; GFX9-SUNK-NEXT: [[V_ADD_U32_e64_49:%[0-9]+]]:vgpr_32 = V_ADD_U32_e64 [[DEF]].sub6, [[DEF1]].sub2, 0, implicit $exec
1151+
; GFX9-SUNK-NEXT: [[V_ADD_U32_e64_50:%[0-9]+]]:vgpr_32 = V_ADD_U32_e64 [[DEF]].sub7, [[DEF1]].sub2, 0, implicit $exec
1152+
; GFX9-SUNK-NEXT: [[V_ADD_U32_e64_51:%[0-9]+]]:vgpr_32 = V_ADD_U32_e64 [[DEF]].sub0, [[DEF1]].sub3, 0, implicit $exec
1153+
; GFX9-SUNK-NEXT: [[V_ADD_U32_e64_52:%[0-9]+]]:vgpr_32 = V_ADD_U32_e64 [[DEF]].sub1, [[DEF1]].sub3, 0, implicit $exec
1154+
; GFX9-SUNK-NEXT: [[V_ADD_U32_e64_53:%[0-9]+]]:vgpr_32 = V_ADD_U32_e64 [[DEF]].sub2, [[DEF1]].sub3, 0, implicit $exec
11281155
; GFX9-SUNK-NEXT: INLINEASM &"", 1 /* sideeffect attdialect */, implicit [[DEF]], implicit [[DEF1]], implicit [[DEF2]], implicit [[DEF3]], implicit [[DEF4]]
1129-
; GFX9-SUNK-NEXT: INLINEASM &"", 1 /* sideeffect attdialect */, implicit [[V_ADD_U32_e64_]], implicit [[V_ADD_U32_e64_1]], implicit [[V_ADD_U32_e64_2]], implicit [[V_ADD_U32_e64_3]], implicit [[V_ADD_U32_e64_4]], implicit [[V_ADD_U32_e64_5]], implicit [[V_ADD_U32_e64_6]], implicit [[V_ADD_U32_e64_7]], implicit [[V_ADD_U32_e64_8]], implicit [[V_ADD_U32_e64_9]], implicit [[V_ADD_U32_e64_10]], implicit [[V_ADD_U32_e64_11]], implicit [[V_ADD_U32_e64_12]], implicit [[V_ADD_U32_e64_13]], implicit [[V_ADD_U32_e64_14]], implicit [[V_ADD_U32_e64_15]], implicit [[V_ADD_U32_e64_16]], implicit [[V_ADD_U32_e64_17]], implicit [[V_ADD_U32_e64_18]], implicit [[V_ADD_U32_e64_19]], implicit [[V_ADD_U32_e64_20]], implicit [[V_ADD_U32_e64_21]], implicit [[V_ADD_U32_e64_22]], implicit [[V_ADD_U32_e64_23]], implicit [[V_ADD_U32_e64_24]], implicit [[V_ADD_U32_e64_25]], implicit [[V_ADD_U32_e64_26]]
1156+
; GFX9-SUNK-NEXT: INLINEASM &"", 1 /* sideeffect attdialect */, implicit [[V_ADD_U32_e64_27]], implicit [[V_ADD_U32_e64_28]], implicit [[V_ADD_U32_e64_29]], implicit [[V_ADD_U32_e64_30]], implicit [[V_ADD_U32_e64_31]], implicit [[V_ADD_U32_e64_32]], implicit [[V_ADD_U32_e64_33]], implicit [[V_ADD_U32_e64_34]], implicit [[V_ADD_U32_e64_35]], implicit [[V_ADD_U32_e64_36]], implicit [[V_ADD_U32_e64_37]], implicit [[V_ADD_U32_e64_38]], implicit [[V_ADD_U32_e64_39]], implicit [[V_ADD_U32_e64_40]], implicit [[V_ADD_U32_e64_41]], implicit [[V_ADD_U32_e64_42]], implicit [[V_ADD_U32_e64_43]], implicit [[V_ADD_U32_e64_44]], implicit [[V_ADD_U32_e64_45]], implicit [[V_ADD_U32_e64_46]], implicit [[V_ADD_U32_e64_47]], implicit [[V_ADD_U32_e64_48]], implicit [[V_ADD_U32_e64_49]], implicit [[V_ADD_U32_e64_50]], implicit [[V_ADD_U32_e64_51]], implicit [[V_ADD_U32_e64_52]], implicit [[V_ADD_U32_e64_53]]
11301157
; GFX9-SUNK-NEXT: S_BRANCH %bb.4
11311158
; GFX9-SUNK-NEXT: {{ $}}
11321159
; GFX9-SUNK-NEXT: bb.4:
@@ -1149,6 +1176,19 @@ body: |
11491176
; GFX10-SUNK-NEXT: [[DEF2:%[0-9]+]]:vreg_256 = IMPLICIT_DEF
11501177
; GFX10-SUNK-NEXT: [[DEF3:%[0-9]+]]:vreg_256 = IMPLICIT_DEF
11511178
; GFX10-SUNK-NEXT: [[DEF4:%[0-9]+]]:vreg_256 = IMPLICIT_DEF
1179+
; GFX10-SUNK-NEXT: [[V_ADD_CO_U32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_CO_U32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADD_CO_U32_e64 [[DEF]].sub2, [[DEF1]].sub4, 0, implicit $exec
1180+
; GFX10-SUNK-NEXT: S_BRANCH %bb.1
1181+
; GFX10-SUNK-NEXT: {{ $}}
1182+
; GFX10-SUNK-NEXT: bb.1:
1183+
; GFX10-SUNK-NEXT: successors: %bb.2(0x40000000), %bb.3(0x40000000)
1184+
; GFX10-SUNK-NEXT: {{ $}}
1185+
; GFX10-SUNK-NEXT: INLINEASM &"", 1 /* sideeffect attdialect */, implicit [[V_ADD_CO_U32_e64_]], implicit [[V_ADD_CO_U32_e64_1]]
1186+
; GFX10-SUNK-NEXT: S_CBRANCH_EXECZ %bb.2, implicit $exec
1187+
; GFX10-SUNK-NEXT: S_BRANCH %bb.3
1188+
; GFX10-SUNK-NEXT: {{ $}}
1189+
; GFX10-SUNK-NEXT: bb.2:
1190+
; GFX10-SUNK-NEXT: successors: %bb.4(0x80000000)
1191+
; GFX10-SUNK-NEXT: {{ $}}
11521192
; GFX10-SUNK-NEXT: [[V_ADD_U32_e64_:%[0-9]+]]:vgpr_32 = V_ADD_U32_e64 [[DEF]].sub0, [[DEF1]].sub0, 0, implicit $exec
11531193
; GFX10-SUNK-NEXT: [[V_ADD_U32_e64_1:%[0-9]+]]:vgpr_32 = V_ADD_U32_e64 [[DEF]].sub1, [[DEF1]].sub0, 0, implicit $exec
11541194
; GFX10-SUNK-NEXT: [[V_ADD_U32_e64_2:%[0-9]+]]:vgpr_32 = V_ADD_U32_e64 [[DEF]].sub2, [[DEF1]].sub0, 0, implicit $exec
@@ -1176,28 +1216,42 @@ body: |
11761216
; GFX10-SUNK-NEXT: [[V_ADD_U32_e64_24:%[0-9]+]]:vgpr_32 = V_ADD_U32_e64 [[DEF]].sub0, [[DEF1]].sub3, 0, implicit $exec
11771217
; GFX10-SUNK-NEXT: [[V_ADD_U32_e64_25:%[0-9]+]]:vgpr_32 = V_ADD_U32_e64 [[DEF]].sub1, [[DEF1]].sub3, 0, implicit $exec
11781218
; GFX10-SUNK-NEXT: [[V_ADD_U32_e64_26:%[0-9]+]]:vgpr_32 = V_ADD_U32_e64 [[DEF]].sub2, [[DEF1]].sub3, 0, implicit $exec
1179-
; GFX10-SUNK-NEXT: [[V_ADD_CO_U32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_CO_U32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADD_CO_U32_e64 [[DEF]].sub2, [[DEF1]].sub4, 0, implicit $exec
1180-
; GFX10-SUNK-NEXT: S_BRANCH %bb.1
1181-
; GFX10-SUNK-NEXT: {{ $}}
1182-
; GFX10-SUNK-NEXT: bb.1:
1183-
; GFX10-SUNK-NEXT: successors: %bb.2(0x40000000), %bb.3(0x40000000)
1184-
; GFX10-SUNK-NEXT: {{ $}}
1185-
; GFX10-SUNK-NEXT: INLINEASM &"", 1 /* sideeffect attdialect */, implicit [[V_ADD_CO_U32_e64_]], implicit [[V_ADD_CO_U32_e64_1]]
1186-
; GFX10-SUNK-NEXT: S_CBRANCH_EXECZ %bb.2, implicit $exec
1187-
; GFX10-SUNK-NEXT: S_BRANCH %bb.3
1188-
; GFX10-SUNK-NEXT: {{ $}}
1189-
; GFX10-SUNK-NEXT: bb.2:
1190-
; GFX10-SUNK-NEXT: successors: %bb.4(0x80000000)
1191-
; GFX10-SUNK-NEXT: {{ $}}
11921219
; GFX10-SUNK-NEXT: INLINEASM &"", 1 /* sideeffect attdialect */, implicit [[DEF]], implicit [[DEF1]], implicit [[DEF2]], implicit [[DEF3]], implicit [[DEF4]]
11931220
; GFX10-SUNK-NEXT: INLINEASM &"", 1 /* sideeffect attdialect */, implicit [[V_ADD_U32_e64_]], implicit [[V_ADD_U32_e64_1]], implicit [[V_ADD_U32_e64_2]], implicit [[V_ADD_U32_e64_3]], implicit [[V_ADD_U32_e64_4]], implicit [[V_ADD_U32_e64_5]], implicit [[V_ADD_U32_e64_6]], implicit [[V_ADD_U32_e64_7]], implicit [[V_ADD_U32_e64_8]], implicit [[V_ADD_U32_e64_9]], implicit [[V_ADD_U32_e64_10]], implicit [[V_ADD_U32_e64_11]], implicit [[V_ADD_U32_e64_12]], implicit [[V_ADD_U32_e64_13]], implicit [[V_ADD_U32_e64_14]], implicit [[V_ADD_U32_e64_15]], implicit [[V_ADD_U32_e64_16]], implicit [[V_ADD_U32_e64_17]], implicit [[V_ADD_U32_e64_18]], implicit [[V_ADD_U32_e64_19]], implicit [[V_ADD_U32_e64_20]], implicit [[V_ADD_U32_e64_21]], implicit [[V_ADD_U32_e64_22]], implicit [[V_ADD_U32_e64_23]], implicit [[V_ADD_U32_e64_24]], implicit [[V_ADD_U32_e64_25]], implicit [[V_ADD_U32_e64_26]]
11941221
; GFX10-SUNK-NEXT: S_BRANCH %bb.4
11951222
; GFX10-SUNK-NEXT: {{ $}}
11961223
; GFX10-SUNK-NEXT: bb.3:
11971224
; GFX10-SUNK-NEXT: successors: %bb.4(0x80000000)
11981225
; GFX10-SUNK-NEXT: {{ $}}
1226+
; GFX10-SUNK-NEXT: [[V_ADD_U32_e64_27:%[0-9]+]]:vgpr_32 = V_ADD_U32_e64 [[DEF]].sub0, [[DEF1]].sub0, 0, implicit $exec
1227+
; GFX10-SUNK-NEXT: [[V_ADD_U32_e64_28:%[0-9]+]]:vgpr_32 = V_ADD_U32_e64 [[DEF]].sub1, [[DEF1]].sub0, 0, implicit $exec
1228+
; GFX10-SUNK-NEXT: [[V_ADD_U32_e64_29:%[0-9]+]]:vgpr_32 = V_ADD_U32_e64 [[DEF]].sub2, [[DEF1]].sub0, 0, implicit $exec
1229+
; GFX10-SUNK-NEXT: [[V_ADD_U32_e64_30:%[0-9]+]]:vgpr_32 = V_ADD_U32_e64 [[DEF]].sub3, [[DEF1]].sub0, 0, implicit $exec
1230+
; GFX10-SUNK-NEXT: [[V_ADD_U32_e64_31:%[0-9]+]]:vgpr_32 = V_ADD_U32_e64 [[DEF]].sub4, [[DEF1]].sub0, 0, implicit $exec
1231+
; GFX10-SUNK-NEXT: [[V_ADD_U32_e64_32:%[0-9]+]]:vgpr_32 = V_ADD_U32_e64 [[DEF]].sub5, [[DEF1]].sub0, 0, implicit $exec
1232+
; GFX10-SUNK-NEXT: [[V_ADD_U32_e64_33:%[0-9]+]]:vgpr_32 = V_ADD_U32_e64 [[DEF]].sub6, [[DEF1]].sub0, 0, implicit $exec
1233+
; GFX10-SUNK-NEXT: [[V_ADD_U32_e64_34:%[0-9]+]]:vgpr_32 = V_ADD_U32_e64 [[DEF]].sub7, [[DEF1]].sub0, 0, implicit $exec
1234+
; GFX10-SUNK-NEXT: [[V_ADD_U32_e64_35:%[0-9]+]]:vgpr_32 = V_ADD_U32_e64 [[DEF]].sub0, [[DEF1]].sub1, 0, implicit $exec
1235+
; GFX10-SUNK-NEXT: [[V_ADD_U32_e64_36:%[0-9]+]]:vgpr_32 = V_ADD_U32_e64 [[DEF]].sub1, [[DEF1]].sub1, 0, implicit $exec
1236+
; GFX10-SUNK-NEXT: [[V_ADD_U32_e64_37:%[0-9]+]]:vgpr_32 = V_ADD_U32_e64 [[DEF]].sub2, [[DEF1]].sub1, 0, implicit $exec
1237+
; GFX10-SUNK-NEXT: [[V_ADD_U32_e64_38:%[0-9]+]]:vgpr_32 = V_ADD_U32_e64 [[DEF]].sub3, [[DEF1]].sub1, 0, implicit $exec
1238+
; GFX10-SUNK-NEXT: [[V_ADD_U32_e64_39:%[0-9]+]]:vgpr_32 = V_ADD_U32_e64 [[DEF]].sub4, [[DEF1]].sub1, 0, implicit $exec
1239+
; GFX10-SUNK-NEXT: [[V_ADD_U32_e64_40:%[0-9]+]]:vgpr_32 = V_ADD_U32_e64 [[DEF]].sub5, [[DEF1]].sub1, 0, implicit $exec
1240+
; GFX10-SUNK-NEXT: [[V_ADD_U32_e64_41:%[0-9]+]]:vgpr_32 = V_ADD_U32_e64 [[DEF]].sub6, [[DEF1]].sub1, 0, implicit $exec
1241+
; GFX10-SUNK-NEXT: [[V_ADD_U32_e64_42:%[0-9]+]]:vgpr_32 = V_ADD_U32_e64 [[DEF]].sub7, [[DEF1]].sub1, 0, implicit $exec
1242+
; GFX10-SUNK-NEXT: [[V_ADD_U32_e64_43:%[0-9]+]]:vgpr_32 = V_ADD_U32_e64 [[DEF]].sub0, [[DEF1]].sub2, 0, implicit $exec
1243+
; GFX10-SUNK-NEXT: [[V_ADD_U32_e64_44:%[0-9]+]]:vgpr_32 = V_ADD_U32_e64 [[DEF]].sub1, [[DEF1]].sub2, 0, implicit $exec
1244+
; GFX10-SUNK-NEXT: [[V_ADD_U32_e64_45:%[0-9]+]]:vgpr_32 = V_ADD_U32_e64 [[DEF]].sub2, [[DEF1]].sub2, 0, implicit $exec
1245+
; GFX10-SUNK-NEXT: [[V_ADD_U32_e64_46:%[0-9]+]]:vgpr_32 = V_ADD_U32_e64 [[DEF]].sub3, [[DEF1]].sub2, 0, implicit $exec
1246+
; GFX10-SUNK-NEXT: [[V_ADD_U32_e64_47:%[0-9]+]]:vgpr_32 = V_ADD_U32_e64 [[DEF]].sub4, [[DEF1]].sub2, 0, implicit $exec
1247+
; GFX10-SUNK-NEXT: [[V_ADD_U32_e64_48:%[0-9]+]]:vgpr_32 = V_ADD_U32_e64 [[DEF]].sub5, [[DEF1]].sub2, 0, implicit $exec
1248+
; GFX10-SUNK-NEXT: [[V_ADD_U32_e64_49:%[0-9]+]]:vgpr_32 = V_ADD_U32_e64 [[DEF]].sub6, [[DEF1]].sub2, 0, implicit $exec
1249+
; GFX10-SUNK-NEXT: [[V_ADD_U32_e64_50:%[0-9]+]]:vgpr_32 = V_ADD_U32_e64 [[DEF]].sub7, [[DEF1]].sub2, 0, implicit $exec
1250+
; GFX10-SUNK-NEXT: [[V_ADD_U32_e64_51:%[0-9]+]]:vgpr_32 = V_ADD_U32_e64 [[DEF]].sub0, [[DEF1]].sub3, 0, implicit $exec
1251+
; GFX10-SUNK-NEXT: [[V_ADD_U32_e64_52:%[0-9]+]]:vgpr_32 = V_ADD_U32_e64 [[DEF]].sub1, [[DEF1]].sub3, 0, implicit $exec
1252+
; GFX10-SUNK-NEXT: [[V_ADD_U32_e64_53:%[0-9]+]]:vgpr_32 = V_ADD_U32_e64 [[DEF]].sub2, [[DEF1]].sub3, 0, implicit $exec
11991253
; GFX10-SUNK-NEXT: INLINEASM &"", 1 /* sideeffect attdialect */, implicit [[DEF]], implicit [[DEF1]], implicit [[DEF2]], implicit [[DEF3]], implicit [[DEF4]]
1200-
; GFX10-SUNK-NEXT: INLINEASM &"", 1 /* sideeffect attdialect */, implicit [[V_ADD_U32_e64_]], implicit [[V_ADD_U32_e64_1]], implicit [[V_ADD_U32_e64_2]], implicit [[V_ADD_U32_e64_3]], implicit [[V_ADD_U32_e64_4]], implicit [[V_ADD_U32_e64_5]], implicit [[V_ADD_U32_e64_6]], implicit [[V_ADD_U32_e64_7]], implicit [[V_ADD_U32_e64_8]], implicit [[V_ADD_U32_e64_9]], implicit [[V_ADD_U32_e64_10]], implicit [[V_ADD_U32_e64_11]], implicit [[V_ADD_U32_e64_12]], implicit [[V_ADD_U32_e64_13]], implicit [[V_ADD_U32_e64_14]], implicit [[V_ADD_U32_e64_15]], implicit [[V_ADD_U32_e64_16]], implicit [[V_ADD_U32_e64_17]], implicit [[V_ADD_U32_e64_18]], implicit [[V_ADD_U32_e64_19]], implicit [[V_ADD_U32_e64_20]], implicit [[V_ADD_U32_e64_21]], implicit [[V_ADD_U32_e64_22]], implicit [[V_ADD_U32_e64_23]], implicit [[V_ADD_U32_e64_24]], implicit [[V_ADD_U32_e64_25]], implicit [[V_ADD_U32_e64_26]]
1254+
; GFX10-SUNK-NEXT: INLINEASM &"", 1 /* sideeffect attdialect */, implicit [[V_ADD_U32_e64_27]], implicit [[V_ADD_U32_e64_28]], implicit [[V_ADD_U32_e64_29]], implicit [[V_ADD_U32_e64_30]], implicit [[V_ADD_U32_e64_31]], implicit [[V_ADD_U32_e64_32]], implicit [[V_ADD_U32_e64_33]], implicit [[V_ADD_U32_e64_34]], implicit [[V_ADD_U32_e64_35]], implicit [[V_ADD_U32_e64_36]], implicit [[V_ADD_U32_e64_37]], implicit [[V_ADD_U32_e64_38]], implicit [[V_ADD_U32_e64_39]], implicit [[V_ADD_U32_e64_40]], implicit [[V_ADD_U32_e64_41]], implicit [[V_ADD_U32_e64_42]], implicit [[V_ADD_U32_e64_43]], implicit [[V_ADD_U32_e64_44]], implicit [[V_ADD_U32_e64_45]], implicit [[V_ADD_U32_e64_46]], implicit [[V_ADD_U32_e64_47]], implicit [[V_ADD_U32_e64_48]], implicit [[V_ADD_U32_e64_49]], implicit [[V_ADD_U32_e64_50]], implicit [[V_ADD_U32_e64_51]], implicit [[V_ADD_U32_e64_52]], implicit [[V_ADD_U32_e64_53]]
12011255
; GFX10-SUNK-NEXT: S_BRANCH %bb.4
12021256
; GFX10-SUNK-NEXT: {{ $}}
12031257
; GFX10-SUNK-NEXT: bb.4:

0 commit comments

Comments
 (0)