|
| 1 | +# RUN: llc -mtriple=thumbv8.1m.main -mattr=+lob -run-pass=arm-low-overhead-loops --verify-machineinstrs %s -o - | FileCheck %s |
| 2 | + |
| 3 | +# There are 2 SUBS, and the 2nd one is identified as the def. |
| 4 | +# Thus, the 1st is a use, and we shouldn't optimise away the SUBS. |
| 5 | + |
| 6 | +# CHECK: bb.1.vector.body: |
| 7 | +# CHECK: renamable $r3, dead $cpsr = tSUBi8 killed renamable $r3, 4, 14, $noreg |
| 8 | +# CHECK: renamable $r3, dead $cpsr = tSUBi8 killed renamable $r3, 4, 14, $noreg |
| 9 | +# CHECK: $lr = MVE_LETP renamable $lr, %bb.1 |
| 10 | + |
| 11 | +--- | |
| 12 | + target datalayout = "e-m:e-p:32:32-Fi8-i64:64-v128:64:128-a:0:32-n32-S64" |
| 13 | + target triple = "thumbv8.1m.main-arm-unknown-eabi" |
| 14 | + |
| 15 | + define dso_local void @use_before_def(i32* noalias nocapture %A, i32* noalias nocapture readonly %B, i32* noalias nocapture readonly %C, i32 %N) local_unnamed_addr #0 { |
| 16 | + entry: |
| 17 | + %cmp8 = icmp sgt i32 %N, 0 |
| 18 | + %0 = add i32 %N, 3 |
| 19 | + %1 = lshr i32 %0, 2 |
| 20 | + %2 = shl nuw i32 %1, 2 |
| 21 | + %3 = add i32 %2, -4 |
| 22 | + %4 = lshr i32 %3, 2 |
| 23 | + %5 = add nuw nsw i32 %4, 1 |
| 24 | + br i1 %cmp8, label %vector.ph, label %for.cond.cleanup |
| 25 | + |
| 26 | + vector.ph: ; preds = %entry |
| 27 | + call void @llvm.set.loop.iterations.i32(i32 %5) |
| 28 | + br label %vector.body |
| 29 | + |
| 30 | + vector.body: ; preds = %vector.body, %vector.ph |
| 31 | + %lsr.iv17 = phi i32* [ %scevgep18, %vector.body ], [ %A, %vector.ph ] |
| 32 | + %lsr.iv14 = phi i32* [ %scevgep15, %vector.body ], [ %C, %vector.ph ] |
| 33 | + %lsr.iv = phi i32* [ %scevgep, %vector.body ], [ %B, %vector.ph ] |
| 34 | + %6 = phi i32 [ %5, %vector.ph ], [ %11, %vector.body ] |
| 35 | + %7 = phi i32 [ %N, %vector.ph ], [ %9, %vector.body ] |
| 36 | + %lsr.iv13 = bitcast i32* %lsr.iv to <4 x i32>* |
| 37 | + %lsr.iv1416 = bitcast i32* %lsr.iv14 to <4 x i32>* |
| 38 | + %lsr.iv1719 = bitcast i32* %lsr.iv17 to <4 x i32>* |
| 39 | + %8 = call <4 x i1> @llvm.arm.vctp32(i32 %7) |
| 40 | + %9 = sub i32 %7, 4 |
| 41 | + %wide.masked.load = call <4 x i32> @llvm.masked.load.v4i32.p0v4i32(<4 x i32>* %lsr.iv13, i32 4, <4 x i1> %8, <4 x i32> undef), !tbaa !3 |
| 42 | + %wide.masked.load12 = call <4 x i32> @llvm.masked.load.v4i32.p0v4i32(<4 x i32>* %lsr.iv1416, i32 4, <4 x i1> %8, <4 x i32> undef), !tbaa !3 |
| 43 | + %10 = add nsw <4 x i32> %wide.masked.load12, %wide.masked.load |
| 44 | + call void @llvm.masked.store.v4i32.p0v4i32(<4 x i32> %10, <4 x i32>* %lsr.iv1719, i32 4, <4 x i1> %8), !tbaa !3 |
| 45 | + %scevgep = getelementptr i32, i32* %lsr.iv, i32 4 |
| 46 | + %scevgep15 = getelementptr i32, i32* %lsr.iv14, i32 4 |
| 47 | + %scevgep18 = getelementptr i32, i32* %lsr.iv17, i32 4 |
| 48 | + %11 = call i32 @llvm.loop.decrement.reg.i32.i32.i32(i32 %6, i32 1) |
| 49 | + %12 = icmp ne i32 %11, 0 |
| 50 | + br i1 %12, label %vector.body, label %for.cond.cleanup, !llvm.loop !7 |
| 51 | + |
| 52 | + for.cond.cleanup: ; preds = %vector.body, %entry |
| 53 | + ret void |
| 54 | + } |
| 55 | + declare void @llvm.set.loop.iterations.i32(i32) #1 |
| 56 | + declare <4 x i1> @llvm.arm.vctp32(i32) #2 |
| 57 | + declare i32 @llvm.loop.decrement.reg.i32.i32.i32(i32, i32) #1 |
| 58 | + declare void @llvm.lifetime.start.p0i8(i64 immarg, i8* nocapture) #3 |
| 59 | + declare <4 x i32> @llvm.masked.load.v4i32.p0v4i32(<4 x i32>*, i32 immarg, <4 x i1>, <4 x i32>) #4 |
| 60 | + declare void @llvm.masked.store.v4i32.p0v4i32(<4 x i32>, <4 x i32>*, i32 immarg, <4 x i1>) #3 |
| 61 | + declare void @llvm.stackprotector(i8*, i8**) #5 |
| 62 | + |
| 63 | + attributes #0 = { nofree norecurse nounwind "correctly-rounded-divide-sqrt-fp-math"="false" "disable-tail-calls"="false" "frame-pointer"="all" "less-precise-fpmad"="false" "min-legal-vector-width"="0" "no-infs-fp-math"="true" "no-jump-tables"="false" "no-nans-fp-math"="true" "no-signed-zeros-fp-math"="true" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="generic" "target-features"="+armv8.1-m.main,+fp-armv8d16sp,+fp16,+fpregs,+fullfp16,+hwdiv,+lob,+mve.fp,+ras,+strict-align,+thumb-mode,+vfp2sp,+vfp3d16sp,+vfp4d16sp" "unsafe-fp-math"="true" "use-soft-float"="false" } |
| 64 | + attributes #1 = { noduplicate nounwind } |
| 65 | + attributes #2 = { nounwind readnone } |
| 66 | + attributes #3 = { argmemonly nounwind willreturn } |
| 67 | + attributes #4 = { argmemonly nounwind readonly willreturn } |
| 68 | + attributes #5 = { nounwind } |
| 69 | + |
| 70 | + !llvm.module.flags = !{!0, !1} |
| 71 | + !llvm.ident = !{!2} |
| 72 | + |
| 73 | + !0 = !{i32 1, !"wchar_size", i32 4} |
| 74 | + !1 = !{i32 1, !"min_enum_size", i32 4} |
| 75 | + !2 = !{!"clang version 10.0.0 (http://github.com/llvm/llvm-project 2589b6d9edda73280fe1dc1d944ee34e22fe9a6f)"} |
| 76 | + !3 = !{!4, !4, i64 0} |
| 77 | + !4 = !{!"int", !5, i64 0} |
| 78 | + !5 = !{!"omnipotent char", !6, i64 0} |
| 79 | + !6 = !{!"Simple C++ TBAA"} |
| 80 | + !7 = distinct !{!7, !8} |
| 81 | + !8 = !{!"llvm.loop.isvectorized", i32 1} |
| 82 | + |
| 83 | +... |
| 84 | +--- |
| 85 | +name: use_before_def |
| 86 | +alignment: 2 |
| 87 | +exposesReturnsTwice: false |
| 88 | +legalized: false |
| 89 | +regBankSelected: false |
| 90 | +selected: false |
| 91 | +failedISel: false |
| 92 | +tracksRegLiveness: true |
| 93 | +hasWinCFI: false |
| 94 | +registers: [] |
| 95 | +liveins: |
| 96 | + - { reg: '$r0', virtual-reg: '' } |
| 97 | + - { reg: '$r1', virtual-reg: '' } |
| 98 | + - { reg: '$r2', virtual-reg: '' } |
| 99 | + - { reg: '$r3', virtual-reg: '' } |
| 100 | +frameInfo: |
| 101 | + isFrameAddressTaken: false |
| 102 | + isReturnAddressTaken: false |
| 103 | + hasStackMap: false |
| 104 | + hasPatchPoint: false |
| 105 | + stackSize: 8 |
| 106 | + offsetAdjustment: 0 |
| 107 | + maxAlignment: 4 |
| 108 | + adjustsStack: false |
| 109 | + hasCalls: false |
| 110 | + stackProtector: '' |
| 111 | + maxCallFrameSize: 0 |
| 112 | + cvBytesOfCalleeSavedRegisters: 0 |
| 113 | + hasOpaqueSPAdjustment: false |
| 114 | + hasVAStart: false |
| 115 | + hasMustTailInVarArgFunc: false |
| 116 | + localFrameSize: 0 |
| 117 | + savePoint: '' |
| 118 | + restorePoint: '' |
| 119 | +fixedStack: [] |
| 120 | +stack: |
| 121 | + - { id: 0, name: '', type: spill-slot, offset: -4, size: 4, alignment: 4, |
| 122 | + stack-id: default, callee-saved-register: '$lr', callee-saved-restored: false, |
| 123 | + debug-info-variable: '', debug-info-expression: '', debug-info-location: '' } |
| 124 | + - { id: 1, name: '', type: spill-slot, offset: -8, size: 4, alignment: 4, |
| 125 | + stack-id: default, callee-saved-register: '$r7', callee-saved-restored: true, |
| 126 | + debug-info-variable: '', debug-info-expression: '', debug-info-location: '' } |
| 127 | +callSites: [] |
| 128 | +constants: [] |
| 129 | +machineFunctionInfo: {} |
| 130 | +body: | |
| 131 | + bb.0.entry: |
| 132 | + successors: %bb.1(0x80000000) |
| 133 | + liveins: $r0, $r1, $r2, $r3, $lr |
| 134 | +
|
| 135 | + frame-setup tPUSH 14, $noreg, $r7, killed $lr, implicit-def $sp, implicit $sp |
| 136 | + frame-setup CFI_INSTRUCTION def_cfa_offset 8 |
| 137 | + frame-setup CFI_INSTRUCTION offset $lr, -4 |
| 138 | + frame-setup CFI_INSTRUCTION offset $r7, -8 |
| 139 | + $r7 = frame-setup tMOVr $sp, 14, $noreg |
| 140 | + frame-setup CFI_INSTRUCTION def_cfa_register $r7 |
| 141 | + tCMPi8 renamable $r3, 1, 14, $noreg, implicit-def $cpsr |
| 142 | + t2IT 11, 8, implicit-def $itstate |
| 143 | + tPOP_RET 11, killed $cpsr, def $r7, def $pc, implicit killed $itstate |
| 144 | + renamable $r12 = t2ADDri renamable $r3, 3, 14, $noreg, $noreg |
| 145 | + renamable $lr = t2MOVi 1, 14, $noreg, $noreg |
| 146 | + renamable $r12 = t2BICri killed renamable $r12, 3, 14, $noreg, $noreg |
| 147 | + renamable $r12 = t2SUBri killed renamable $r12, 4, 14, $noreg, $noreg |
| 148 | + renamable $lr = nuw nsw t2ADDrs killed renamable $lr, killed renamable $r12, 19, 14, $noreg, $noreg |
| 149 | + t2DoLoopStart renamable $lr |
| 150 | +
|
| 151 | + bb.1.vector.body: |
| 152 | + successors: %bb.1(0x7c000000), %bb.2(0x04000000) |
| 153 | + liveins: $lr, $r0, $r1, $r2, $r3 |
| 154 | +
|
| 155 | + renamable $vpr = MVE_VCTP32 renamable $r3, 0, $noreg |
| 156 | + MVE_VPST 4, implicit $vpr |
| 157 | + renamable $r1, renamable $q0 = MVE_VLDRWU32_post killed renamable $r1, 16, 1, renamable $vpr :: (load 16 from %ir.lsr.iv13, align 4, !tbaa !3) |
| 158 | + renamable $r2, renamable $q1 = MVE_VLDRWU32_post killed renamable $r2, 16, 1, renamable $vpr :: (load 16 from %ir.lsr.iv1416, align 4, !tbaa !3) |
| 159 | + renamable $r3, dead $cpsr = tSUBi8 killed renamable $r3, 4, 14, $noreg |
| 160 | + renamable $q0 = nsw MVE_VADDi32 killed renamable $q1, killed renamable $q0, 0, $noreg, undef renamable $q0 |
| 161 | + MVE_VPST 8, implicit $vpr |
| 162 | + renamable $r0 = MVE_VSTRWU32_post killed renamable $q0, killed renamable $r0, 16, 1, killed renamable $vpr :: (store 16 into %ir.lsr.iv1719, align 4, !tbaa !3) |
| 163 | + renamable $lr = t2LoopDec killed renamable $lr, 1 |
| 164 | + renamable $r3, dead $cpsr = tSUBi8 killed renamable $r3, 4, 14, $noreg |
| 165 | + t2LoopEnd renamable $lr, %bb.1, implicit-def dead $cpsr |
| 166 | + tB %bb.2, 14, $noreg |
| 167 | +
|
| 168 | + bb.2.for.cond.cleanup: |
| 169 | + tPOP_RET 14, $noreg, def $r7, def $pc |
| 170 | +
|
| 171 | +... |
0 commit comments