1# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py 2# RUN: llc -mtriple=thumbv8.1m.main -mattr=+mve -run-pass=arm-low-overhead-loops %s -o - --verify-machineinstrs | FileCheck %s 3 4# General test for vpsel exclusion from tail predication 5 6--- | 7 define dso_local i32 @vpsel_after_vpt(ptr nocapture readonly %a, ptr nocapture readonly %b, ptr nocapture readonly %c, ptr nocapture readonly %d, i32 %N) local_unnamed_addr #0 { 8 entry: 9 %cmp9 = icmp eq i32 %N, 0 10 %tmp = add i32 %N, 3 11 %tmp1 = lshr i32 %tmp, 2 12 %tmp2 = shl nuw i32 %tmp1, 2 13 %tmp3 = add i32 %tmp2, -4 14 %tmp4 = lshr i32 %tmp3, 2 15 %tmp5 = add nuw nsw i32 %tmp4, 1 16 br i1 %cmp9, label %for.cond.cleanup, label %vector.ph 17 18 vector.ph: ; preds = %entry 19 %start = call i32 @llvm.start.loop.iterations.i32(i32 %tmp5) 20 br label %vector.body 21 22 vector.body: ; preds = %vector.body, %vector.ph 23 %lsr.iv1 = phi i32 [ %lsr.iv.next, %vector.body ], [ %start, %vector.ph ] 24 %lsr.iv.d = phi ptr [ %scevgep.d, %vector.body ], [ %d, %vector.ph ] 25 %lsr.iv.c = phi ptr [ %scevgep.c, %vector.body ], [ %c, %vector.ph ] 26 %lsr.iv18 = phi ptr [ %scevgep19, %vector.body ], [ %b, %vector.ph ] 27 %lsr.iv = phi ptr [ %scevgep, %vector.body ], [ %a, %vector.ph ] 28 %vec.phi = phi <4 x i32> [ zeroinitializer, %vector.ph ], [ %tmp14, %vector.body ] 29 %tmp7 = phi i32 [ %N, %vector.ph ], [ %tmp9, %vector.body ] 30 %lsr.iv17 = bitcast ptr %lsr.iv to ptr 31 %lsr.iv1820 = bitcast ptr %lsr.iv18 to ptr 32 %lsr.iv1820.c = bitcast ptr %lsr.iv.c to ptr 33 %lsr.iv17.d = bitcast ptr %lsr.iv.d to ptr 34 %tmp8 = call <4 x i1> @llvm.arm.mve.vctp32(i32 %tmp7) 35 %tmp9 = sub i32 %tmp7, 4 36 %wide.masked.load = call <4 x i16> @llvm.masked.load.v4i16.p0(ptr %lsr.iv17, i32 2, <4 x i1> %tmp8, <4 x i16> undef) 37 %tmp10 = sext <4 x i16> %wide.masked.load to <4 x i32> 38 %wide.masked.load14 = call <4 x i16> @llvm.masked.load.v4i16.p0(ptr %lsr.iv1820, i32 2, <4 x i1> %tmp8, <4 x i16> undef) 39 %tmp11 = sext <4 x i16> %wide.masked.load14 to <4 x i32> 40 %wide.masked.load.c = call <4 x i16> @llvm.masked.load.v4i16.p0(ptr %lsr.iv1820.c, i32 2, <4 x i1> %tmp8, <4 x i16> undef) 41 %sext.load.c = sext <4 x i16> %wide.masked.load.c to <4 x i32> 42 %wide.masked.load.d = call <4 x i16> @llvm.masked.load.v4i16.p0(ptr %lsr.iv17.d, i32 2, <4 x i1> %tmp8, <4 x i16> undef) 43 %sext.load.d = sext <4 x i16> %wide.masked.load.d to <4 x i32> 44 %tmp12 = mul nsw <4 x i32> %tmp11, %tmp10 45 %mul.2 = mul nsw <4 x i32> %sext.load.c, %sext.load.d 46 %tmp13 = add <4 x i32> %tmp12, %mul.2 47 %acc = add <4 x i32> %tmp13, %vec.phi 48 %tmp14 = select <4 x i1> %tmp8, <4 x i32> %acc, <4 x i32> %vec.phi 49 %scevgep = getelementptr i16, ptr %lsr.iv, i32 4 50 %scevgep19 = getelementptr i16, ptr %lsr.iv18, i32 4 51 %scevgep.c = getelementptr i16, ptr %lsr.iv.c, i32 4 52 %scevgep.d = getelementptr i16, ptr %lsr.iv.d, i32 4 53 %tmp15 = call i32 @llvm.loop.decrement.reg.i32.i32.i32(i32 %lsr.iv1, i32 1) 54 %tmp16 = icmp ne i32 %tmp15, 0 55 %lsr.iv.next = add nsw i32 %lsr.iv1, -1 56 br i1 %tmp16, label %vector.body, label %middle.block 57 58 middle.block: ; preds = %vector.body 59 %tmp17 = call i32 @llvm.vector.reduce.add.v4i32(<4 x i32> %tmp14) 60 br label %for.cond.cleanup 61 62 for.cond.cleanup: ; preds = %middle.block, %entry 63 %res.0.lcssa = phi i32 [ 0, %entry ], [ %tmp17, %middle.block ] 64 ret i32 %res.0.lcssa 65 } 66 declare <4 x i16> @llvm.masked.load.v4i16.p0(ptr, i32 immarg, <4 x i1>, <4 x i16>) #1 67 declare i32 @llvm.vector.reduce.add.v4i32(<4 x i32>) #2 68 declare i32 @llvm.start.loop.iterations.i32(i32) #3 69 declare i32 @llvm.loop.decrement.reg.i32.i32.i32(i32, i32) #3 70 declare <4 x i1> @llvm.arm.mve.vctp32(i32) #4 71 72... 73--- 74name: vpsel_after_vpt 75alignment: 2 76exposesReturnsTwice: false 77legalized: false 78regBankSelected: false 79selected: false 80failedISel: false 81tracksRegLiveness: true 82hasWinCFI: false 83registers: [] 84liveins: 85 - { reg: '$r0', virtual-reg: '' } 86 - { reg: '$r1', virtual-reg: '' } 87 - { reg: '$r2', virtual-reg: '' } 88 - { reg: '$r3', virtual-reg: '' } 89frameInfo: 90 isFrameAddressTaken: false 91 isReturnAddressTaken: false 92 hasStackMap: false 93 hasPatchPoint: false 94 stackSize: 16 95 offsetAdjustment: 0 96 maxAlignment: 4 97 adjustsStack: false 98 hasCalls: false 99 stackProtector: '' 100 maxCallFrameSize: 0 101 cvBytesOfCalleeSavedRegisters: 0 102 hasOpaqueSPAdjustment: false 103 hasVAStart: false 104 hasMustTailInVarArgFunc: false 105 localFrameSize: 0 106 savePoint: '' 107 restorePoint: '' 108fixedStack: 109 - { id: 0, type: default, offset: 0, size: 4, alignment: 8, stack-id: default, 110 isImmutable: true, isAliased: false, callee-saved-register: '', callee-saved-restored: true, 111 debug-info-variable: '', debug-info-expression: '', debug-info-location: '' } 112stack: 113 - { id: 0, name: '', type: spill-slot, offset: -4, size: 4, alignment: 4, 114 stack-id: default, callee-saved-register: '$lr', callee-saved-restored: false, 115 debug-info-variable: '', debug-info-expression: '', debug-info-location: '' } 116 - { id: 1, name: '', type: spill-slot, offset: -8, size: 4, alignment: 4, 117 stack-id: default, callee-saved-register: '$r7', callee-saved-restored: true, 118 debug-info-variable: '', debug-info-expression: '', debug-info-location: '' } 119 - { id: 2, name: '', type: spill-slot, offset: -12, size: 4, alignment: 4, 120 stack-id: default, callee-saved-register: '$r5', callee-saved-restored: true, 121 debug-info-variable: '', debug-info-expression: '', debug-info-location: '' } 122 - { id: 3, name: '', type: spill-slot, offset: -16, size: 4, alignment: 4, 123 stack-id: default, callee-saved-register: '$r4', callee-saved-restored: true, 124 debug-info-variable: '', debug-info-expression: '', debug-info-location: '' } 125callSites: [] 126constants: [] 127machineFunctionInfo: {} 128body: | 129 ; CHECK-LABEL: name: vpsel_after_vpt 130 ; CHECK: bb.0.entry: 131 ; CHECK-NEXT: successors: %bb.4(0x30000000), %bb.1(0x50000000) 132 ; CHECK-NEXT: liveins: $lr, $r0, $r1, $r2, $r3, $r4, $r5, $r7 133 ; CHECK-NEXT: {{ $}} 134 ; CHECK-NEXT: frame-setup tPUSH 14 /* CC::al */, $noreg, killed $r4, killed $r5, killed $r7, killed $lr, implicit-def $sp, implicit $sp 135 ; CHECK-NEXT: frame-setup CFI_INSTRUCTION def_cfa_offset 16 136 ; CHECK-NEXT: frame-setup CFI_INSTRUCTION offset $lr, -4 137 ; CHECK-NEXT: frame-setup CFI_INSTRUCTION offset $r7, -8 138 ; CHECK-NEXT: frame-setup CFI_INSTRUCTION offset $r5, -12 139 ; CHECK-NEXT: frame-setup CFI_INSTRUCTION offset $r4, -16 140 ; CHECK-NEXT: renamable $r12 = t2LDRi12 $sp, 16, 14 /* CC::al */, $noreg :: (load (s32) from %fixed-stack.0, align 8) 141 ; CHECK-NEXT: t2CMPri renamable $r12, 0, 14 /* CC::al */, $noreg, implicit-def $cpsr 142 ; CHECK-NEXT: tBcc %bb.4, 0 /* CC::eq */, killed $cpsr 143 ; CHECK-NEXT: {{ $}} 144 ; CHECK-NEXT: bb.1.vector.ph: 145 ; CHECK-NEXT: successors: %bb.2(0x80000000) 146 ; CHECK-NEXT: liveins: $r0, $r1, $r2, $r3, $r12 147 ; CHECK-NEXT: {{ $}} 148 ; CHECK-NEXT: renamable $lr = t2ADDri renamable $r12, 3, 14 /* CC::al */, $noreg, $noreg 149 ; CHECK-NEXT: renamable $r4, dead $cpsr = tMOVi8 1, 14 /* CC::al */, $noreg 150 ; CHECK-NEXT: renamable $lr = t2BICri killed renamable $lr, 3, 14 /* CC::al */, $noreg, $noreg 151 ; CHECK-NEXT: renamable $q0 = MVE_VMOVimmi32 0, 0, $noreg, $noreg, undef renamable $q0 152 ; CHECK-NEXT: renamable $lr = t2SUBri killed renamable $lr, 4, 14 /* CC::al */, $noreg, $noreg 153 ; CHECK-NEXT: renamable $r5 = nuw nsw t2ADDrs killed renamable $r4, killed renamable $lr, 19, 14 /* CC::al */, $noreg, $noreg 154 ; CHECK-NEXT: dead $lr = t2DLS renamable $r5 155 ; CHECK-NEXT: $r4 = tMOVr killed $r5, 14 /* CC::al */, $noreg 156 ; CHECK-NEXT: {{ $}} 157 ; CHECK-NEXT: bb.2.vector.body: 158 ; CHECK-NEXT: successors: %bb.2(0x7c000000), %bb.3(0x04000000) 159 ; CHECK-NEXT: liveins: $q0, $r0, $r1, $r2, $r3, $r4, $r12 160 ; CHECK-NEXT: {{ $}} 161 ; CHECK-NEXT: renamable $vpr = MVE_VCTP32 renamable $r12, 0, $noreg, $noreg 162 ; CHECK-NEXT: MVE_VPST 4, implicit $vpr 163 ; CHECK-NEXT: renamable $r3, renamable $q1 = MVE_VLDRHS32_post killed renamable $r3, 8, 1, renamable $vpr, $noreg :: (load (s64) from %ir.lsr.iv17.d, align 2) 164 ; CHECK-NEXT: renamable $r2, renamable $q2 = MVE_VLDRHS32_post killed renamable $r2, 8, 1, renamable $vpr, $noreg :: (load (s64) from %ir.lsr.iv1820.c, align 2) 165 ; CHECK-NEXT: renamable $q1 = nsw MVE_VMULi32 killed renamable $q2, killed renamable $q1, 0, $noreg, $noreg, undef renamable $q1 166 ; CHECK-NEXT: MVE_VPST 4, implicit $vpr 167 ; CHECK-NEXT: renamable $r0, renamable $q2 = MVE_VLDRHS32_post killed renamable $r0, 8, 1, renamable $vpr, $noreg :: (load (s64) from %ir.lsr.iv17, align 2) 168 ; CHECK-NEXT: renamable $r1, renamable $q3 = MVE_VLDRHS32_post killed renamable $r1, 8, 1, renamable $vpr, $noreg :: (load (s64) from %ir.lsr.iv1820, align 2) 169 ; CHECK-NEXT: renamable $q2 = nsw MVE_VMULi32 killed renamable $q3, killed renamable $q2, 0, $noreg, $noreg, undef renamable $q2 170 ; CHECK-NEXT: $lr = tMOVr $r4, 14 /* CC::al */, $noreg 171 ; CHECK-NEXT: renamable $q1 = MVE_VADDi32 killed renamable $q2, killed renamable $q1, 0, $noreg, $noreg, undef renamable $q1 172 ; CHECK-NEXT: renamable $r4, dead $cpsr = nsw tSUBi8 killed $r4, 1, 14 /* CC::al */, $noreg 173 ; CHECK-NEXT: renamable $q1 = MVE_VADDi32 killed renamable $q1, renamable $q0, 0, $noreg, $noreg, undef renamable $q1 174 ; CHECK-NEXT: renamable $r12 = t2SUBri killed renamable $r12, 4, 14 /* CC::al */, $noreg, $noreg 175 ; CHECK-NEXT: renamable $q0 = MVE_VPSEL killed renamable $q1, killed renamable $q0, 0, killed renamable $vpr, $noreg 176 ; CHECK-NEXT: dead $lr = t2LEUpdate killed renamable $lr, %bb.2 177 ; CHECK-NEXT: {{ $}} 178 ; CHECK-NEXT: bb.3.middle.block: 179 ; CHECK-NEXT: liveins: $q0 180 ; CHECK-NEXT: {{ $}} 181 ; CHECK-NEXT: renamable $r0 = MVE_VADDVu32no_acc killed renamable $q0, 0, $noreg, $noreg 182 ; CHECK-NEXT: tPOP_RET 14 /* CC::al */, $noreg, def $r4, def $r5, def $r7, def $pc, implicit killed $r0 183 ; CHECK-NEXT: {{ $}} 184 ; CHECK-NEXT: bb.4: 185 ; CHECK-NEXT: renamable $r0, dead $cpsr = tMOVi8 0, 14 /* CC::al */, $noreg 186 ; CHECK-NEXT: tPOP_RET 14 /* CC::al */, $noreg, def $r4, def $r5, def $r7, def $pc, implicit killed $r0 187 bb.0.entry: 188 successors: %bb.4(0x30000000), %bb.1(0x50000000) 189 liveins: $r0, $r1, $r2, $r3, $r4, $r5, $r7, $lr 190 191 frame-setup tPUSH 14, $noreg, killed $r4, killed $r5, killed $r7, killed $lr, implicit-def $sp, implicit $sp 192 frame-setup CFI_INSTRUCTION def_cfa_offset 16 193 frame-setup CFI_INSTRUCTION offset $lr, -4 194 frame-setup CFI_INSTRUCTION offset $r7, -8 195 frame-setup CFI_INSTRUCTION offset $r5, -12 196 frame-setup CFI_INSTRUCTION offset $r4, -16 197 renamable $r12 = t2LDRi12 $sp, 16, 14, $noreg :: (load (s32) from %fixed-stack.0, align 8) 198 t2CMPri renamable $r12, 0, 14, $noreg, implicit-def $cpsr 199 tBcc %bb.4, 0, killed $cpsr 200 201 bb.1.vector.ph: 202 successors: %bb.2(0x80000000) 203 liveins: $r0, $r1, $r2, $r3, $r12 204 205 renamable $lr = t2ADDri renamable $r12, 3, 14, $noreg, $noreg 206 renamable $r4, dead $cpsr = tMOVi8 1, 14, $noreg 207 renamable $lr = t2BICri killed renamable $lr, 3, 14, $noreg, $noreg 208 renamable $q0 = MVE_VMOVimmi32 0, 0, $noreg, $noreg, undef renamable $q0 209 renamable $lr = t2SUBri killed renamable $lr, 4, 14, $noreg, $noreg 210 renamable $r5 = nuw nsw t2ADDrs killed renamable $r4, killed renamable $lr, 19, 14, $noreg, $noreg 211 $lr = t2DoLoopStart renamable $r5 212 $r4 = tMOVr killed $r5, 14, $noreg 213 214 bb.2.vector.body: 215 successors: %bb.2(0x7c000000), %bb.3(0x04000000) 216 liveins: $q0, $r0, $r1, $r2, $r3, $r4, $r12 217 218 renamable $vpr = MVE_VCTP32 renamable $r12, 0, $noreg, $noreg 219 MVE_VPST 4, implicit $vpr 220 renamable $r3, renamable $q1 = MVE_VLDRHS32_post killed renamable $r3, 8, 1, renamable $vpr, $noreg :: (load (s64) from %ir.lsr.iv17.d, align 2) 221 renamable $r2, renamable $q2 = MVE_VLDRHS32_post killed renamable $r2, 8, 1, renamable $vpr, $noreg :: (load (s64) from %ir.lsr.iv1820.c, align 2) 222 renamable $q1 = nsw MVE_VMULi32 killed renamable $q2, killed renamable $q1, 0, $noreg, $noreg, undef renamable $q1 223 MVE_VPST 4, implicit $vpr 224 renamable $r0, renamable $q2 = MVE_VLDRHS32_post killed renamable $r0, 8, 1, renamable $vpr, $noreg :: (load (s64) from %ir.lsr.iv17, align 2) 225 renamable $r1, renamable $q3 = MVE_VLDRHS32_post killed renamable $r1, 8, 1, renamable $vpr, $noreg :: (load (s64) from %ir.lsr.iv1820, align 2) 226 renamable $q2 = nsw MVE_VMULi32 killed renamable $q3, killed renamable $q2, 0, $noreg, $noreg, undef renamable $q2 227 $lr = tMOVr $r4, 14, $noreg 228 renamable $q1 = MVE_VADDi32 killed renamable $q2, killed renamable $q1, 0, $noreg, $noreg, undef renamable $q1 229 renamable $r4, dead $cpsr = nsw tSUBi8 killed $r4, 1, 14, $noreg 230 renamable $q1 = MVE_VADDi32 killed renamable $q1, renamable $q0, 0, $noreg, $noreg, undef renamable $q1 231 renamable $r12 = t2SUBri killed renamable $r12, 4, 14, $noreg, $noreg 232 renamable $q0 = MVE_VPSEL killed renamable $q1, killed renamable $q0, 0, killed renamable $vpr, $noreg 233 renamable $lr = t2LoopDec killed renamable $lr, 1 234 t2LoopEnd killed renamable $lr, %bb.2, implicit-def dead $cpsr 235 tB %bb.3, 14, $noreg 236 237 bb.3.middle.block: 238 liveins: $q0 239 240 renamable $r0 = MVE_VADDVu32no_acc killed renamable $q0, 0, $noreg, $noreg 241 tPOP_RET 14, $noreg, def $r4, def $r5, def $r7, def $pc, implicit killed $r0 242 243 bb.4: 244 renamable $r0, dead $cpsr = tMOVi8 0, 14, $noreg 245 tPOP_RET 14, $noreg, def $r4, def $r5, def $r7, def $pc, implicit killed $r0 246 247... 248