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
4 # Test that a VPNOT is not added to a max sized VPT block.
7 define dso_local void @inloop_vpnot(ptr nocapture readonly %a, ptr nocapture readonly %b, ptr nocapture readonly %c, ptr nocapture readonly %d, ptr nocapture %e, i32 %N) local_unnamed_addr #0 {
9 %cmp9 = icmp eq i32 %N, 0
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
18 vector.ph: ; preds = %entry
19 %start = call i32 @llvm.start.loop.iterations.i32(i32 %tmp5)
22 vector.body: ; preds = %vector.body, %vector.ph
23 %lsr.iv1 = phi i32 [ %lsr.iv.next, %vector.body ], [ %start, %vector.ph ]
24 %lsr.iv.e = phi ptr [ %scevgep.e, %vector.body ], [ %e, %vector.ph ]
25 %lsr.iv.d = phi ptr [ %scevgep.d, %vector.body ], [ %d, %vector.ph ]
26 %lsr.iv.c = phi ptr [ %scevgep.c, %vector.body ], [ %c, %vector.ph ]
27 %lsr.iv18 = phi ptr [ %scevgep19, %vector.body ], [ %b, %vector.ph ]
28 %lsr.iv = phi ptr [ %scevgep, %vector.body ], [ %a, %vector.ph ]
29 %vec.phi = phi <4 x i32> [ zeroinitializer, %vector.ph ], [ %tmp14, %vector.body ]
30 %tmp7 = phi i32 [ %N, %vector.ph ], [ %tmp9, %vector.body ]
31 %lsr.iv17 = bitcast ptr %lsr.iv to ptr
32 %lsr.iv1820 = bitcast ptr %lsr.iv18 to ptr
33 %lsr.iv1820.c = bitcast ptr %lsr.iv.c to ptr
34 %lsr.iv17.d = bitcast ptr %lsr.iv.d to ptr
35 %lsr.cast.e = bitcast ptr %lsr.iv.e to ptr
36 %tmp8 = call <4 x i1> @llvm.arm.mve.vctp32(i32 %tmp7)
37 %tmp9 = sub i32 %tmp7, 4
38 %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)
39 %tmp10 = sext <4 x i16> %wide.masked.load to <4 x i32>
40 %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)
41 %tmp11 = sext <4 x i16> %wide.masked.load14 to <4 x i32>
42 %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)
43 %sext.load.c = sext <4 x i16> %wide.masked.load.c to <4 x i32>
44 %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)
45 %sext.load.d = sext <4 x i16> %wide.masked.load.d to <4 x i32>
46 %tmp12 = mul nsw <4 x i32> %tmp11, %tmp10
47 %mul.2 = mul nsw <4 x i32> %sext.load.c, %sext.load.d
48 %tmp13 = add <4 x i32> %tmp12, %mul.2
49 %tmp14 = add <4 x i32> %tmp13, %vec.phi
50 %vpnot = xor <4 x i1> %tmp8, <i1 true, i1 true, i1 true, i1 true>
51 call void @llvm.masked.store.v4i32.p0(<4 x i32> %tmp14, ptr %lsr.cast.e, i32 4, <4 x i1> %vpnot)
52 %scevgep = getelementptr i16, ptr %lsr.iv, i32 4
53 %scevgep19 = getelementptr i16, ptr %lsr.iv18, i32 4
54 %scevgep.c = getelementptr i16, ptr %lsr.iv.c, i32 4
55 %scevgep.d = getelementptr i16, ptr %lsr.iv.d, i32 4
56 %scevgep.e = getelementptr i32, ptr %lsr.iv.e, i32 4
57 %tmp15 = call i32 @llvm.loop.decrement.reg.i32.i32.i32(i32 %lsr.iv1, i32 1)
58 %tmp16 = icmp ne i32 %tmp15, 0
59 %lsr.iv.next = add nsw i32 %lsr.iv1, -1
60 br i1 %tmp16, label %vector.body, label %for.cond.cleanup
62 for.cond.cleanup: ; preds = %vector.body, %entry
65 declare <4 x i16> @llvm.masked.load.v4i16.p0(ptr, i32 immarg, <4 x i1>, <4 x i16>) #1
66 declare void @llvm.masked.store.v4i32.p0(<4 x i32>, ptr, i32 immarg, <4 x i1>) #2
67 declare i32 @llvm.start.loop.iterations.i32(i32) #3
68 declare i32 @llvm.loop.decrement.reg.i32.i32.i32(i32, i32) #3
69 declare <4 x i1> @llvm.arm.mve.vctp32(i32) #4
75 exposesReturnsTwice: false
77 regBankSelected: false
80 tracksRegLiveness: true
84 - { reg: '$r0', virtual-reg: '' }
85 - { reg: '$r1', virtual-reg: '' }
86 - { reg: '$r2', virtual-reg: '' }
87 - { reg: '$r3', virtual-reg: '' }
89 isFrameAddressTaken: false
90 isReturnAddressTaken: false
100 cvBytesOfCalleeSavedRegisters: 0
101 hasOpaqueSPAdjustment: false
103 hasMustTailInVarArgFunc: false
108 - { id: 0, type: default, offset: 4, size: 4, alignment: 4, stack-id: default,
109 isImmutable: true, isAliased: false, callee-saved-register: '', callee-saved-restored: true,
110 debug-info-variable: '', debug-info-expression: '', debug-info-location: '' }
111 - { id: 1, type: default, offset: 0, size: 4, alignment: 8, stack-id: default,
112 isImmutable: true, isAliased: false, callee-saved-register: '', callee-saved-restored: true,
113 debug-info-variable: '', debug-info-expression: '', debug-info-location: '' }
115 - { id: 0, name: '', type: spill-slot, offset: -4, size: 4, alignment: 4,
116 stack-id: default, callee-saved-register: '$lr', callee-saved-restored: false,
117 debug-info-variable: '', debug-info-expression: '', debug-info-location: '' }
118 - { id: 1, name: '', type: spill-slot, offset: -8, size: 4, alignment: 4,
119 stack-id: default, callee-saved-register: '$r7', callee-saved-restored: true,
120 debug-info-variable: '', debug-info-expression: '', debug-info-location: '' }
121 - { id: 2, name: '', type: spill-slot, offset: -12, size: 4, alignment: 4,
122 stack-id: default, callee-saved-register: '$r5', callee-saved-restored: true,
123 debug-info-variable: '', debug-info-expression: '', debug-info-location: '' }
124 - { id: 3, name: '', type: spill-slot, offset: -16, size: 4, alignment: 4,
125 stack-id: default, callee-saved-register: '$r4', callee-saved-restored: true,
126 debug-info-variable: '', debug-info-expression: '', debug-info-location: '' }
129 machineFunctionInfo: {}
131 ; CHECK-LABEL: name: inloop_vpnot
133 ; CHECK-NEXT: successors: %bb.3(0x30000000), %bb.1(0x50000000)
134 ; CHECK-NEXT: liveins: $lr, $r0, $r1, $r2, $r3, $r4, $r5, $r7
136 ; CHECK-NEXT: frame-setup tPUSH 14 /* CC::al */, $noreg, killed $r4, killed $r5, killed $r7, killed $lr, implicit-def $sp, implicit $sp
137 ; CHECK-NEXT: frame-setup CFI_INSTRUCTION def_cfa_offset 16
138 ; CHECK-NEXT: frame-setup CFI_INSTRUCTION offset $lr, -4
139 ; CHECK-NEXT: frame-setup CFI_INSTRUCTION offset $r7, -8
140 ; CHECK-NEXT: frame-setup CFI_INSTRUCTION offset $r5, -12
141 ; CHECK-NEXT: frame-setup CFI_INSTRUCTION offset $r4, -16
142 ; CHECK-NEXT: renamable $r12 = t2LDRi12 $sp, 20, 14 /* CC::al */, $noreg :: (load (s32) from %fixed-stack.1)
143 ; CHECK-NEXT: t2CMPri renamable $r12, 0, 14 /* CC::al */, $noreg, implicit-def $cpsr
144 ; CHECK-NEXT: tBcc %bb.3, 0 /* CC::eq */, killed $cpsr
146 ; CHECK-NEXT: bb.1.vector.ph:
147 ; CHECK-NEXT: successors: %bb.2(0x80000000)
148 ; CHECK-NEXT: liveins: $r0, $r1, $r2, $r3, $r12
150 ; CHECK-NEXT: renamable $lr = t2ADDri renamable $r12, 3, 14 /* CC::al */, $noreg, $noreg
151 ; CHECK-NEXT: renamable $r4, dead $cpsr = tMOVi8 1, 14 /* CC::al */, $noreg
152 ; CHECK-NEXT: renamable $lr = t2BICri killed renamable $lr, 3, 14 /* CC::al */, $noreg, $noreg
153 ; CHECK-NEXT: renamable $r5 = tLDRspi $sp, 4, 14 /* CC::al */, $noreg :: (load (s32) from %fixed-stack.0, align 8)
154 ; CHECK-NEXT: renamable $lr = t2SUBri killed renamable $lr, 4, 14 /* CC::al */, $noreg, $noreg
155 ; CHECK-NEXT: renamable $q0 = MVE_VMOVimmi32 0, 0, $noreg, $noreg, undef renamable $q0
156 ; CHECK-NEXT: renamable $lr = nuw nsw t2ADDrs killed renamable $r4, killed renamable $lr, 19, 14 /* CC::al */, $noreg, $noreg
157 ; CHECK-NEXT: $r4 = tMOVr killed $lr, 14 /* CC::al */, $noreg
159 ; CHECK-NEXT: bb.2.vector.body:
160 ; CHECK-NEXT: successors: %bb.2(0x7c000000), %bb.3(0x04000000)
161 ; CHECK-NEXT: liveins: $q0, $r0, $r1, $r2, $r3, $r4, $r5, $r12
163 ; CHECK-NEXT: renamable $vpr = MVE_VCTP32 renamable $r12, 0, $noreg, $noreg
164 ; CHECK-NEXT: MVE_VPST 4, implicit $vpr
165 ; 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)
166 ; 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)
167 ; CHECK-NEXT: renamable $q1 = nsw MVE_VMULi32 killed renamable $q2, killed renamable $q1, 0, $noreg, $noreg, undef renamable $q1
168 ; CHECK-NEXT: MVE_VPST 4, implicit $vpr
169 ; 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)
170 ; 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)
171 ; CHECK-NEXT: $lr = tMOVr $r4, 14 /* CC::al */, $noreg
172 ; CHECK-NEXT: renamable $r4, dead $cpsr = nsw tSUBi8 killed $r4, 1, 14 /* CC::al */, $noreg
173 ; CHECK-NEXT: renamable $r12 = t2SUBri killed renamable $r12, 4, 14 /* CC::al */, $noreg, $noreg
174 ; CHECK-NEXT: MVE_VPST 2, implicit $vpr
175 ; CHECK-NEXT: renamable $q2 = nsw MVE_VMULi32 killed renamable $q3, killed renamable $q2, 0, $noreg, $noreg, undef renamable $q2
176 ; CHECK-NEXT: renamable $q1 = MVE_VADDi32 killed renamable $q2, killed renamable $q1, 0, renamable $vpr, $noreg, undef renamable $q1
177 ; CHECK-NEXT: renamable $q0 = MVE_VADDi32 killed renamable $q1, killed renamable $q0, 0, renamable $vpr, $noreg, undef renamable $q0
178 ; CHECK-NEXT: renamable $r5 = MVE_VSTRWU32_post renamable $q0, killed renamable $r5, 16, 1, renamable $vpr, $noreg :: (store (s128) into %ir.lsr.cast.e, align 4)
179 ; CHECK-NEXT: dead renamable $vpr = MVE_VPNOT killed renamable $vpr, 0, $noreg, $noreg
180 ; CHECK-NEXT: dead $lr = t2LEUpdate killed renamable $lr, %bb.2
182 ; CHECK-NEXT: bb.3.for.cond.cleanup:
183 ; CHECK-NEXT: tPOP_RET 14 /* CC::al */, $noreg, def $r4, def $r5, def $r7, def $pc
185 successors: %bb.3(0x30000000), %bb.1(0x50000000)
186 liveins: $r0, $r1, $r2, $r3, $r4, $r5, $r7, $lr
188 frame-setup tPUSH 14, $noreg, killed $r4, killed $r5, killed $r7, killed $lr, implicit-def $sp, implicit $sp
189 frame-setup CFI_INSTRUCTION def_cfa_offset 16
190 frame-setup CFI_INSTRUCTION offset $lr, -4
191 frame-setup CFI_INSTRUCTION offset $r7, -8
192 frame-setup CFI_INSTRUCTION offset $r5, -12
193 frame-setup CFI_INSTRUCTION offset $r4, -16
194 renamable $r12 = t2LDRi12 $sp, 20, 14, $noreg :: (load (s32) from %fixed-stack.0)
195 t2CMPri renamable $r12, 0, 14, $noreg, implicit-def $cpsr
196 tBcc %bb.3, 0, killed $cpsr
199 successors: %bb.2(0x80000000)
200 liveins: $r0, $r1, $r2, $r3, $r12
202 renamable $lr = t2ADDri renamable $r12, 3, 14, $noreg, $noreg
203 renamable $r4, dead $cpsr = tMOVi8 1, 14, $noreg
204 renamable $lr = t2BICri killed renamable $lr, 3, 14, $noreg, $noreg
205 renamable $r5 = tLDRspi $sp, 4, 14, $noreg :: (load (s32) from %fixed-stack.1, align 8)
206 renamable $lr = t2SUBri killed renamable $lr, 4, 14, $noreg, $noreg
207 renamable $q0 = MVE_VMOVimmi32 0, 0, $noreg, $noreg, undef renamable $q0
208 renamable $lr = nuw nsw t2ADDrs killed renamable $r4, killed renamable $lr, 19, 14, $noreg, $noreg
209 $lr = t2DoLoopStart renamable $lr
210 $r4 = tMOVr killed $lr, 14, $noreg
213 successors: %bb.2(0x7c000000), %bb.3(0x04000000)
214 liveins: $q0, $r0, $r1, $r2, $r3, $r4, $r5, $r12
216 renamable $vpr = MVE_VCTP32 renamable $r12, 0, $noreg, $noreg
217 MVE_VPST 4, implicit $vpr
218 renamable $r3, renamable $q1 = MVE_VLDRHS32_post killed renamable $r3, 8, 1, renamable $vpr, $noreg :: (load (s64) from %ir.lsr.iv17.d, align 2)
219 renamable $r2, renamable $q2 = MVE_VLDRHS32_post killed renamable $r2, 8, 1, renamable $vpr, $noreg :: (load (s64) from %ir.lsr.iv1820.c, align 2)
220 renamable $q1 = nsw MVE_VMULi32 killed renamable $q2, killed renamable $q1, 0, $noreg, $noreg, undef renamable $q1
221 MVE_VPST 4, implicit $vpr
222 renamable $r0, renamable $q2 = MVE_VLDRHS32_post killed renamable $r0, 8, 1, renamable $vpr, $noreg :: (load (s64) from %ir.lsr.iv17, align 2)
223 renamable $r1, renamable $q3 = MVE_VLDRHS32_post killed renamable $r1, 8, 1, renamable $vpr, $noreg :: (load (s64) from %ir.lsr.iv1820, align 2)
224 $lr = tMOVr $r4, 14, $noreg
225 renamable $r4, dead $cpsr = nsw tSUBi8 killed $r4, 1, 14, $noreg
226 renamable $r12 = t2SUBri killed renamable $r12, 4, 14, $noreg, $noreg
227 MVE_VPST 2, implicit $vpr
228 renamable $q2 = nsw MVE_VMULi32 killed renamable $q3, killed renamable $q2, 0, $noreg, $noreg, undef renamable $q2
229 renamable $q1 = MVE_VADDi32 killed renamable $q2, killed renamable $q1, 0, renamable $vpr, $noreg, undef renamable $q1
230 renamable $q0 = MVE_VADDi32 killed renamable $q1, killed renamable $q0, 0, renamable $vpr, $noreg, undef renamable $q0
231 renamable $r5 = MVE_VSTRWU32_post renamable $q0, killed renamable $r5, 16, 1, renamable $vpr, $noreg :: (store (s128) into %ir.lsr.cast.e, align 4)
232 renamable $vpr = MVE_VPNOT renamable $vpr, 0, $noreg, $noreg
233 renamable $lr = t2LoopDec killed renamable $lr, 1
234 t2LoopEnd killed renamable $lr, %bb.2, implicit-def dead $cpsr
237 bb.3.for.cond.cleanup:
238 tPOP_RET 14, $noreg, def $r4, def $r5, def $r7, def $pc