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, even if vpsels are allowed, that they aren't inserted into a vpt
5 # block - which would cause an assert here because of the number of insts in
9 define dso_local i32 @vpsel_after_vpt(i16* nocapture readonly %a, i16* nocapture readonly %b, i16* nocapture readonly %c, i16* nocapture readonly %d, i32 %N) local_unnamed_addr #0 {
11 %cmp9 = icmp eq i32 %N, 0
13 %tmp1 = lshr i32 %tmp, 2
14 %tmp2 = shl nuw i32 %tmp1, 2
15 %tmp3 = add i32 %tmp2, -4
16 %tmp4 = lshr i32 %tmp3, 2
17 %tmp5 = add nuw nsw i32 %tmp4, 1
18 br i1 %cmp9, label %for.cond.cleanup, label %vector.ph
20 vector.ph: ; preds = %entry
21 %start = call i32 @llvm.start.loop.iterations.i32(i32 %tmp5)
24 vector.body: ; preds = %vector.body, %vector.ph
25 %lsr.iv1 = phi i32 [ %lsr.iv.next, %vector.body ], [ %start, %vector.ph ]
26 %lsr.iv.d = phi i16* [ %scevgep.d, %vector.body ], [ %d, %vector.ph ]
27 %lsr.iv.c = phi i16* [ %scevgep.c, %vector.body ], [ %c, %vector.ph ]
28 %lsr.iv18 = phi i16* [ %scevgep19, %vector.body ], [ %b, %vector.ph ]
29 %lsr.iv = phi i16* [ %scevgep, %vector.body ], [ %a, %vector.ph ]
30 %vec.phi = phi <4 x i32> [ zeroinitializer, %vector.ph ], [ %tmp14, %vector.body ]
31 %tmp7 = phi i32 [ %N, %vector.ph ], [ %tmp9, %vector.body ]
32 %lsr.iv17 = bitcast i16* %lsr.iv to <4 x i16>*
33 %lsr.iv1820 = bitcast i16* %lsr.iv18 to <4 x i16>*
34 %lsr.iv1820.c = bitcast i16* %lsr.iv.c to <4 x i16>*
35 %lsr.iv17.d = bitcast i16* %lsr.iv.d to <4 x i16>*
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.p0v4i16(<4 x i16>* %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.p0v4i16(<4 x i16>* %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.p0v4i16(<4 x i16>* %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.p0v4i16(<4 x i16>* %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 %acc = add <4 x i32> %tmp13, %vec.phi
50 %tmp14 = select <4 x i1> %tmp8, <4 x i32> %acc, <4 x i32> %vec.phi
51 %scevgep = getelementptr i16, i16* %lsr.iv, i32 4
52 %scevgep19 = getelementptr i16, i16* %lsr.iv18, i32 4
53 %scevgep.c = getelementptr i16, i16* %lsr.iv.c, i32 4
54 %scevgep.d = getelementptr i16, i16* %lsr.iv.d, i32 4
55 %tmp15 = call i32 @llvm.loop.decrement.reg.i32.i32.i32(i32 %lsr.iv1, i32 1)
56 %tmp16 = icmp ne i32 %tmp15, 0
57 %lsr.iv.next = add nsw i32 %lsr.iv1, -1
58 br i1 %tmp16, label %vector.body, label %middle.block
60 middle.block: ; preds = %vector.body
61 %tmp17 = call i32 @llvm.vector.reduce.add.v4i32(<4 x i32> %tmp14)
62 br label %for.cond.cleanup
64 for.cond.cleanup: ; preds = %middle.block, %entry
65 %res.0.lcssa = phi i32 [ 0, %entry ], [ %tmp17, %middle.block ]
68 declare <4 x i16> @llvm.masked.load.v4i16.p0v4i16(<4 x i16>*, i32 immarg, <4 x i1>, <4 x i16>) #1
69 declare i32 @llvm.vector.reduce.add.v4i32(<4 x i32>) #2
70 declare i32 @llvm.start.loop.iterations.i32(i32) #3
71 declare i32 @llvm.loop.decrement.reg.i32.i32.i32(i32, i32) #3
72 declare <4 x i1> @llvm.arm.mve.vctp32(i32) #4
78 exposesReturnsTwice: false
80 regBankSelected: false
83 tracksRegLiveness: true
87 - { reg: '$r0', virtual-reg: '' }
88 - { reg: '$r1', virtual-reg: '' }
89 - { reg: '$r2', virtual-reg: '' }
90 - { reg: '$r3', virtual-reg: '' }
92 isFrameAddressTaken: false
93 isReturnAddressTaken: false
103 cvBytesOfCalleeSavedRegisters: 0
104 hasOpaqueSPAdjustment: false
106 hasMustTailInVarArgFunc: false
111 - { id: 0, 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: vpsel_after_vpt
133 ; CHECK: successors: %bb.4(0x30000000), %bb.1(0x50000000)
134 ; CHECK: liveins: $lr, $r0, $r1, $r2, $r3, $r4, $r5, $r7
135 ; CHECK: frame-setup tPUSH 14 /* CC::al */, $noreg, killed $r4, killed $r5, killed $r7, killed $lr, implicit-def $sp, implicit $sp
136 ; CHECK: frame-setup CFI_INSTRUCTION def_cfa_offset 16
137 ; CHECK: frame-setup CFI_INSTRUCTION offset $lr, -4
138 ; CHECK: frame-setup CFI_INSTRUCTION offset $r7, -8
139 ; CHECK: frame-setup CFI_INSTRUCTION offset $r5, -12
140 ; CHECK: frame-setup CFI_INSTRUCTION offset $r4, -16
141 ; CHECK: renamable $r12 = t2LDRi12 $sp, 16, 14 /* CC::al */, $noreg :: (load (s32) from %fixed-stack.0, align 8)
142 ; CHECK: t2CMPri renamable $r12, 0, 14 /* CC::al */, $noreg, implicit-def $cpsr
143 ; CHECK: tBcc %bb.4, 0 /* CC::eq */, killed $cpsr
144 ; CHECK: bb.1.vector.ph:
145 ; CHECK: successors: %bb.2(0x80000000)
146 ; CHECK: liveins: $r0, $r1, $r2, $r3, $r12
147 ; CHECK: renamable $lr = t2ADDri renamable $r12, 3, 14 /* CC::al */, $noreg, $noreg
148 ; CHECK: renamable $r4, dead $cpsr = tMOVi8 1, 14 /* CC::al */, $noreg
149 ; CHECK: renamable $lr = t2BICri killed renamable $lr, 3, 14 /* CC::al */, $noreg, $noreg
150 ; CHECK: renamable $q0 = MVE_VMOVimmi32 0, 0, $noreg, undef renamable $q0
151 ; CHECK: renamable $lr = t2SUBri killed renamable $lr, 4, 14 /* CC::al */, $noreg, $noreg
152 ; CHECK: renamable $r5 = nuw nsw t2ADDrs killed renamable $r4, killed renamable $lr, 19, 14 /* CC::al */, $noreg, $noreg
153 ; CHECK: dead $lr = t2DLS renamable $r5
154 ; CHECK: $r4 = tMOVr killed $r5, 14 /* CC::al */, $noreg
155 ; CHECK: bb.2.vector.body:
156 ; CHECK: successors: %bb.2(0x7c000000), %bb.3(0x04000000)
157 ; CHECK: liveins: $q0, $r0, $r1, $r2, $r3, $r4, $r12
158 ; CHECK: renamable $vpr = MVE_VCTP32 renamable $r12, 0, $noreg
159 ; CHECK: MVE_VPST 2, implicit $vpr
160 ; CHECK: renamable $r3, renamable $q1 = MVE_VLDRHS32_post killed renamable $r3, 8, 1, renamable $vpr :: (load (s64) from %ir.lsr.iv17.d, align 2)
161 ; CHECK: renamable $r2, renamable $q2 = MVE_VLDRHS32_post killed renamable $r2, 8, 1, renamable $vpr :: (load (s64) from %ir.lsr.iv1820.c, align 2)
162 ; CHECK: renamable $r1, renamable $q3 = MVE_VLDRHS32_post killed renamable $r1, 8, 1, renamable $vpr :: (load (s64) from %ir.lsr.iv1820, align 2)
163 ; CHECK: renamable $r0, renamable $q4 = MVE_VLDRHS32_post killed renamable $r0, 8, 1, renamable $vpr :: (load (s64) from %ir.lsr.iv17, align 2)
164 ; CHECK: renamable $q1 = nsw MVE_VMULi32 killed renamable $q2, killed renamable $q1, 0, $noreg, undef renamable $q1
165 ; CHECK: renamable $q2 = nsw MVE_VMULi32 killed renamable $q3, killed renamable $q4, 0, $noreg, undef renamable $q2
166 ; CHECK: $lr = tMOVr $r4, 14 /* CC::al */, $noreg
167 ; CHECK: renamable $q1 = MVE_VADDi32 killed renamable $q2, killed renamable $q1, 0, $noreg, undef renamable $q1
168 ; CHECK: renamable $r4, dead $cpsr = nsw tSUBi8 killed $r4, 1, 14 /* CC::al */, $noreg
169 ; CHECK: renamable $q1 = MVE_VADDi32 killed renamable $q1, renamable $q0, 0, $noreg, undef renamable $q1
170 ; CHECK: renamable $r12 = t2SUBri killed renamable $r12, 4, 14 /* CC::al */, $noreg, $noreg
171 ; CHECK: renamable $q0 = MVE_VPSEL killed renamable $q1, killed renamable $q0, 0, killed renamable $vpr
172 ; CHECK: dead $lr = t2LEUpdate killed renamable $lr, %bb.2
173 ; CHECK: bb.3.middle.block:
174 ; CHECK: liveins: $q0
175 ; CHECK: renamable $r0 = MVE_VADDVu32no_acc killed renamable $q0, 0, $noreg
176 ; CHECK: tPOP_RET 14 /* CC::al */, $noreg, def $r4, def $r5, def $r7, def $pc, implicit killed $r0
178 ; CHECK: renamable $r0, dead $cpsr = tMOVi8 0, 14 /* CC::al */, $noreg
179 ; CHECK: tPOP_RET 14 /* CC::al */, $noreg, def $r4, def $r5, def $r7, def $pc, implicit killed $r0
181 successors: %bb.4(0x30000000), %bb.1(0x50000000)
182 liveins: $r0, $r1, $r2, $r3, $r4, $r5, $r7, $lr
184 frame-setup tPUSH 14, $noreg, killed $r4, killed $r5, killed $r7, killed $lr, implicit-def $sp, implicit $sp
185 frame-setup CFI_INSTRUCTION def_cfa_offset 16
186 frame-setup CFI_INSTRUCTION offset $lr, -4
187 frame-setup CFI_INSTRUCTION offset $r7, -8
188 frame-setup CFI_INSTRUCTION offset $r5, -12
189 frame-setup CFI_INSTRUCTION offset $r4, -16
190 renamable $r12 = t2LDRi12 $sp, 16, 14, $noreg :: (load (s32) from %fixed-stack.0, align 8)
191 t2CMPri renamable $r12, 0, 14, $noreg, implicit-def $cpsr
192 tBcc %bb.4, 0, killed $cpsr
195 successors: %bb.2(0x80000000)
196 liveins: $r0, $r1, $r2, $r3, $r12
198 renamable $lr = t2ADDri renamable $r12, 3, 14, $noreg, $noreg
199 renamable $r4, dead $cpsr = tMOVi8 1, 14, $noreg
200 renamable $lr = t2BICri killed renamable $lr, 3, 14, $noreg, $noreg
201 renamable $q0 = MVE_VMOVimmi32 0, 0, $noreg, undef renamable $q0
202 renamable $lr = t2SUBri killed renamable $lr, 4, 14, $noreg, $noreg
203 renamable $r5 = nuw nsw t2ADDrs killed renamable $r4, killed renamable $lr, 19, 14, $noreg, $noreg
204 $lr = t2DoLoopStart renamable $r5
205 $r4 = tMOVr killed $r5, 14, $noreg
208 successors: %bb.2(0x7c000000), %bb.3(0x04000000)
209 liveins: $q0, $r0, $r1, $r2, $r3, $r4, $r12
211 renamable $vpr = MVE_VCTP32 renamable $r12, 0, $noreg
212 MVE_VPST 2, implicit $vpr
213 renamable $r3, renamable $q1 = MVE_VLDRHS32_post killed renamable $r3, 8, 1, renamable $vpr :: (load (s64) from %ir.lsr.iv17.d, align 2)
214 renamable $r2, renamable $q2 = MVE_VLDRHS32_post killed renamable $r2, 8, 1, renamable $vpr :: (load (s64) from %ir.lsr.iv1820.c, align 2)
215 renamable $r1, renamable $q3 = MVE_VLDRHS32_post killed renamable $r1, 8, 1, renamable $vpr :: (load (s64) from %ir.lsr.iv1820, align 2)
216 renamable $r0, renamable $q4 = MVE_VLDRHS32_post killed renamable $r0, 8, 1, renamable $vpr :: (load (s64) from %ir.lsr.iv17, align 2)
217 renamable $q1 = nsw MVE_VMULi32 killed renamable $q2, killed renamable $q1, 0, $noreg, undef renamable $q1
218 renamable $q2 = nsw MVE_VMULi32 killed renamable $q3, killed renamable $q4, 0, $noreg, undef renamable $q2
219 $lr = tMOVr $r4, 14, $noreg
220 renamable $q1 = MVE_VADDi32 killed renamable $q2, killed renamable $q1, 0, $noreg, undef renamable $q1
221 renamable $r4, dead $cpsr = nsw tSUBi8 killed $r4, 1, 14, $noreg
222 renamable $q1 = MVE_VADDi32 killed renamable $q1, renamable $q0, 0, $noreg, undef renamable $q1
223 renamable $r12 = t2SUBri killed renamable $r12, 4, 14, $noreg, $noreg
224 renamable $q0 = MVE_VPSEL killed renamable $q1, killed renamable $q0, 0, killed renamable $vpr
225 renamable $lr = t2LoopDec killed renamable $lr, 1
226 t2LoopEnd killed renamable $lr, %bb.2, implicit-def dead $cpsr
232 renamable $r0 = MVE_VADDVu32no_acc killed renamable $q0, 0, $noreg
233 tPOP_RET 14, $noreg, def $r4, def $r5, def $r7, def $pc, implicit killed $r0
236 renamable $r0, dead $cpsr = tMOVi8 0, 14, $noreg
237 tPOP_RET 14, $noreg, def $r4, def $r5, def $r7, def $pc, implicit killed $r0