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# TODO: We should be able to handle the VCMP -> VPST -> VCMP -> VCTP case.
5
6--- |
7  define dso_local arm_aapcs_vfpcc void @test(i32* noalias nocapture %a, i32* nocapture readonly %b, 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    %div = lshr i32 %N, 1
20    %trip.count.minus.1 = add i32 %N, -1
21    %broadcast.splatinsert = insertelement <4 x i32> undef, i32 %trip.count.minus.1, i32 0
22    %broadcast.splat = shufflevector <4 x i32> %broadcast.splatinsert, <4 x i32> undef, <4 x i32> zeroinitializer
23    %start = call i32 @llvm.start.loop.iterations.i32(i32 %tmp5)
24    br label %vector.body
25
26  vector.body:                                      ; preds = %vector.body, %vector.ph
27    %lsr.iv = phi i32 [ %lsr.iv.next, %vector.body ], [ %start, %vector.ph ]
28    %lsr.iv3 = phi i32* [ %scevgep4, %vector.body ], [ %b, %vector.ph ]
29    %lsr.iv1 = phi i32* [ %scevgep, %vector.body ], [ %a, %vector.ph ]
30    %vec.ind = phi <4 x i32> [ <i32 0, i32 1, i32 2, i32 3>, %vector.ph ], [ %vec.ind.next, %vector.body ]
31    %elts.rem = phi i32 [ %N, %vector.ph ], [ %elts.rem.next, %vector.body ]
32    %lsr.iv12 = bitcast i32* %lsr.iv1 to <4 x i32>*
33    %lsr.iv35 = bitcast i32* %lsr.iv3 to <4 x i32>*
34    %tmp7 = insertelement <4 x i32> undef, i32 %div, i32 0
35    %tmp8 = shufflevector <4 x i32> %tmp7, <4 x i32> undef, <4 x i32> zeroinitializer
36    %tmp9 = icmp ult <4 x i32> %vec.ind, %tmp8
37    %lower = icmp uge <4 x i32> %vec.ind, <i32 1, i32 1, i32 1, i32 1>
38    %tmp10 = call <4 x i1> @llvm.arm.mve.vctp32(i32 %elts.rem)
39    %tmp11 = and <4 x i1> %tmp9, %tmp10
40    %pred = and <4 x i1> %tmp11, %lower
41    %wide.masked.load = call <4 x i32> @llvm.masked.load.v4i32.p0v4i32(<4 x i32>* %lsr.iv35, i32 4, <4 x i1> %pred, <4 x i32> undef)
42    call void @llvm.masked.store.v4i32.p0v4i32(<4 x i32> %wide.masked.load, <4 x i32>* %lsr.iv12, i32 4, <4 x i1> %pred)
43    %vec.ind.next = add <4 x i32> %vec.ind, <i32 4, i32 4, i32 4, i32 4>
44    %elts.rem.next = sub i32 %elts.rem, 4
45    %scevgep = getelementptr i32, i32* %lsr.iv1, i32 4
46    %scevgep4 = getelementptr i32, i32* %lsr.iv3, i32 4
47    %tmp12 = call i32 @llvm.loop.decrement.reg.i32.i32.i32(i32 %lsr.iv, i32 1)
48    %tmp13 = icmp ne i32 %tmp12, 0
49    %lsr.iv.next = add nsw i32 %lsr.iv, -1
50    br i1 %tmp13, label %vector.body, label %for.cond.cleanup
51
52  for.cond.cleanup:                                 ; preds = %vector.body, %entry
53    ret void
54  }
55
56  declare <4 x i32> @llvm.masked.load.v4i32.p0v4i32(<4 x i32>*, i32 immarg, <4 x i1>, <4 x i32>)
57  declare void @llvm.masked.store.v4i32.p0v4i32(<4 x i32>, <4 x i32>*, i32 immarg, <4 x i1>)
58  declare <4 x i1> @llvm.arm.mve.vctp32(i32)
59  declare i32 @llvm.start.loop.iterations.i32(i32)
60  declare i32 @llvm.loop.decrement.reg.i32.i32.i32(i32, i32)
61
62...
63---
64name:            test
65alignment:       16
66tracksRegLiveness: true
67registers:       []
68liveins:
69  - { reg: '$r0', virtual-reg: '' }
70  - { reg: '$r1', virtual-reg: '' }
71  - { reg: '$r2', virtual-reg: '' }
72frameInfo:
73  stackSize:       24
74  offsetAdjustment: 0
75  maxAlignment:    8
76fixedStack:      []
77stack:
78  - { id: 0, name: '', type: spill-slot, offset: -4, size: 4, alignment: 4,
79      stack-id: default, callee-saved-register: '$lr', callee-saved-restored: false,
80      debug-info-variable: '', debug-info-expression: '', debug-info-location: '' }
81  - { id: 1, name: '', type: spill-slot, offset: -8, size: 4, alignment: 4,
82      stack-id: default, callee-saved-register: '$r4', callee-saved-restored: true,
83      debug-info-variable: '', debug-info-expression: '', debug-info-location: '' }
84  - { id: 2, name: '', type: spill-slot, offset: -16, size: 8, alignment: 8,
85      stack-id: default, callee-saved-register: '$d9', callee-saved-restored: true,
86      debug-info-variable: '', debug-info-expression: '', debug-info-location: '' }
87  - { id: 3, name: '', type: spill-slot, offset: -24, size: 8, alignment: 8,
88      stack-id: default, callee-saved-register: '$d8', callee-saved-restored: true,
89      debug-info-variable: '', debug-info-expression: '', debug-info-location: '' }
90callSites:       []
91constants:
92  - id:              0
93    value:           '<4 x i32> <i32 0, i32 1, i32 2, i32 3>'
94    alignment:       16
95    isTargetSpecific: false
96machineFunctionInfo: {}
97body:             |
98  ; CHECK-LABEL: name: test
99  ; CHECK: bb.0.entry:
100  ; CHECK:   successors: %bb.3(0x30000000), %bb.1(0x50000000)
101  ; CHECK:   liveins: $lr, $d8, $d9, $r0, $r1, $r2, $r4
102  ; CHECK:   frame-setup tPUSH 14 /* CC::al */, $noreg, killed $r4, killed $lr, implicit-def $sp, implicit $sp
103  ; CHECK:   frame-setup CFI_INSTRUCTION def_cfa_offset 8
104  ; CHECK:   frame-setup CFI_INSTRUCTION offset $lr, -4
105  ; CHECK:   frame-setup CFI_INSTRUCTION offset $r4, -8
106  ; CHECK:   $sp = frame-setup VSTMDDB_UPD $sp, 14 /* CC::al */, $noreg, killed $d8, killed $d9
107  ; CHECK:   frame-setup CFI_INSTRUCTION def_cfa_offset 24
108  ; CHECK:   frame-setup CFI_INSTRUCTION offset $d9, -16
109  ; CHECK:   frame-setup CFI_INSTRUCTION offset $d8, -24
110  ; CHECK:   tCBZ $r2, %bb.3
111  ; CHECK: bb.1.vector.ph:
112  ; CHECK:   successors: %bb.2(0x80000000)
113  ; CHECK:   liveins: $r0, $r1, $r2
114  ; CHECK:   renamable $r3, dead $cpsr = tADDi3 renamable $r2, 3, 14 /* CC::al */, $noreg
115  ; CHECK:   renamable $q2 = MVE_VMOVimmi32 1, 0, $noreg, undef renamable $q2
116  ; CHECK:   renamable $r3 = t2BICri killed renamable $r3, 3, 14 /* CC::al */, $noreg, $noreg
117  ; CHECK:   renamable $q3 = MVE_VMOVimmi32 4, 0, $noreg, undef renamable $q3
118  ; CHECK:   renamable $r12 = t2SUBri killed renamable $r3, 4, 14 /* CC::al */, $noreg, $noreg
119  ; CHECK:   renamable $r3, dead $cpsr = tMOVi8 1, 14 /* CC::al */, $noreg
120  ; CHECK:   renamable $r3 = nuw nsw t2ADDrs killed renamable $r3, killed renamable $r12, 19, 14 /* CC::al */, $noreg, $noreg
121  ; CHECK:   dead $lr = t2DLS renamable $r3
122  ; CHECK:   $r4 = tMOVr killed $r3, 14 /* CC::al */, $noreg
123  ; CHECK:   renamable $r3 = tLEApcrel %const.0, 14 /* CC::al */, $noreg
124  ; CHECK:   renamable $q0 = MVE_VLDRWU32 killed renamable $r3, 0, 0, $noreg :: (load 16 from constant-pool)
125  ; CHECK:   renamable $r3, dead $cpsr = tLSRri renamable $r2, 1, 14 /* CC::al */, $noreg
126  ; CHECK:   renamable $q1 = MVE_VDUP32 killed renamable $r3, 0, $noreg, undef renamable $q1
127  ; CHECK: bb.2.vector.body:
128  ; CHECK:   successors: %bb.2(0x7c000000), %bb.3(0x04000000)
129  ; CHECK:   liveins: $q0, $q1, $q2, $q3, $r0, $r1, $r2, $r4
130  ; CHECK:   $lr = tMOVr $r4, 14 /* CC::al */, $noreg
131  ; CHECK:   renamable $r4, dead $cpsr = nsw tSUBi8 killed $r4, 1, 14 /* CC::al */, $noreg
132  ; CHECK:   renamable $r2, dead $cpsr = tSUBi8 killed renamable $r2, 4, 14 /* CC::al */, $noreg
133  ; CHECK:   renamable $vpr = MVE_VCMPu32 renamable $q1, renamable $q0, 8, 0, $noreg
134  ; CHECK:   MVE_VPST 1, implicit $vpr
135  ; CHECK:   renamable $vpr = MVE_VCMPu32 renamable $q0, renamable $q2, 2, 1, killed renamable $vpr
136  ; CHECK:   renamable $vpr = MVE_VCTP32 renamable $r2, 1, killed renamable $vpr
137  ; CHECK:   renamable $r1, renamable $q4 = MVE_VLDRWU32_post killed renamable $r1, 16, 1, renamable $vpr :: (load 16 from %ir.lsr.iv35, align 4)
138  ; CHECK:   renamable $r0 = MVE_VSTRWU32_post killed renamable $q4, killed renamable $r0, 16, 1, killed renamable $vpr :: (store 16 into %ir.lsr.iv12, align 4)
139  ; CHECK:   renamable $q0 = MVE_VADDi32 killed renamable $q0, renamable $q3, 0, $noreg, undef renamable $q0
140  ; CHECK:   dead $lr = t2LEUpdate killed renamable $lr, %bb.2
141  ; CHECK: bb.3.for.cond.cleanup:
142  ; CHECK:   $sp = frame-destroy VLDMDIA_UPD $sp, 14 /* CC::al */, $noreg, def $d8, def $d9
143  ; CHECK:   frame-destroy tPOP_RET 14 /* CC::al */, $noreg, def $r4, def $pc
144  ; CHECK: bb.4 (align 16):
145  ; CHECK:   CONSTPOOL_ENTRY 0, %const.0, 16
146  bb.0.entry:
147    successors: %bb.3(0x30000000), %bb.1(0x50000000)
148    liveins: $r0, $r1, $r2, $r4, $lr, $d8, $d9
149
150    frame-setup tPUSH 14 /* CC::al */, $noreg, killed $r4, killed $lr, implicit-def $sp, implicit $sp
151    frame-setup CFI_INSTRUCTION def_cfa_offset 8
152    frame-setup CFI_INSTRUCTION offset $lr, -4
153    frame-setup CFI_INSTRUCTION offset $r4, -8
154    $sp = frame-setup VSTMDDB_UPD $sp, 14 /* CC::al */, $noreg, killed $d8, killed $d9
155    frame-setup CFI_INSTRUCTION def_cfa_offset 24
156    frame-setup CFI_INSTRUCTION offset $d9, -16
157    frame-setup CFI_INSTRUCTION offset $d8, -24
158    tCBZ $r2, %bb.3
159
160  bb.1.vector.ph:
161    successors: %bb.2(0x80000000)
162    liveins: $r0, $r1, $r2
163
164    renamable $r3, dead $cpsr = tADDi3 renamable $r2, 3, 14 /* CC::al */, $noreg
165    renamable $q2 = MVE_VMOVimmi32 1, 0, $noreg, undef renamable $q2
166    renamable $r3 = t2BICri killed renamable $r3, 3, 14 /* CC::al */, $noreg, $noreg
167    renamable $q3 = MVE_VMOVimmi32 4, 0, $noreg, undef renamable $q3
168    renamable $r12 = t2SUBri killed renamable $r3, 4, 14 /* CC::al */, $noreg, $noreg
169    renamable $r3, dead $cpsr = tMOVi8 1, 14 /* CC::al */, $noreg
170    renamable $r3 = nuw nsw t2ADDrs killed renamable $r3, killed renamable $r12, 19, 14 /* CC::al */, $noreg, $noreg
171    $lr = t2DoLoopStart renamable $r3
172    $r4 = tMOVr killed $r3, 14 /* CC::al */, $noreg
173    renamable $r3 = tLEApcrel %const.0, 14 /* CC::al */, $noreg
174    renamable $q0 = MVE_VLDRWU32 killed renamable $r3, 0, 0, $noreg :: (load 16 from constant-pool)
175    renamable $r3, dead $cpsr = tLSRri renamable $r2, 1, 14 /* CC::al */, $noreg
176    renamable $q1 = MVE_VDUP32 killed renamable $r3, 0, $noreg, undef renamable $q1
177
178  bb.2.vector.body:
179    successors: %bb.2(0x7c000000), %bb.3(0x04000000)
180    liveins: $q0, $q1, $q2, $q3, $r0, $r1, $r2, $r4
181
182    $lr = tMOVr $r4, 14 /* CC::al */, $noreg
183    renamable $r4, dead $cpsr = nsw tSUBi8 killed $r4, 1, 14 /* CC::al */, $noreg
184    renamable $r2, dead $cpsr = tSUBi8 killed renamable $r2, 4, 14 /* CC::al */, $noreg
185    renamable $vpr = MVE_VCMPu32 renamable $q1, renamable $q0, 8, 0, $noreg
186    MVE_VPST 1, implicit $vpr
187    renamable $vpr = MVE_VCMPu32 renamable $q0, renamable $q2, 2, 1, killed renamable $vpr
188    renamable $vpr = MVE_VCTP32 renamable $r2, 1, killed renamable $vpr
189    renamable $r1, renamable $q4 = MVE_VLDRWU32_post killed renamable $r1, 16, 1, renamable $vpr :: (load 16 from %ir.lsr.iv35, align 4)
190    renamable $r0 = MVE_VSTRWU32_post killed renamable $q4, killed renamable $r0, 16, 1, killed renamable $vpr :: (store 16 into %ir.lsr.iv12, align 4)
191    renamable $q0 = MVE_VADDi32 killed renamable $q0, renamable $q3, 0, $noreg, undef renamable $q0
192    renamable $lr = t2LoopDec killed renamable $lr, 1
193    t2LoopEnd killed renamable $lr, %bb.2, implicit-def dead $cpsr
194    tB %bb.3, 14 /* CC::al */, $noreg
195
196  bb.3.for.cond.cleanup:
197    $sp = frame-destroy VLDMDIA_UPD $sp, 14 /* CC::al */, $noreg, def $d8, def $d9
198    frame-destroy tPOP_RET 14 /* CC::al */, $noreg, def $r4, def $pc
199
200  bb.4 (align 16):
201    CONSTPOOL_ENTRY 0, %const.0, 16
202
203...
204