xref: /llvm-project/llvm/test/Transforms/LoopVectorize/AArch64/tail-folding-styles.ll (revision 56c091ea7106507b36015297ee9005c9d5fab0bf)
1; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
2; RUN: opt -S -passes=loop-vectorize -force-tail-folding-style=none < %s | FileCheck %s --check-prefix=NONE
3; RUN: opt -S -passes=loop-vectorize -force-tail-folding-style=data < %s | FileCheck %s --check-prefix=DATA
4; RUN: opt -S -passes=loop-vectorize -force-tail-folding-style=data-without-lane-mask < %s | FileCheck %s --check-prefix=DATA_NO_LANEMASK
5; RUN: opt -S -passes=loop-vectorize -force-tail-folding-style=data-and-control < %s | FileCheck %s --check-prefix=DATA_AND_CONTROL
6; RUN: opt -S -passes=loop-vectorize -force-tail-folding-style=data-and-control-without-rt-check < %s | FileCheck %s --check-prefix=DATA_AND_CONTROL_NO_RT_CHECK
7
8target triple = "aarch64-unknown-linux-gnu"
9
10; Test the different tail folding styles.
11
12define void @simple_memset_tailfold(i32 %val, ptr %ptr, i64 %n) "target-features" = "+sve" {
13; NONE-LABEL: @simple_memset_tailfold(
14; NONE-NEXT:  entry:
15; NONE-NEXT:    [[UMAX:%.*]] = call i64 @llvm.umax.i64(i64 [[N:%.*]], i64 1)
16; NONE-NEXT:    [[TMP0:%.*]] = call i64 @llvm.vscale.i64()
17; NONE-NEXT:    [[TMP1:%.*]] = mul i64 [[TMP0]], 4
18; NONE-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[UMAX]], [[TMP1]]
19; NONE-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
20; NONE:       vector.ph:
21; NONE-NEXT:    [[TMP2:%.*]] = call i64 @llvm.vscale.i64()
22; NONE-NEXT:    [[TMP3:%.*]] = mul i64 [[TMP2]], 4
23; NONE-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[UMAX]], [[TMP3]]
24; NONE-NEXT:    [[N_VEC:%.*]] = sub i64 [[UMAX]], [[N_MOD_VF]]
25; NONE-NEXT:    [[TMP7:%.*]] = call i64 @llvm.vscale.i64()
26; NONE-NEXT:    [[TMP8:%.*]] = mul i64 [[TMP7]], 4
27; NONE-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <vscale x 4 x i32> poison, i32 [[VAL:%.*]], i64 0
28; NONE-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <vscale x 4 x i32> [[BROADCAST_SPLATINSERT]], <vscale x 4 x i32> poison, <vscale x 4 x i32> zeroinitializer
29; NONE-NEXT:    br label [[VECTOR_BODY:%.*]]
30; NONE:       vector.body:
31; NONE-NEXT:    [[INDEX1:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT2:%.*]], [[VECTOR_BODY]] ]
32; NONE-NEXT:    [[TMP4:%.*]] = add i64 [[INDEX1]], 0
33; NONE-NEXT:    [[TMP5:%.*]] = getelementptr i32, ptr [[PTR:%.*]], i64 [[TMP4]]
34; NONE-NEXT:    [[TMP6:%.*]] = getelementptr i32, ptr [[TMP5]], i32 0
35; NONE-NEXT:    store <vscale x 4 x i32> [[BROADCAST_SPLAT]], ptr [[TMP6]], align 4
36; NONE-NEXT:    [[INDEX_NEXT2]] = add nuw i64 [[INDEX1]], [[TMP8]]
37; NONE-NEXT:    [[TMP9:%.*]] = icmp eq i64 [[INDEX_NEXT2]], [[N_VEC]]
38; NONE-NEXT:    br i1 [[TMP9]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
39; NONE:       middle.block:
40; NONE-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[UMAX]], [[N_VEC]]
41; NONE-NEXT:    br i1 [[CMP_N]], label [[WHILE_END_LOOPEXIT:%.*]], label [[SCALAR_PH]]
42; NONE:       scalar.ph:
43; NONE-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
44; NONE-NEXT:    br label [[WHILE_BODY:%.*]]
45; NONE:       while.body:
46; NONE-NEXT:    [[INDEX:%.*]] = phi i64 [ [[INDEX_NEXT:%.*]], [[WHILE_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
47; NONE-NEXT:    [[GEP:%.*]] = getelementptr i32, ptr [[PTR]], i64 [[INDEX]]
48; NONE-NEXT:    store i32 [[VAL]], ptr [[GEP]], align 4
49; NONE-NEXT:    [[INDEX_NEXT]] = add nsw i64 [[INDEX]], 1
50; NONE-NEXT:    [[CMP10:%.*]] = icmp ult i64 [[INDEX_NEXT]], [[N]]
51; NONE-NEXT:    br i1 [[CMP10]], label [[WHILE_BODY]], label [[WHILE_END_LOOPEXIT]], !llvm.loop [[LOOP3:![0-9]+]]
52; NONE:       while.end.loopexit:
53; NONE-NEXT:    ret void
54;
55; DATA-LABEL: @simple_memset_tailfold(
56; DATA-NEXT:  entry:
57; DATA-NEXT:    [[UMAX:%.*]] = call i64 @llvm.umax.i64(i64 [[N:%.*]], i64 1)
58; DATA-NEXT:    [[TMP0:%.*]] = sub i64 -1, [[UMAX]]
59; DATA-NEXT:    [[TMP1:%.*]] = call i64 @llvm.vscale.i64()
60; DATA-NEXT:    [[TMP2:%.*]] = mul i64 [[TMP1]], 4
61; DATA-NEXT:    [[TMP3:%.*]] = icmp ult i64 [[TMP0]], [[TMP2]]
62; DATA-NEXT:    br i1 [[TMP3]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
63; DATA:       vector.ph:
64; DATA-NEXT:    [[TMP4:%.*]] = call i64 @llvm.vscale.i64()
65; DATA-NEXT:    [[TMP5:%.*]] = mul i64 [[TMP4]], 4
66; DATA-NEXT:    [[TMP8:%.*]] = sub i64 [[TMP5]], 1
67; DATA-NEXT:    [[N_RND_UP:%.*]] = add i64 [[UMAX]], [[TMP8]]
68; DATA-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N_RND_UP]], [[TMP5]]
69; DATA-NEXT:    [[N_VEC:%.*]] = sub i64 [[N_RND_UP]], [[N_MOD_VF]]
70; DATA-NEXT:    [[TMP12:%.*]] = call i64 @llvm.vscale.i64()
71; DATA-NEXT:    [[TMP13:%.*]] = mul i64 [[TMP12]], 4
72; DATA-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <vscale x 4 x i32> poison, i32 [[VAL:%.*]], i64 0
73; DATA-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <vscale x 4 x i32> [[BROADCAST_SPLATINSERT]], <vscale x 4 x i32> poison, <vscale x 4 x i32> zeroinitializer
74; DATA-NEXT:    br label [[VECTOR_BODY:%.*]]
75; DATA:       vector.body:
76; DATA-NEXT:    [[INDEX1:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT2:%.*]], [[VECTOR_BODY]] ]
77; DATA-NEXT:    [[TMP9:%.*]] = add i64 [[INDEX1]], 0
78; DATA-NEXT:    [[ACTIVE_LANE_MASK:%.*]] = call <vscale x 4 x i1> @llvm.get.active.lane.mask.nxv4i1.i64(i64 [[TMP9]], i64 [[UMAX]])
79; DATA-NEXT:    [[TMP10:%.*]] = getelementptr i32, ptr [[PTR:%.*]], i64 [[TMP9]]
80; DATA-NEXT:    [[TMP11:%.*]] = getelementptr i32, ptr [[TMP10]], i32 0
81; DATA-NEXT:    call void @llvm.masked.store.nxv4i32.p0(<vscale x 4 x i32> [[BROADCAST_SPLAT]], ptr [[TMP11]], i32 4, <vscale x 4 x i1> [[ACTIVE_LANE_MASK]])
82; DATA-NEXT:    [[INDEX_NEXT2]] = add i64 [[INDEX1]], [[TMP13]]
83; DATA-NEXT:    [[TMP14:%.*]] = icmp eq i64 [[INDEX_NEXT2]], [[N_VEC]]
84; DATA-NEXT:    br i1 [[TMP14]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
85; DATA:       middle.block:
86; DATA-NEXT:    br i1 true, label [[WHILE_END_LOOPEXIT:%.*]], label [[SCALAR_PH]]
87; DATA:       scalar.ph:
88; DATA-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
89; DATA-NEXT:    br label [[WHILE_BODY:%.*]]
90; DATA:       while.body:
91; DATA-NEXT:    [[INDEX:%.*]] = phi i64 [ [[INDEX_NEXT:%.*]], [[WHILE_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
92; DATA-NEXT:    [[GEP:%.*]] = getelementptr i32, ptr [[PTR]], i64 [[INDEX]]
93; DATA-NEXT:    store i32 [[VAL]], ptr [[GEP]], align 4
94; DATA-NEXT:    [[INDEX_NEXT]] = add nsw i64 [[INDEX]], 1
95; DATA-NEXT:    [[CMP10:%.*]] = icmp ult i64 [[INDEX_NEXT]], [[N]]
96; DATA-NEXT:    br i1 [[CMP10]], label [[WHILE_BODY]], label [[WHILE_END_LOOPEXIT]], !llvm.loop [[LOOP3:![0-9]+]]
97; DATA:       while.end.loopexit:
98; DATA-NEXT:    ret void
99;
100; DATA_NO_LANEMASK-LABEL: @simple_memset_tailfold(
101; DATA_NO_LANEMASK-NEXT:  entry:
102; DATA_NO_LANEMASK-NEXT:    [[UMAX:%.*]] = call i64 @llvm.umax.i64(i64 [[N:%.*]], i64 1)
103; DATA_NO_LANEMASK-NEXT:    [[TMP0:%.*]] = sub i64 -1, [[UMAX]]
104; DATA_NO_LANEMASK-NEXT:    [[TMP1:%.*]] = call i64 @llvm.vscale.i64()
105; DATA_NO_LANEMASK-NEXT:    [[TMP2:%.*]] = mul i64 [[TMP1]], 4
106; DATA_NO_LANEMASK-NEXT:    [[TMP3:%.*]] = icmp ult i64 [[TMP0]], [[TMP2]]
107; DATA_NO_LANEMASK-NEXT:    br i1 [[TMP3]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
108; DATA_NO_LANEMASK:       vector.ph:
109; DATA_NO_LANEMASK-NEXT:    [[TMP4:%.*]] = call i64 @llvm.vscale.i64()
110; DATA_NO_LANEMASK-NEXT:    [[TMP5:%.*]] = mul i64 [[TMP4]], 4
111; DATA_NO_LANEMASK-NEXT:    [[TMP8:%.*]] = sub i64 [[TMP5]], 1
112; DATA_NO_LANEMASK-NEXT:    [[N_RND_UP:%.*]] = add i64 [[UMAX]], [[TMP8]]
113; DATA_NO_LANEMASK-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N_RND_UP]], [[TMP5]]
114; DATA_NO_LANEMASK-NEXT:    [[N_VEC:%.*]] = sub i64 [[N_RND_UP]], [[N_MOD_VF]]
115; DATA_NO_LANEMASK-NEXT:    [[TRIP_COUNT_MINUS_1:%.*]] = sub i64 [[UMAX]], 1
116; DATA_NO_LANEMASK-NEXT:    [[TMP15:%.*]] = call i64 @llvm.vscale.i64()
117; DATA_NO_LANEMASK-NEXT:    [[TMP16:%.*]] = mul i64 [[TMP15]], 4
118; DATA_NO_LANEMASK-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <vscale x 4 x i64> poison, i64 [[TRIP_COUNT_MINUS_1]], i64 0
119; DATA_NO_LANEMASK-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <vscale x 4 x i64> [[BROADCAST_SPLATINSERT]], <vscale x 4 x i64> poison, <vscale x 4 x i32> zeroinitializer
120; DATA_NO_LANEMASK-NEXT:    [[BROADCAST_SPLATINSERT4:%.*]] = insertelement <vscale x 4 x i32> poison, i32 [[VAL:%.*]], i64 0
121; DATA_NO_LANEMASK-NEXT:    [[BROADCAST_SPLAT5:%.*]] = shufflevector <vscale x 4 x i32> [[BROADCAST_SPLATINSERT4]], <vscale x 4 x i32> poison, <vscale x 4 x i32> zeroinitializer
122; DATA_NO_LANEMASK-NEXT:    br label [[VECTOR_BODY:%.*]]
123; DATA_NO_LANEMASK:       vector.body:
124; DATA_NO_LANEMASK-NEXT:    [[INDEX1:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT6:%.*]], [[VECTOR_BODY]] ]
125; DATA_NO_LANEMASK-NEXT:    [[TMP9:%.*]] = add i64 [[INDEX1]], 0
126; DATA_NO_LANEMASK-NEXT:    [[BROADCAST_SPLATINSERT2:%.*]] = insertelement <vscale x 4 x i64> poison, i64 [[INDEX1]], i64 0
127; DATA_NO_LANEMASK-NEXT:    [[BROADCAST_SPLAT3:%.*]] = shufflevector <vscale x 4 x i64> [[BROADCAST_SPLATINSERT2]], <vscale x 4 x i64> poison, <vscale x 4 x i32> zeroinitializer
128; DATA_NO_LANEMASK-NEXT:    [[TMP10:%.*]] = call <vscale x 4 x i64> @llvm.stepvector.nxv4i64()
129; DATA_NO_LANEMASK-NEXT:    [[TMP11:%.*]] = add <vscale x 4 x i64> zeroinitializer, [[TMP10]]
130; DATA_NO_LANEMASK-NEXT:    [[VEC_IV:%.*]] = add <vscale x 4 x i64> [[BROADCAST_SPLAT3]], [[TMP11]]
131; DATA_NO_LANEMASK-NEXT:    [[TMP12:%.*]] = icmp ule <vscale x 4 x i64> [[VEC_IV]], [[BROADCAST_SPLAT]]
132; DATA_NO_LANEMASK-NEXT:    [[TMP13:%.*]] = getelementptr i32, ptr [[PTR:%.*]], i64 [[TMP9]]
133; DATA_NO_LANEMASK-NEXT:    [[TMP14:%.*]] = getelementptr i32, ptr [[TMP13]], i32 0
134; DATA_NO_LANEMASK-NEXT:    call void @llvm.masked.store.nxv4i32.p0(<vscale x 4 x i32> [[BROADCAST_SPLAT5]], ptr [[TMP14]], i32 4, <vscale x 4 x i1> [[TMP12]])
135; DATA_NO_LANEMASK-NEXT:    [[INDEX_NEXT6]] = add i64 [[INDEX1]], [[TMP16]]
136; DATA_NO_LANEMASK-NEXT:    [[TMP17:%.*]] = icmp eq i64 [[INDEX_NEXT6]], [[N_VEC]]
137; DATA_NO_LANEMASK-NEXT:    br i1 [[TMP17]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
138; DATA_NO_LANEMASK:       middle.block:
139; DATA_NO_LANEMASK-NEXT:    br i1 true, label [[WHILE_END_LOOPEXIT:%.*]], label [[SCALAR_PH]]
140; DATA_NO_LANEMASK:       scalar.ph:
141; DATA_NO_LANEMASK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
142; DATA_NO_LANEMASK-NEXT:    br label [[WHILE_BODY:%.*]]
143; DATA_NO_LANEMASK:       while.body:
144; DATA_NO_LANEMASK-NEXT:    [[INDEX:%.*]] = phi i64 [ [[INDEX_NEXT:%.*]], [[WHILE_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
145; DATA_NO_LANEMASK-NEXT:    [[GEP:%.*]] = getelementptr i32, ptr [[PTR]], i64 [[INDEX]]
146; DATA_NO_LANEMASK-NEXT:    store i32 [[VAL]], ptr [[GEP]], align 4
147; DATA_NO_LANEMASK-NEXT:    [[INDEX_NEXT]] = add nsw i64 [[INDEX]], 1
148; DATA_NO_LANEMASK-NEXT:    [[CMP10:%.*]] = icmp ult i64 [[INDEX_NEXT]], [[N]]
149; DATA_NO_LANEMASK-NEXT:    br i1 [[CMP10]], label [[WHILE_BODY]], label [[WHILE_END_LOOPEXIT]], !llvm.loop [[LOOP3:![0-9]+]]
150; DATA_NO_LANEMASK:       while.end.loopexit:
151; DATA_NO_LANEMASK-NEXT:    ret void
152;
153; DATA_AND_CONTROL-LABEL: @simple_memset_tailfold(
154; DATA_AND_CONTROL-NEXT:  entry:
155; DATA_AND_CONTROL-NEXT:    [[UMAX:%.*]] = call i64 @llvm.umax.i64(i64 [[N:%.*]], i64 1)
156; DATA_AND_CONTROL-NEXT:    [[TMP0:%.*]] = sub i64 -1, [[UMAX]]
157; DATA_AND_CONTROL-NEXT:    [[TMP1:%.*]] = call i64 @llvm.vscale.i64()
158; DATA_AND_CONTROL-NEXT:    [[TMP2:%.*]] = mul i64 [[TMP1]], 4
159; DATA_AND_CONTROL-NEXT:    [[TMP3:%.*]] = icmp ult i64 [[TMP0]], [[TMP2]]
160; DATA_AND_CONTROL-NEXT:    br i1 [[TMP3]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
161; DATA_AND_CONTROL:       vector.ph:
162; DATA_AND_CONTROL-NEXT:    [[TMP4:%.*]] = call i64 @llvm.vscale.i64()
163; DATA_AND_CONTROL-NEXT:    [[TMP5:%.*]] = mul i64 [[TMP4]], 4
164; DATA_AND_CONTROL-NEXT:    [[TMP8:%.*]] = sub i64 [[TMP5]], 1
165; DATA_AND_CONTROL-NEXT:    [[N_RND_UP:%.*]] = add i64 [[UMAX]], [[TMP8]]
166; DATA_AND_CONTROL-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N_RND_UP]], [[TMP5]]
167; DATA_AND_CONTROL-NEXT:    [[N_VEC:%.*]] = sub i64 [[N_RND_UP]], [[N_MOD_VF]]
168; DATA_AND_CONTROL-NEXT:    [[TMP12:%.*]] = call i64 @llvm.vscale.i64()
169; DATA_AND_CONTROL-NEXT:    [[TMP13:%.*]] = mul i64 [[TMP12]], 4
170; DATA_AND_CONTROL-NEXT:    [[ACTIVE_LANE_MASK_ENTRY:%.*]] = call <vscale x 4 x i1> @llvm.get.active.lane.mask.nxv4i1.i64(i64 0, i64 [[UMAX]])
171; DATA_AND_CONTROL-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <vscale x 4 x i32> poison, i32 [[VAL:%.*]], i64 0
172; DATA_AND_CONTROL-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <vscale x 4 x i32> [[BROADCAST_SPLATINSERT]], <vscale x 4 x i32> poison, <vscale x 4 x i32> zeroinitializer
173; DATA_AND_CONTROL-NEXT:    br label [[VECTOR_BODY:%.*]]
174; DATA_AND_CONTROL:       vector.body:
175; DATA_AND_CONTROL-NEXT:    [[INDEX1:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT2:%.*]], [[VECTOR_BODY]] ]
176; DATA_AND_CONTROL-NEXT:    [[ACTIVE_LANE_MASK:%.*]] = phi <vscale x 4 x i1> [ [[ACTIVE_LANE_MASK_ENTRY]], [[VECTOR_PH]] ], [ [[ACTIVE_LANE_MASK_NEXT:%.*]], [[VECTOR_BODY]] ]
177; DATA_AND_CONTROL-NEXT:    [[TMP9:%.*]] = add i64 [[INDEX1]], 0
178; DATA_AND_CONTROL-NEXT:    [[TMP10:%.*]] = getelementptr i32, ptr [[PTR:%.*]], i64 [[TMP9]]
179; DATA_AND_CONTROL-NEXT:    [[TMP11:%.*]] = getelementptr i32, ptr [[TMP10]], i32 0
180; DATA_AND_CONTROL-NEXT:    call void @llvm.masked.store.nxv4i32.p0(<vscale x 4 x i32> [[BROADCAST_SPLAT]], ptr [[TMP11]], i32 4, <vscale x 4 x i1> [[ACTIVE_LANE_MASK]])
181; DATA_AND_CONTROL-NEXT:    [[INDEX_NEXT2]] = add i64 [[INDEX1]], [[TMP13]]
182; DATA_AND_CONTROL-NEXT:    [[ACTIVE_LANE_MASK_NEXT]] = call <vscale x 4 x i1> @llvm.get.active.lane.mask.nxv4i1.i64(i64 [[INDEX_NEXT2]], i64 [[UMAX]])
183; DATA_AND_CONTROL-NEXT:    [[TMP14:%.*]] = xor <vscale x 4 x i1> [[ACTIVE_LANE_MASK_NEXT]], splat (i1 true)
184; DATA_AND_CONTROL-NEXT:    [[TMP15:%.*]] = extractelement <vscale x 4 x i1> [[TMP14]], i32 0
185; DATA_AND_CONTROL-NEXT:    br i1 [[TMP15]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
186; DATA_AND_CONTROL:       middle.block:
187; DATA_AND_CONTROL-NEXT:    br i1 true, label [[WHILE_END_LOOPEXIT:%.*]], label [[SCALAR_PH]]
188; DATA_AND_CONTROL:       scalar.ph:
189; DATA_AND_CONTROL-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
190; DATA_AND_CONTROL-NEXT:    br label [[WHILE_BODY:%.*]]
191; DATA_AND_CONTROL:       while.body:
192; DATA_AND_CONTROL-NEXT:    [[INDEX:%.*]] = phi i64 [ [[INDEX_NEXT:%.*]], [[WHILE_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
193; DATA_AND_CONTROL-NEXT:    [[GEP:%.*]] = getelementptr i32, ptr [[PTR]], i64 [[INDEX]]
194; DATA_AND_CONTROL-NEXT:    store i32 [[VAL]], ptr [[GEP]], align 4
195; DATA_AND_CONTROL-NEXT:    [[INDEX_NEXT]] = add nsw i64 [[INDEX]], 1
196; DATA_AND_CONTROL-NEXT:    [[CMP10:%.*]] = icmp ult i64 [[INDEX_NEXT]], [[N]]
197; DATA_AND_CONTROL-NEXT:    br i1 [[CMP10]], label [[WHILE_BODY]], label [[WHILE_END_LOOPEXIT]], !llvm.loop [[LOOP3:![0-9]+]]
198; DATA_AND_CONTROL:       while.end.loopexit:
199; DATA_AND_CONTROL-NEXT:    ret void
200;
201; DATA_AND_CONTROL_NO_RT_CHECK-LABEL: @simple_memset_tailfold(
202; DATA_AND_CONTROL_NO_RT_CHECK-NEXT:  entry:
203; DATA_AND_CONTROL_NO_RT_CHECK-NEXT:    [[UMAX:%.*]] = call i64 @llvm.umax.i64(i64 [[N:%.*]], i64 1)
204; DATA_AND_CONTROL_NO_RT_CHECK-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
205; DATA_AND_CONTROL_NO_RT_CHECK:       vector.ph:
206; DATA_AND_CONTROL_NO_RT_CHECK-NEXT:    [[TMP0:%.*]] = call i64 @llvm.vscale.i64()
207; DATA_AND_CONTROL_NO_RT_CHECK-NEXT:    [[TMP1:%.*]] = mul i64 [[TMP0]], 4
208; DATA_AND_CONTROL_NO_RT_CHECK-NEXT:    [[TMP4:%.*]] = sub i64 [[TMP1]], 1
209; DATA_AND_CONTROL_NO_RT_CHECK-NEXT:    [[N_RND_UP:%.*]] = add i64 [[UMAX]], [[TMP4]]
210; DATA_AND_CONTROL_NO_RT_CHECK-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N_RND_UP]], [[TMP1]]
211; DATA_AND_CONTROL_NO_RT_CHECK-NEXT:    [[N_VEC:%.*]] = sub i64 [[N_RND_UP]], [[N_MOD_VF]]
212; DATA_AND_CONTROL_NO_RT_CHECK-NEXT:    [[TMP13:%.*]] = call i64 @llvm.vscale.i64()
213; DATA_AND_CONTROL_NO_RT_CHECK-NEXT:    [[TMP14:%.*]] = mul i64 [[TMP13]], 4
214; DATA_AND_CONTROL_NO_RT_CHECK-NEXT:    [[TMP5:%.*]] = call i64 @llvm.vscale.i64()
215; DATA_AND_CONTROL_NO_RT_CHECK-NEXT:    [[TMP6:%.*]] = mul i64 [[TMP5]], 4
216; DATA_AND_CONTROL_NO_RT_CHECK-NEXT:    [[TMP7:%.*]] = sub i64 [[UMAX]], [[TMP6]]
217; DATA_AND_CONTROL_NO_RT_CHECK-NEXT:    [[TMP8:%.*]] = icmp ugt i64 [[UMAX]], [[TMP6]]
218; DATA_AND_CONTROL_NO_RT_CHECK-NEXT:    [[TMP9:%.*]] = select i1 [[TMP8]], i64 [[TMP7]], i64 0
219; DATA_AND_CONTROL_NO_RT_CHECK-NEXT:    [[ACTIVE_LANE_MASK_ENTRY:%.*]] = call <vscale x 4 x i1> @llvm.get.active.lane.mask.nxv4i1.i64(i64 0, i64 [[UMAX]])
220; DATA_AND_CONTROL_NO_RT_CHECK-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <vscale x 4 x i32> poison, i32 [[VAL:%.*]], i64 0
221; DATA_AND_CONTROL_NO_RT_CHECK-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <vscale x 4 x i32> [[BROADCAST_SPLATINSERT]], <vscale x 4 x i32> poison, <vscale x 4 x i32> zeroinitializer
222; DATA_AND_CONTROL_NO_RT_CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
223; DATA_AND_CONTROL_NO_RT_CHECK:       vector.body:
224; DATA_AND_CONTROL_NO_RT_CHECK-NEXT:    [[INDEX1:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT2:%.*]], [[VECTOR_BODY]] ]
225; DATA_AND_CONTROL_NO_RT_CHECK-NEXT:    [[ACTIVE_LANE_MASK:%.*]] = phi <vscale x 4 x i1> [ [[ACTIVE_LANE_MASK_ENTRY]], [[VECTOR_PH]] ], [ [[ACTIVE_LANE_MASK_NEXT:%.*]], [[VECTOR_BODY]] ]
226; DATA_AND_CONTROL_NO_RT_CHECK-NEXT:    [[TMP10:%.*]] = add i64 [[INDEX1]], 0
227; DATA_AND_CONTROL_NO_RT_CHECK-NEXT:    [[TMP11:%.*]] = getelementptr i32, ptr [[PTR:%.*]], i64 [[TMP10]]
228; DATA_AND_CONTROL_NO_RT_CHECK-NEXT:    [[TMP12:%.*]] = getelementptr i32, ptr [[TMP11]], i32 0
229; DATA_AND_CONTROL_NO_RT_CHECK-NEXT:    call void @llvm.masked.store.nxv4i32.p0(<vscale x 4 x i32> [[BROADCAST_SPLAT]], ptr [[TMP12]], i32 4, <vscale x 4 x i1> [[ACTIVE_LANE_MASK]])
230; DATA_AND_CONTROL_NO_RT_CHECK-NEXT:    [[INDEX_NEXT2]] = add i64 [[INDEX1]], [[TMP14]]
231; DATA_AND_CONTROL_NO_RT_CHECK-NEXT:    [[ACTIVE_LANE_MASK_NEXT]] = call <vscale x 4 x i1> @llvm.get.active.lane.mask.nxv4i1.i64(i64 [[INDEX1]], i64 [[TMP9]])
232; DATA_AND_CONTROL_NO_RT_CHECK-NEXT:    [[TMP15:%.*]] = xor <vscale x 4 x i1> [[ACTIVE_LANE_MASK_NEXT]], splat (i1 true)
233; DATA_AND_CONTROL_NO_RT_CHECK-NEXT:    [[TMP16:%.*]] = extractelement <vscale x 4 x i1> [[TMP15]], i32 0
234; DATA_AND_CONTROL_NO_RT_CHECK-NEXT:    br i1 [[TMP16]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
235; DATA_AND_CONTROL_NO_RT_CHECK:       middle.block:
236; DATA_AND_CONTROL_NO_RT_CHECK-NEXT:    br i1 true, label [[WHILE_END_LOOPEXIT:%.*]], label [[SCALAR_PH]]
237; DATA_AND_CONTROL_NO_RT_CHECK:       scalar.ph:
238; DATA_AND_CONTROL_NO_RT_CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
239; DATA_AND_CONTROL_NO_RT_CHECK-NEXT:    br label [[WHILE_BODY:%.*]]
240; DATA_AND_CONTROL_NO_RT_CHECK:       while.body:
241; DATA_AND_CONTROL_NO_RT_CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ [[INDEX_NEXT:%.*]], [[WHILE_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
242; DATA_AND_CONTROL_NO_RT_CHECK-NEXT:    [[GEP:%.*]] = getelementptr i32, ptr [[PTR]], i64 [[INDEX]]
243; DATA_AND_CONTROL_NO_RT_CHECK-NEXT:    store i32 [[VAL]], ptr [[GEP]], align 4
244; DATA_AND_CONTROL_NO_RT_CHECK-NEXT:    [[INDEX_NEXT]] = add nsw i64 [[INDEX]], 1
245; DATA_AND_CONTROL_NO_RT_CHECK-NEXT:    [[CMP10:%.*]] = icmp ult i64 [[INDEX_NEXT]], [[N]]
246; DATA_AND_CONTROL_NO_RT_CHECK-NEXT:    br i1 [[CMP10]], label [[WHILE_BODY]], label [[WHILE_END_LOOPEXIT]], !llvm.loop [[LOOP3:![0-9]+]]
247; DATA_AND_CONTROL_NO_RT_CHECK:       while.end.loopexit:
248; DATA_AND_CONTROL_NO_RT_CHECK-NEXT:    ret void
249;
250entry:
251  br label %while.body
252
253while.body:                                       ; preds = %while.body, %entry
254  %index = phi i64 [ %index.next, %while.body ], [ 0, %entry ]
255  %gep = getelementptr i32, ptr %ptr, i64 %index
256  store i32 %val, ptr %gep
257  %index.next = add nsw i64 %index, 1
258  %cmp10 = icmp ult i64 %index.next, %n
259  br i1 %cmp10, label %while.body, label %while.end.loopexit, !llvm.loop !0
260
261while.end.loopexit:                               ; preds = %while.body
262  ret void
263}
264
265!0 = distinct !{!0, !1, !2, !3, !4}
266!1 = !{!"llvm.loop.vectorize.predicate.enable", i1 true}
267!2 = !{!"llvm.loop.vectorize.scalable.enable", i1 true}
268!3 = !{!"llvm.loop.interleave.count", i32 1}
269!4 = !{!"llvm.loop.vectorize.width", i32 4}
270