| blob | a91f92348ab261bb13f25c429241d16e7b842c7e |
1 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 4
2 ; RUN: opt -passes=loop-vectorize -mtriple=riscv64 -mattr=+v -S %s | FileCheck %s
4 define void @test(ptr %p, i64 %a, i8 %b) {
5 ; CHECK-LABEL: define void @test(
6 ; CHECK-SAME: ptr [[P:%.*]], i64 [[A:%.*]], i8 [[B:%.*]]) #[[ATTR0:[0-9]+]] {
7 ; CHECK-NEXT: entry:
8 ; CHECK-NEXT: br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
9 ; CHECK: vector.ph:
10 ; CHECK-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <4 x i64> poison, i64 [[A]], i64 0
11 ; CHECK-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x i64> [[BROADCAST_SPLATINSERT]], <4 x i64> poison, <4 x i32> zeroinitializer
12 ; CHECK-NEXT: [[BROADCAST_SPLATINSERT1:%.*]] = insertelement <4 x i8> poison, i8 [[B]], i64 0
13 ; CHECK-NEXT: [[BROADCAST_SPLAT2:%.*]] = shufflevector <4 x i8> [[BROADCAST_SPLATINSERT1]], <4 x i8> poison, <4 x i32> zeroinitializer
14 ; CHECK-NEXT: br label [[VECTOR_BODY1:%.*]]
15 ; CHECK: vector.body:
16 ; CHECK-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[PRED_STORE_CONTINUE8:%.*]] ]
17 ; CHECK-NEXT: [[VEC_IND:%.*]] = phi <4 x i32> [ <i32 0, i32 1, i32 2, i32 3>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[PRED_STORE_CONTINUE8]] ]
18 ; CHECK-NEXT: [[TMP0:%.*]] = add i32 [[INDEX]], 0
19 ; CHECK-NEXT: [[ACTIVE_LANE_MASK:%.*]] = call <4 x i1> @llvm.get.active.lane.mask.v4i1.i32(i32 [[TMP0]], i32 3)
20 ; CHECK-NEXT: [[TMP1:%.*]] = icmp slt <4 x i32> [[VEC_IND]], <i32 2, i32 2, i32 2, i32 2>
21 ; CHECK-NEXT: [[TMP2:%.*]] = shl <4 x i64> [[BROADCAST_SPLAT]], <i64 48, i64 48, i64 48, i64 48>
22 ; CHECK-NEXT: [[TMP3:%.*]] = ashr <4 x i64> [[TMP2]], <i64 52, i64 52, i64 52, i64 52>
23 ; CHECK-NEXT: [[TMP4:%.*]] = trunc <4 x i64> [[TMP3]] to <4 x i32>
24 ; CHECK-NEXT: [[TMP5:%.*]] = select <4 x i1> [[ACTIVE_LANE_MASK]], <4 x i1> [[TMP1]], <4 x i1> zeroinitializer
25 ; CHECK-NEXT: [[TMP6:%.*]] = zext <4 x i8> [[BROADCAST_SPLAT2]] to <4 x i32>
26 ; CHECK-NEXT: [[PREDPHI:%.*]] = select <4 x i1> [[TMP5]], <4 x i32> [[TMP6]], <4 x i32> [[TMP4]]
27 ; CHECK-NEXT: [[TMP7:%.*]] = shl <4 x i32> [[PREDPHI]], <i32 8, i32 8, i32 8, i32 8>
28 ; CHECK-NEXT: [[TMP8:%.*]] = trunc <4 x i32> [[TMP7]] to <4 x i8>
29 ; CHECK-NEXT: [[TMP9:%.*]] = extractelement <4 x i1> [[ACTIVE_LANE_MASK]], i32 0
30 ; CHECK-NEXT: br i1 [[TMP9]], label [[PRED_STORE_IF:%.*]], label [[VECTOR_BODY:%.*]]
31 ; CHECK: pred.store.if:
32 ; CHECK-NEXT: [[TMP10:%.*]] = extractelement <4 x i8> [[TMP8]], i32 0
33 ; CHECK-NEXT: store i8 [[TMP10]], ptr [[P]], align 1
34 ; CHECK-NEXT: br label [[VECTOR_BODY]]
35 ; CHECK: pred.store.continue:
36 ; CHECK-NEXT: [[TMP11:%.*]] = extractelement <4 x i1> [[ACTIVE_LANE_MASK]], i32 1
37 ; CHECK-NEXT: br i1 [[TMP11]], label [[PRED_STORE_IF3:%.*]], label [[PRED_STORE_CONTINUE4:%.*]]
38 ; CHECK: pred.store.if3:
39 ; CHECK-NEXT: [[TMP12:%.*]] = extractelement <4 x i8> [[TMP8]], i32 1
40 ; CHECK-NEXT: store i8 [[TMP12]], ptr [[P]], align 1
41 ; CHECK-NEXT: br label [[PRED_STORE_CONTINUE4]]
42 ; CHECK: pred.store.continue4:
43 ; CHECK-NEXT: [[TMP13:%.*]] = extractelement <4 x i1> [[ACTIVE_LANE_MASK]], i32 2
44 ; CHECK-NEXT: br i1 [[TMP13]], label [[PRED_STORE_IF5:%.*]], label [[PRED_STORE_CONTINUE6:%.*]]
45 ; CHECK: pred.store.if5:
46 ; CHECK-NEXT: [[TMP14:%.*]] = extractelement <4 x i8> [[TMP8]], i32 2
47 ; CHECK-NEXT: store i8 [[TMP14]], ptr [[P]], align 1
48 ; CHECK-NEXT: br label [[PRED_STORE_CONTINUE6]]
49 ; CHECK: pred.store.continue6:
50 ; CHECK-NEXT: [[TMP15:%.*]] = extractelement <4 x i1> [[ACTIVE_LANE_MASK]], i32 3
51 ; CHECK-NEXT: br i1 [[TMP15]], label [[PRED_STORE_IF7:%.*]], label [[PRED_STORE_CONTINUE8]]
52 ; CHECK: pred.store.if7:
53 ; CHECK-NEXT: [[TMP16:%.*]] = extractelement <4 x i8> [[TMP8]], i32 3
54 ; CHECK-NEXT: store i8 [[TMP16]], ptr [[P]], align 1
55 ; CHECK-NEXT: br label [[PRED_STORE_CONTINUE8]]
56 ; CHECK: pred.store.continue8:
57 ; CHECK-NEXT: [[VEC_IND_NEXT]] = add <4 x i32> [[VEC_IND]], <i32 4, i32 4, i32 4, i32 4>
58 ; CHECK-NEXT: [[INDEX_NEXT]] = add i32 [[INDEX]], 4
59 ; CHECK-NEXT: br i1 true, label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY1]], !llvm.loop [[LOOP0:![0-9]+]]
60 ; CHECK: middle.block:
61 ; CHECK-NEXT: br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
62 ; CHECK: scalar.ph:
63 ; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i32 [ 4, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
64 ; CHECK-NEXT: br label [[FOR_COND:%.*]]
65 ; CHECK: for.cond:
66 ; CHECK-NEXT: [[IV:%.*]] = phi i32 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[ADD:%.*]], [[FOR_BODY:%.*]] ]
67 ; CHECK-NEXT: [[ADD]] = add i32 [[IV]], 1
68 ; CHECK-NEXT: [[CMP_SLT:%.*]] = icmp slt i32 [[IV]], 2
69 ; CHECK-NEXT: [[SHL:%.*]] = shl i64 [[A]], 48
70 ; CHECK-NEXT: [[ASHR:%.*]] = ashr i64 [[SHL]], 52
71 ; CHECK-NEXT: [[TRUNC_I32:%.*]] = trunc i64 [[ASHR]] to i32
72 ; CHECK-NEXT: br i1 [[CMP_SLT]], label [[COND_FALSE:%.*]], label [[FOR_BODY]]
73 ; CHECK: cond.false:
74 ; CHECK-NEXT: [[ZEXT:%.*]] = zext i8 [[B]] to i32
75 ; CHECK-NEXT: br label [[FOR_BODY]]
76 ; CHECK: for.body:
77 ; CHECK-NEXT: [[COND:%.*]] = phi i32 [ [[TRUNC_I32]], [[FOR_COND]] ], [ [[ZEXT]], [[COND_FALSE]] ]
78 ; CHECK-NEXT: [[SHL_I32:%.*]] = shl i32 [[COND]], 8
79 ; CHECK-NEXT: [[TRUNC:%.*]] = trunc i32 [[SHL_I32]] to i8
80 ; CHECK-NEXT: store i8 [[TRUNC]], ptr [[P]], align 1
81 ; CHECK-NEXT: [[CMP:%.*]] = icmp slt i32 [[IV]], 2
82 ; CHECK-NEXT: br i1 [[CMP]], label [[FOR_COND]], label [[EXIT]], !llvm.loop [[LOOP3:![0-9]+]]
83 ; CHECK: exit:
84 ; CHECK-NEXT: ret void
85 ;
86 entry:
87 br label %for.cond
89 for.cond: ; preds = %for.body, %entry
90 %iv = phi i32 [ 0, %entry ], [ %add, %for.body ]
91 %add = add i32 %iv, 1
92 %cmp.slt = icmp slt i32 %iv, 2
93 %shl = shl i64 %a, 48
94 %ashr = ashr i64 %shl, 52
95 %trunc.i32 = trunc i64 %ashr to i32
96 br i1 %cmp.slt, label %cond.false, label %for.body
98 cond.false: ; preds = %for.cond
99 %zext = zext i8 %b to i32
100 br label %for.body
102 for.body: ; preds = %cond.false, %for.cond
103 %cond = phi i32 [ %trunc.i32, %for.cond ], [ %zext, %cond.false ]
104 %shl.i32 = shl i32 %cond, 8
105 %trunc = trunc i32 %shl.i32 to i8
106 store i8 %trunc, ptr %p, align 1
107 %cmp = icmp slt i32 %iv, 2
108 br i1 %cmp, label %for.cond, label %exit
110 exit: ; preds = %for.body
111 ret void
112 }
113 ;.
114 ; CHECK: [[LOOP0]] = distinct !{[[LOOP0]], [[META1:![0-9]+]], [[META2:![0-9]+]]}
115 ; CHECK: [[META1]] = !{!"llvm.loop.isvectorized", i32 1}
116 ; CHECK: [[META2]] = !{!"llvm.loop.unroll.runtime.disable"}
117 ; CHECK: [[LOOP3]] = distinct !{[[LOOP3]], [[META2]], [[META1]]}
118 ;.