150
|
1 ; RUN: opt < %s -S -debug-only=loop-unroll -loop-unroll 2>&1 | FileCheck %s
|
|
2 ; RUN: opt < %s -S -debug-only=loop-unroll -passes='require<profile-summary>,function(require<opt-remark-emit>,unroll)' 2>&1 | FileCheck %s
|
|
3 ; Confirm that peeling is disabled if the number of counts required to reach
|
|
4 ; the hot percentile is above the threshold.
|
|
5 ; RUN: opt < %s -S -profile-summary-huge-working-set-size-threshold=9 -debug-only=loop-unroll -passes='require<profile-summary>,function(require<opt-remark-emit>,unroll)' 2>&1 | FileCheck %s --check-prefix=NOPEEL
|
|
6 ; REQUIRES: asserts
|
|
7
|
|
8 ; Make sure we use the profile information correctly to peel-off 3 iterations
|
|
9 ; from the loop, and update the branch weights for the peeled loop properly.
|
|
10
|
|
11 ; CHECK: Loop Unroll: F[basic]
|
|
12 ; CHECK: PEELING loop %for.body with iteration count 4!
|
|
13 ; CHECK: Loop Unroll: F[optsize]
|
|
14 ; CHECK-NOT: PEELING
|
|
15
|
|
16 ; Confirm that no peeling occurs when we are performing full unrolling.
|
|
17 ; RUN: opt < %s -S -debug-only=loop-unroll -passes='require<opt-remark-emit>,loop(unroll-full)' 2>&1 | FileCheck %s --check-prefix=NOPEEL
|
|
18 ; NOPEEL-NOT: PEELING
|
|
19
|
|
20 ; CHECK-LABEL: @basic
|
|
21 ; CHECK: br i1 %{{.*}}, label %[[NEXT0:.*]], label %for.cond.for.end_crit_edge, !prof !15
|
|
22 ; CHECK: [[NEXT0]]:
|
|
23 ; CHECK: br i1 %{{.*}}, label %[[NEXT1:.*]], label %for.cond.for.end_crit_edge, !prof !16
|
|
24 ; CHECK: [[NEXT1]]:
|
|
25 ; CHECK: br i1 %{{.*}}, label %[[NEXT2:.*]], label %for.cond.for.end_crit_edge, !prof !17
|
|
26 ; CHECK: [[NEXT2]]:
|
|
27 ; CHECK: br i1 %{{.*}}, label %for.body, label %{{.*}}, !prof !18
|
|
28
|
|
29 define void @basic(i32* %p, i32 %k) #0 !prof !15 {
|
|
30 entry:
|
|
31 %cmp3 = icmp slt i32 0, %k
|
|
32 br i1 %cmp3, label %for.body.lr.ph, label %for.end
|
|
33
|
|
34 for.body.lr.ph: ; preds = %entry
|
|
35 br label %for.body
|
|
36
|
|
37 for.body: ; preds = %for.body.lr.ph, %for.body
|
|
38 %i.05 = phi i32 [ 0, %for.body.lr.ph ], [ %inc, %for.body ]
|
|
39 %p.addr.04 = phi i32* [ %p, %for.body.lr.ph ], [ %incdec.ptr, %for.body ]
|
|
40 %incdec.ptr = getelementptr inbounds i32, i32* %p.addr.04, i32 1
|
|
41 store i32 %i.05, i32* %p.addr.04, align 4
|
|
42 %inc = add nsw i32 %i.05, 1
|
|
43 %cmp = icmp slt i32 %inc, %k
|
|
44 br i1 %cmp, label %for.body, label %for.cond.for.end_crit_edge, !prof !16
|
|
45
|
|
46 for.cond.for.end_crit_edge: ; preds = %for.body
|
|
47 br label %for.end
|
|
48
|
|
49 for.end: ; preds = %for.cond.for.end_crit_edge, %entry
|
|
50 ret void
|
|
51 }
|
|
52
|
|
53 ; We don't want to peel loops when optimizing for size.
|
|
54 ; CHECK-LABEL: @optsize
|
|
55 ; CHECK: for.body.lr.ph:
|
|
56 ; CHECK-NEXT: br label %for.body
|
|
57 ; CHECK: for.body:
|
|
58 ; CHECK-NOT: br
|
|
59 ; CHECK: br i1 %cmp, label %for.body, label %for.cond.for.end_crit_edge
|
|
60 define void @optsize(i32* %p, i32 %k) #1 !prof !15 {
|
|
61 entry:
|
|
62 %cmp3 = icmp slt i32 0, %k
|
|
63 br i1 %cmp3, label %for.body.lr.ph, label %for.end
|
|
64
|
|
65 for.body.lr.ph: ; preds = %entry
|
|
66 br label %for.body
|
|
67
|
|
68 for.body: ; preds = %for.body.lr.ph, %for.body
|
|
69 %i.05 = phi i32 [ 0, %for.body.lr.ph ], [ %inc, %for.body ]
|
|
70 %p.addr.04 = phi i32* [ %p, %for.body.lr.ph ], [ %incdec.ptr, %for.body ]
|
|
71 %incdec.ptr = getelementptr inbounds i32, i32* %p.addr.04, i32 1
|
|
72 store i32 %i.05, i32* %p.addr.04, align 4
|
|
73 %inc = add nsw i32 %i.05, 1
|
|
74 %cmp = icmp slt i32 %inc, %k
|
|
75 br i1 %cmp, label %for.body, label %for.cond.for.end_crit_edge, !prof !16
|
|
76
|
|
77 for.cond.for.end_crit_edge: ; preds = %for.body
|
|
78 br label %for.end
|
|
79
|
|
80 for.end: ; preds = %for.cond.for.end_crit_edge, %entry
|
|
81 ret void
|
|
82 }
|
|
83
|
|
84 attributes #0 = { nounwind }
|
|
85 attributes #1 = { nounwind optsize }
|
|
86
|
|
87 !llvm.module.flags = !{!1}
|
|
88
|
|
89 !1 = !{i32 1, !"ProfileSummary", !2}
|
|
90 !2 = !{!3, !4, !5, !6, !7, !8, !9, !10}
|
|
91 !3 = !{!"ProfileFormat", !"InstrProf"}
|
|
92 !4 = !{!"TotalCount", i64 10}
|
|
93 !5 = !{!"MaxCount", i64 3}
|
|
94 !6 = !{!"MaxInternalCount", i64 1}
|
|
95 !7 = !{!"MaxFunctionCount", i64 3}
|
|
96 !8 = !{!"NumCounts", i64 2}
|
|
97 !9 = !{!"NumFunctions", i64 2}
|
|
98 !10 = !{!"DetailedSummary", !11}
|
|
99 !11 = !{!12, !13, !14}
|
|
100 !12 = !{i32 10000, i64 3, i32 2}
|
|
101 !13 = !{i32 999000, i64 1, i32 10}
|
|
102 !14 = !{i32 999999, i64 1, i32 10}
|
|
103 !15 = !{!"function_entry_count", i64 1}
|
|
104 !16 = !{!"branch_weights", i32 3001, i32 1001}
|
|
105
|
|
106 ;CHECK: !15 = !{!"branch_weights", i32 3001, i32 1001}
|
|
107 ;CHECK: !16 = !{!"branch_weights", i32 2000, i32 1001}
|
|
108 ;CHECK: !17 = !{!"branch_weights", i32 999, i32 1001}
|
|
109 ;CHECK: !18 = !{!"branch_weights", i32 1, i32 1001}
|
|
110
|