xref: /llvm-project/llvm/test/Analysis/ScalarEvolution/nsw-offset.ll (revision 8b5b294ec2cf876bc5eb5bd5fcb56ef487e36d60)
1; NOTE: Assertions have been autogenerated by utils/update_analyze_test_checks.py
2; RUN: opt < %s -S -disable-output "-passes=print<scalar-evolution>" 2>&1 | FileCheck %s
3
4; ScalarEvolution should be able to fold away the sign-extensions
5; on this loop with a primary induction variable incremented with
6; a nsw add of 2.
7
8target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128"
9
10define void @foo(i32 %no, ptr nocapture %d, ptr nocapture %q) nounwind {
11; CHECK-LABEL: 'foo'
12; CHECK-NEXT:  Classifying expressions for: @foo
13; CHECK-NEXT:    %n = and i32 %no, -2
14; CHECK-NEXT:    --> (2 * (%no /u 2))<nuw> U: [0,-1) S: [-2147483648,2147483647)
15; CHECK-NEXT:    %i.01 = phi i32 [ %16, %bb1 ], [ 0, %bb.nph ]
16; CHECK-NEXT:    --> {0,+,2}<nuw><nsw><%bb> U: [0,2147483645) S: [0,2147483645) Exits: (2 * ((-1 + (2 * (%no /u 2))<nuw>) /u 2))<nuw> LoopDispositions: { %bb: Computable }
17; CHECK-NEXT:    %1 = sext i32 %i.01 to i64
18; CHECK-NEXT:    --> {0,+,2}<nuw><nsw><%bb> U: [0,2147483645) S: [0,2147483645) Exits: (2 * ((1 + (zext i32 (-2 + (2 * (%no /u 2))<nuw>) to i64))<nuw><nsw> /u 2))<nuw><nsw> LoopDispositions: { %bb: Computable }
19; CHECK-NEXT:    %2 = getelementptr inbounds double, ptr %d, i64 %1
20; CHECK-NEXT:    --> {%d,+,16}<nuw><%bb> U: full-set S: full-set Exits: ((16 * ((1 + (zext i32 (-2 + (2 * (%no /u 2))<nuw>) to i64))<nuw><nsw> /u 2))<nuw><nsw> + %d) LoopDispositions: { %bb: Computable }
21; CHECK-NEXT:    %4 = sext i32 %i.01 to i64
22; CHECK-NEXT:    --> {0,+,2}<nuw><nsw><%bb> U: [0,2147483645) S: [0,2147483645) Exits: (2 * ((1 + (zext i32 (-2 + (2 * (%no /u 2))<nuw>) to i64))<nuw><nsw> /u 2))<nuw><nsw> LoopDispositions: { %bb: Computable }
23; CHECK-NEXT:    %5 = getelementptr inbounds double, ptr %q, i64 %4
24; CHECK-NEXT:    --> {%q,+,16}<nuw><%bb> U: full-set S: full-set Exits: ((16 * ((1 + (zext i32 (-2 + (2 * (%no /u 2))<nuw>) to i64))<nuw><nsw> /u 2))<nuw><nsw> + %q) LoopDispositions: { %bb: Computable }
25; CHECK-NEXT:    %7 = or disjoint i32 %i.01, 1
26; CHECK-NEXT:    --> {1,+,2}<nuw><nsw><%bb> U: [1,2147483646) S: [1,2147483646) Exits: (1 + (2 * ((-1 + (2 * (%no /u 2))<nuw>) /u 2))<nuw>)<nuw><nsw> LoopDispositions: { %bb: Computable }
27; CHECK-NEXT:    %8 = sext i32 %7 to i64
28; CHECK-NEXT:    --> {1,+,2}<nuw><nsw><%bb> U: [1,2147483646) S: [1,2147483646) Exits: (1 + (2 * ((1 + (zext i32 (-2 + (2 * (%no /u 2))<nuw>) to i64))<nuw><nsw> /u 2))<nuw><nsw>)<nuw><nsw> LoopDispositions: { %bb: Computable }
29; CHECK-NEXT:    %9 = getelementptr inbounds double, ptr %q, i64 %8
30; CHECK-NEXT:    --> {(8 + %q),+,16}<nuw><%bb> U: full-set S: full-set Exits: (8 + (16 * ((1 + (zext i32 (-2 + (2 * (%no /u 2))<nuw>) to i64))<nuw><nsw> /u 2))<nuw><nsw> + %q) LoopDispositions: { %bb: Computable }
31; CHECK-NEXT:    %t7 = add nsw i32 %i.01, 1
32; CHECK-NEXT:    --> {1,+,2}<nuw><nsw><%bb> U: [1,2147483646) S: [1,2147483646) Exits: (1 + (2 * ((-1 + (2 * (%no /u 2))<nuw>) /u 2))<nuw>)<nuw><nsw> LoopDispositions: { %bb: Computable }
33; CHECK-NEXT:    %t8 = sext i32 %t7 to i64
34; CHECK-NEXT:    --> {1,+,2}<nuw><nsw><%bb> U: [1,2147483646) S: [1,2147483646) Exits: (1 + (2 * ((1 + (zext i32 (-2 + (2 * (%no /u 2))<nuw>) to i64))<nuw><nsw> /u 2))<nuw><nsw>)<nuw><nsw> LoopDispositions: { %bb: Computable }
35; CHECK-NEXT:    %t9 = getelementptr inbounds double, ptr %q, i64 %t8
36; CHECK-NEXT:    --> {(8 + %q),+,16}<nuw><%bb> U: full-set S: full-set Exits: (8 + (16 * ((1 + (zext i32 (-2 + (2 * (%no /u 2))<nuw>) to i64))<nuw><nsw> /u 2))<nuw><nsw> + %q) LoopDispositions: { %bb: Computable }
37; CHECK-NEXT:    %14 = sext i32 %i.01 to i64
38; CHECK-NEXT:    --> {0,+,2}<nuw><nsw><%bb> U: [0,2147483645) S: [0,2147483645) Exits: (2 * ((1 + (zext i32 (-2 + (2 * (%no /u 2))<nuw>) to i64))<nuw><nsw> /u 2))<nuw><nsw> LoopDispositions: { %bb: Computable }
39; CHECK-NEXT:    %15 = getelementptr inbounds double, ptr %d, i64 %14
40; CHECK-NEXT:    --> {%d,+,16}<nuw><%bb> U: full-set S: full-set Exits: ((16 * ((1 + (zext i32 (-2 + (2 * (%no /u 2))<nuw>) to i64))<nuw><nsw> /u 2))<nuw><nsw> + %d) LoopDispositions: { %bb: Computable }
41; CHECK-NEXT:    %16 = add nsw i32 %i.01, 2
42; CHECK-NEXT:    --> {2,+,2}<nuw><nsw><%bb> U: [2,2147483647) S: [2,2147483647) Exits: (2 + (2 * ((-1 + (2 * (%no /u 2))<nuw>) /u 2))<nuw>) LoopDispositions: { %bb: Computable }
43; CHECK-NEXT:  Determining loop execution counts for: @foo
44; CHECK-NEXT:  Loop %bb: backedge-taken count is ((-1 + (2 * (%no /u 2))<nuw>) /u 2)
45; CHECK-NEXT:  Loop %bb: constant max backedge-taken count is i32 1073741822
46; CHECK-NEXT:  Loop %bb: symbolic max backedge-taken count is ((-1 + (2 * (%no /u 2))<nuw>) /u 2)
47; CHECK-NEXT:  Loop %bb: Trip multiple is 1
48;
49entry:
50  %n = and i32 %no, 4294967294
51  %0 = icmp sgt i32 %n, 0                         ; <i1> [#uses=1]
52  br i1 %0, label %bb.nph, label %return
53
54bb.nph:                                           ; preds = %entry
55  br label %bb
56
57bb:                                               ; preds = %bb.nph, %bb1
58  %i.01 = phi i32 [ %16, %bb1 ], [ 0, %bb.nph ]   ; <i32> [#uses=5]
59
60  %1 = sext i32 %i.01 to i64                      ; <i64> [#uses=1]
61
62  %2 = getelementptr inbounds double, ptr %d, i64 %1  ; <ptr> [#uses=1]
63
64  %3 = load double, ptr %2, align 8                   ; <double> [#uses=1]
65  %4 = sext i32 %i.01 to i64                      ; <i64> [#uses=1]
66  %5 = getelementptr inbounds double, ptr %q, i64 %4  ; <ptr> [#uses=1]
67  %6 = load double, ptr %5, align 8                   ; <double> [#uses=1]
68  %7 = or disjoint i32 %i.01, 1                            ; <i32> [#uses=1]
69
70  %8 = sext i32 %7 to i64                         ; <i64> [#uses=1]
71
72  %9 = getelementptr inbounds double, ptr %q, i64 %8  ; <ptr> [#uses=1]
73
74; Artificially repeat the above three instructions, this time using
75; add nsw instead of or.
76  %t7 = add nsw i32 %i.01, 1                            ; <i32> [#uses=1]
77
78  %t8 = sext i32 %t7 to i64                         ; <i64> [#uses=1]
79
80  %t9 = getelementptr inbounds double, ptr %q, i64 %t8  ; <ptr> [#uses=1]
81
82  %10 = load double, ptr %9, align 8                  ; <double> [#uses=1]
83  %11 = fadd double %6, %10                       ; <double> [#uses=1]
84  %12 = fadd double %11, 3.200000e+00             ; <double> [#uses=1]
85  %13 = fmul double %3, %12                       ; <double> [#uses=1]
86  %14 = sext i32 %i.01 to i64                     ; <i64> [#uses=1]
87  %15 = getelementptr inbounds double, ptr %d, i64 %14 ; <ptr> [#uses=1]
88  store double %13, ptr %15, align 8
89  %16 = add nsw i32 %i.01, 2                      ; <i32> [#uses=2]
90  br label %bb1
91
92bb1:                                              ; preds = %bb
93  %17 = icmp slt i32 %16, %n                      ; <i1> [#uses=1]
94  br i1 %17, label %bb, label %bb1.return_crit_edge
95
96bb1.return_crit_edge:                             ; preds = %bb1
97  br label %return
98
99return:                                           ; preds = %bb1.return_crit_edge, %entry
100  ret void
101}
102