xref: /llvm-project/polly/test/ScopInfo/memmove.ll (revision e1f056f692d869708c1898d9d65a69ac5584a0ed)
1; RUN: opt %loadNPMPolly -aa-pipeline=basic-aa -polly-allow-differing-element-types '-passes=print<polly-function-scops>' -disable-output < %s 2>&1 | FileCheck %s
2; RUN: opt %loadNPMPolly -S -aa-pipeline=basic-aa -polly-allow-differing-element-types -passes=polly-codegen < %s 2>&1 | FileCheck --check-prefix=IR %s
3;
4; CHECK:         Arrays {
5; CHECK-NEXT:        i8 MemRef_A[*]; // Element size 1
6; CHECK-NEXT:        i8 MemRef_B[*]; // Element size 1
7; CHECK-NEXT:    }
8; CHECK:         Statements {
9; CHECK-NEXT:       Stmt_for_body3
10; CHECK-NEXT:            Domain :=
11; CHECK-NEXT:                { Stmt_for_body3[i0, i1] : 0 <= i0 <= 1023 and 0 <= i1 <= 1023 };
12; CHECK-NEXT:            Schedule :=
13; CHECK-NEXT:                { Stmt_for_body3[i0, i1] -> [i0, i1] };
14; CHECK-NEXT:            MustWriteAccess :=	[Reduction Type: NONE] [Scalar: 0]
15; CHECK-NEXT:                { Stmt_for_body3[i0, i1] -> MemRef_A[o0] : -16 <= o0 <= 15 };
16; CHECK-NEXT:            ReadAccess :=	[Reduction Type: NONE] [Scalar: 0]
17; CHECK-NEXT:                { Stmt_for_body3[i0, i1] -> MemRef_B[o0] : 64 <= o0 <= 95 };
18;
19; IR: polly.loop_preheader:
20; IR:   %[[r1:[a-zA-Z0-9]*]] = getelementptr i8, ptr %A, i64 -16
21; IR:   %[[r3:[a-zA-Z0-9]*]] = getelementptr i8, ptr %B, i64 64
22;
23; IR: polly.stmt.for.body3:
24; IR:   call void @llvm.memmove.p0.p0.i64(ptr align 4 %[[r1]], ptr align 4 %[[r3]], i64 32, i1 false)
25;
26;    #include <string.h>
27;
28;    void jd(int *restrict A, long *restrict B) {
29;      for (int i = 0; i < 1024; i++)
30;        for (int j = 0; j < 1024; j++)
31;          memmove(A - 4, B + 8, 32);
32;    }
33;
34target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
35
36define void @jd(ptr noalias %A, ptr noalias %B) {
37entry:
38  br label %for.cond
39
40for.cond:                                         ; preds = %for.inc5, %entry
41  %i.0 = phi i32 [ 0, %entry ], [ %inc6, %for.inc5 ]
42  %exitcond1 = icmp ne i32 %i.0, 1024
43  br i1 %exitcond1, label %for.body, label %for.end7
44
45for.body:                                         ; preds = %for.cond
46  br label %for.cond1
47
48for.cond1:                                        ; preds = %for.inc, %for.body
49  %j.0 = phi i32 [ 0, %for.body ], [ %inc, %for.inc ]
50  %exitcond = icmp ne i32 %j.0, 1024
51  br i1 %exitcond, label %for.body3, label %for.end
52
53for.body3:                                        ; preds = %for.cond1
54  %add.ptr = getelementptr inbounds i32, ptr %A, i64 -4
55  %add.ptr4 = getelementptr inbounds i64, ptr %B, i64 8
56  call void @llvm.memmove.p0.p0.i64(ptr %add.ptr, ptr %add.ptr4, i64 32, i32 4, i1 false)
57  br label %for.inc
58
59for.inc:                                          ; preds = %for.body3
60  %inc = add nsw i32 %j.0, 1
61  br label %for.cond1
62
63for.end:                                          ; preds = %for.cond1
64  br label %for.inc5
65
66for.inc5:                                         ; preds = %for.end
67  %inc6 = add nsw i32 %i.0, 1
68  br label %for.cond
69
70for.end7:                                         ; preds = %for.cond
71  ret void
72}
73
74declare void @llvm.memmove.p0.p0.i64(ptr nocapture, ptr nocapture readonly, i64, i32, i1) #1
75
76