• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1; RUN: opt %loadPolly -analyze -polly-scops -polly-invariant-load-hoisting \
2; RUN:     -polly-detect-full-functions < %s | FileCheck %s
3;
4; This testcase checks for compatibility of the -detect-full-functions
5; flag in combination with the -invariant-load-hoisting option. More
6; specifically, ScopHelper.cpp::isHoistableLoad only gets called if
7; -invariant-load-hoisting is enabled. This function, however, had a bug
8; which caused a crash if the region argument was top-level. This test
9; is a minimal example that hits this specific code path.
10;
11; Also note that this file's IR is in no way optimized, i.e. it was
12; generated with clang -O0 from the following C-code:
13;
14;    void test() {
15;      int A[] = {1, 2, 3, 4, 5};
16;      int len = (sizeof A) / sizeof(int);
17;      for (int i = 0; i < len; ++i) {
18;        A[i] = A[i] * 2;
19;      }
20;    }
21;
22; This is also the reason why polly does not detect any scops (the loop
23; variable i is loaded from and stored to memory in each iteration):
24;
25; CHECK:      region: 'for.cond => for.end' in function 'test':
26; CHECK-NEXT: Invalid Scop!
27; CHECK-NEXT: region: 'entry => <Function Return>' in function 'test':
28; CHECK-NEXT: Invalid Scop!
29;
30target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
31
32@test.A = private unnamed_addr constant [5 x i32] [i32 1, i32 2, i32 3, i32 4, i32 5], align 16
33
34define void @test() {
35entry:
36  %A = alloca [5 x i32], align 16
37  %len = alloca i32, align 4
38  %i = alloca i32, align 4
39  %0 = bitcast [5 x i32]* %A to i8*
40  call void @llvm.memcpy.p0i8.p0i8.i64(i8* %0, i8* bitcast ([5 x i32]* @test.A to i8*), i64 20, i32 16, i1 false)
41  store i32 5, i32* %len, align 4
42  store i32 0, i32* %i, align 4
43  br label %for.cond
44
45for.cond:                                         ; preds = %for.inc, %entry
46  %1 = load i32, i32* %i, align 4
47  %2 = load i32, i32* %len, align 4
48  %cmp = icmp slt i32 %1, %2
49  br i1 %cmp, label %for.body, label %for.end
50
51for.body:                                         ; preds = %for.cond
52  %3 = load i32, i32* %i, align 4
53  %idxprom = sext i32 %3 to i64
54  %arrayidx = getelementptr inbounds [5 x i32], [5 x i32]* %A, i64 0, i64 %idxprom
55  %4 = load i32, i32* %arrayidx, align 4
56  %mul = mul nsw i32 %4, 2
57  %5 = load i32, i32* %i, align 4
58  %idxprom1 = sext i32 %5 to i64
59  %arrayidx2 = getelementptr inbounds [5 x i32], [5 x i32]* %A, i64 0, i64 %idxprom1
60  store i32 %mul, i32* %arrayidx2, align 4
61  br label %for.inc
62
63for.inc:                                          ; preds = %for.body
64  %6 = load i32, i32* %i, align 4
65  %inc = add nsw i32 %6, 1
66  store i32 %inc, i32* %i, align 4
67  br label %for.cond
68
69for.end:                                          ; preds = %for.cond
70  ret void
71}
72
73declare void @llvm.memcpy.p0i8.p0i8.i64(i8* nocapture writeonly, i8* nocapture readonly, i64, i32, i1)
74