llvm-project

1
; RUN: opt %loadNPMPolly '-passes=print<polly-function-scops>' -disable-output < %s 2>&1 | FileCheck %s
2
; RUN: opt %loadNPMPolly -passes=polly-codegen -S < %s | FileCheck %s --check-prefix=IR
3
;
4
; Check that we guard the divisions because we moved them and thereby increased
5
; their domain.
6
;
7
; CHECK:         Invalid Context:
8
; CHECK-NEXT:    [p_0] -> {  : false }
9
; CHECK:         p0: (((zext i32 %a to i64) /u (zext i32 %b to i64)) /u ((zext i32 %c to i64) /u (zext i32 %d to i64)))
10
;
11
;    void f(unsigned *A, unsigned a, unsigned b, unsigned c, unsigned d) {
12
;      for (unsigned i; i < 100; i++)
13
;        A[i] += A[(a / b) / (c / d)];
14
;    }
15
;
16
; IR:       %[[A:[.a-zA-Z0-9]*]] = zext i32 %a to i64
17
; IR-NEXT:  %[[B:[.a-zA-Z0-9]*]] = zext i32 %b to i64
18
; IR-NEXT:  %[[R1:[.a-zA-Z0-9]*]] = call i64 @llvm.umax.i64(i64 %[[B]], i64 1)
19
; IR-NEXT:  %[[R2:[.a-zA-Z0-9]*]] = udiv i64 %[[A]], %[[R1]]
20
; IR-NEXT:  %[[C:[.a-zA-Z0-9]*]] = zext i32 %c to i64
21
; IR-NEXT:  %[[D:[.a-zA-Z0-9]*]] = zext i32 %d to i64
22
; IR-NEXT:  %[[R6:[.a-zA-Z0-9]*]] = call i64 @llvm.umax.i64(i64 %[[D]], i64 1)
23
; IR-NEXT:  %[[R7:[.a-zA-Z0-9]*]] = udiv i64 %[[C]], %[[R6]]
24
; IR-NEXT:  %[[R4:[.a-zA-Z0-9]*]] = call i64 @llvm.umax.i64(i64 %[[R7]], i64 1)
25
; IR-NEXT:  %[[R8:[.a-zA-Z0-9]*]] = udiv i64 %[[R2]], %[[R4]]
26
;
27
target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
28

29
define void @f(ptr %A, i32 %a, i32 %b, i32 %c, i32 %d) {
30
entry:
31
  br label %for.cond
32

33
for.cond:                                         ; preds = %for.inc, %entry
34
  %indvars.iv = phi i64 [ %indvars.iv.next, %for.inc ], [ 0, %entry ]
35
  %cmp = icmp ult i64 %indvars.iv, 100
36
  br i1 %cmp, label %for.body, label %for.end
37

38
for.body:                                         ; preds = %for.cond
39
  %div = udiv i32 %a, %b
40
  %div1 = udiv i32 %c, %d
41
  %div2 = udiv i32 %div, %div1
42
  %idxprom = zext i32 %div2 to i64
43
  %arrayidx = getelementptr inbounds i32, ptr %A, i64 %idxprom
44
  %tmp = load i32, ptr %arrayidx, align 4
45
  %arrayidx4 = getelementptr inbounds i32, ptr %A, i64 %indvars.iv
46
  %tmp1 = load i32, ptr %arrayidx4, align 4
47
  %add = add i32 %tmp1, %tmp
48
  store i32 %add, ptr %arrayidx4, align 4
49
  br label %for.inc
50

51
for.inc:                                          ; preds = %for.body
52
  %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1
53
  br label %for.cond
54

55
for.end:                                          ; preds = %for.cond
56
  ret void
57
}
58