|
236
|
1 ; RUN: opt %loadPolly -polly-print-scops \
|
|
|
2 ; RUN: -polly-invariant-load-hoisting=true -disable-output < %s | FileCheck %s
|
|
150
|
3 ; RUN: opt %loadPolly -S -polly-codegen \
|
|
|
4 ; RUN: -polly-invariant-load-hoisting=true < %s | FileCheck %s --check-prefix=IR
|
|
|
5 ;
|
|
|
6 ; void f(long *A, long *B, long *ptr, long val) {
|
|
|
7 ; for (long i = 0; i < 100; i++) {
|
|
|
8 ; long ptrV = ((long)(ptr + 1)) + 1;
|
|
|
9 ; long valP = (long)(((long *)(val + 1)) + 1);
|
|
|
10 ; A[ptrV] += B[valP];
|
|
|
11 ; }
|
|
|
12 ; }
|
|
|
13 ;
|
|
|
14 ; CHECK: ReadAccess := [Reduction Type: NONE] [Scalar: 0]
|
|
|
15 ; CHECK-NEXT: [val, ptr] -> { Stmt_for_body[i0] -> MemRef_B[9 + val] };
|
|
221
|
16 ; CHECK-NEXT: Execution Context: [val, ptr] -> { : -4097 <= val <= 4086 }
|
|
150
|
17 ;
|
|
|
18 ; CHECK: ReadAccess := [Reduction Type: +] [Scalar: 0]
|
|
|
19 ; CHECK-NEXT: [val, ptr] -> { Stmt_for_body[i0] -> MemRef_A[9 + ptr] };
|
|
|
20 ; CHECK-NEXT: MustWriteAccess := [Reduction Type: +] [Scalar: 0]
|
|
|
21 ; CHECK-NEXT: [val, ptr] -> { Stmt_for_body[i0] -> MemRef_A[9 + ptr] };
|
|
221
|
22
|
|
|
23 ; IR: entry:
|
|
|
24 ; IR-NEXT: %ptr13 = ptrtoint i64* %ptr to i16
|
|
150
|
25 ;
|
|
|
26 ; IR: polly.stmt.for.body:
|
|
236
|
27 ; IR-NEXT: %tmp4_p_scalar_ = load i64, i64* %scevgep, align 8, !alias.scope !3, !noalias !0
|
|
221
|
28 ; IR-NEXT: %p_add4 = add nsw i64 %tmp4_p_scalar_, %polly.preload.tmp3.merge
|
|
236
|
29 ; IR-NEXT: store i64 %p_add4, i64* %scevgep, align 8, !alias.scope !3, !noalias !0
|
|
221
|
30 ; IR-NEXT: %polly.indvar_next = add nsw i64 %polly.indvar, 1
|
|
|
31 ; IR-NEXT: %polly.loop_cond = icmp sle i64 %polly.indvar_next, 99
|
|
|
32 ; IR-NEXT: br i1 %polly.loop_cond, label %polly.loop_header, label %polly.loop_exit
|
|
|
33
|
|
150
|
34 ; IR: polly.loop_preheader:
|
|
221
|
35 ; IR-NEXT: %41 = add i16 %val, 1
|
|
|
36 ; IR-NEXT: %42 = add i16 %ptr13, 9
|
|
|
37 ; IR-NEXT: %scevgep = getelementptr i64, i64* %A, i16 %42
|
|
150
|
38 ; IR-NEXT: br label %polly.loop_header
|
|
|
39 ;
|
|
|
40 target datalayout = "e-p:16:16:16-m:e-i64:64-f80:128-n8:16:16:64-S128"
|
|
|
41
|
|
|
42 define void @f(i64* %A, i64* %B, i64* %ptr, i16 %val) {
|
|
|
43 entry:
|
|
|
44 br label %for.cond
|
|
|
45
|
|
|
46 for.cond: ; preds = %for.inc, %entry
|
|
|
47 %i.0 = phi i64 [ 0, %entry ], [ %inc, %for.inc ]
|
|
|
48 %exitcond = icmp ne i64 %i.0, 100
|
|
|
49 br i1 %exitcond, label %for.body, label %for.end
|
|
|
50
|
|
|
51 for.body: ; preds = %for.cond
|
|
|
52 %add.ptr = getelementptr inbounds i64, i64* %ptr, i64 1
|
|
|
53 %tmp = ptrtoint i64* %add.ptr to i16
|
|
|
54 %add = add nsw i16 %tmp, 1
|
|
|
55 %add1 = add nsw i16 %val, 1
|
|
|
56 %tmp1 = inttoptr i16 %add1 to i64*
|
|
|
57 %add.ptr2 = getelementptr inbounds i64, i64* %tmp1, i64 1
|
|
|
58 %tmp2 = ptrtoint i64* %add.ptr2 to i16
|
|
|
59 %arrayidx = getelementptr inbounds i64, i64* %B, i16 %tmp2
|
|
|
60 %tmp3 = load i64, i64* %arrayidx
|
|
|
61 %arrayidx3 = getelementptr inbounds i64, i64* %A, i16 %add
|
|
|
62 %tmp4 = load i64, i64* %arrayidx3
|
|
|
63 %add4 = add nsw i64 %tmp4, %tmp3
|
|
|
64 store i64 %add4, i64* %arrayidx3
|
|
|
65 br label %for.inc
|
|
|
66
|
|
|
67 for.inc: ; preds = %for.body
|
|
|
68 %inc = add nuw nsw i64 %i.0, 1
|
|
|
69 br label %for.cond
|
|
|
70
|
|
|
71 for.end: ; preds = %for.cond
|
|
|
72 ret void
|
|
|
73 }
|
