1
; RUN: llc < %s -mtriple=x86_64-apple-macosx -disable-lsr -post-RA-scheduler=1 -break-anti-dependencies=critical | FileCheck %s
3
; In PR20020, the critical anti-dependency breaker algorithm mistakenly
4
; changes the register operands of an 'xorl %eax, %eax' to 'xorl %ecx, %ecx'
5
; and then immediately reloads %rcx with a value based on the wrong %rax
7
; CHECK-NOT: xorl %ecx, %ecx
8
; CHECK: leaq 1(%rax), %rcx
11
%struct.planet = type { double, double, double }
13
; Function Attrs: nounwind ssp uwtable
14
define void @advance(i32 %nbodies, %struct.planet* nocapture %bodies) #0 {
16
%cmp4 = icmp sgt i32 %nbodies, 0
17
br i1 %cmp4, label %for.body.preheader, label %for.end38
19
for.body.preheader: ; preds = %entry
20
%gep = getelementptr %struct.planet, %struct.planet* %bodies, i64 1, i32 1
21
%gep13 = bitcast double* %gep to %struct.planet*
22
%0 = add i32 %nbodies, -1
25
for.body: ; preds = %for.body.preheader, %for.inc20
26
%iv19 = phi i32 [ %0, %for.body.preheader ], [ %iv.next, %for.inc20 ]
27
%iv = phi %struct.planet* [ %gep13, %for.body.preheader ], [ %gep14, %for.inc20 ]
28
%iv9 = phi i64 [ %iv.next10, %for.inc20 ], [ 0, %for.body.preheader ]
29
%iv.next10 = add nuw nsw i64 %iv9, 1
30
%1 = trunc i64 %iv.next10 to i32
31
%cmp22 = icmp slt i32 %1, %nbodies
32
br i1 %cmp22, label %for.body3.lr.ph, label %for.inc20
34
for.body3.lr.ph: ; preds = %for.body
35
%x = getelementptr inbounds %struct.planet, %struct.planet* %bodies, i64 %iv9, i32 0
36
%y = getelementptr inbounds %struct.planet, %struct.planet* %bodies, i64 %iv9, i32 1
37
%vx = getelementptr inbounds %struct.planet, %struct.planet* %bodies, i64 %iv9, i32 2
40
for.body3: ; preds = %for.body3, %for.body3.lr.ph
41
%iv20 = phi i32 [ %iv.next21, %for.body3 ], [ %iv19, %for.body3.lr.ph ]
42
%iv15 = phi %struct.planet* [ %gep16, %for.body3 ], [ %iv, %for.body3.lr.ph ]
43
%iv1517 = bitcast %struct.planet* %iv15 to double*
44
%2 = load double, double* %x, align 8
45
%gep18 = getelementptr double, double* %iv1517, i64 -1
46
%3 = load double, double* %gep18, align 8
47
%sub = fsub double %2, %3
48
%4 = load double, double* %y, align 8
49
%5 = load double, double* %iv1517, align 8
50
%sub8 = fsub double %4, %5
51
%add10 = fadd double %sub, %sub8
52
%call = tail call double @sqrt(double %sub8) #2
53
store double %add10, double* %vx, align 8
54
%gep16 = getelementptr %struct.planet, %struct.planet* %iv15, i64 1
55
%iv.next21 = add i32 %iv20, -1
56
%exitcond = icmp eq i32 %iv.next21, 0
57
br i1 %exitcond, label %for.inc20, label %for.body3
59
for.inc20: ; preds = %for.body3, %for.body
60
%lftr.wideiv11 = trunc i64 %iv.next10 to i32
61
%gep14 = getelementptr %struct.planet, %struct.planet* %iv, i64 1
62
%iv.next = add i32 %iv19, -1
63
%exitcond12 = icmp eq i32 %lftr.wideiv11, %nbodies
64
br i1 %exitcond12, label %for.end38, label %for.body
66
for.end38: ; preds = %for.inc20, %entry
70
; Function Attrs: nounwind
71
declare double @sqrt(double) #1
73
attributes #0 = { "no-frame-pointer-elim-non-leaf" }