<html>
<head>
<base href="https://bugs.llvm.org/">
</head>
<body><table border="1" cellspacing="0" cellpadding="8">
<tr>
<th>Bug ID</th>
<td><a class="bz_bug_link
bz_status_NEW "
title="NEW - -indvars regression after r346397"
href="https://bugs.llvm.org/show_bug.cgi?id=39673">39673</a>
</td>
</tr>
<tr>
<th>Summary</th>
<td>-indvars regression after r346397
</td>
</tr>
<tr>
<th>Product</th>
<td>libraries
</td>
</tr>
<tr>
<th>Version</th>
<td>trunk
</td>
</tr>
<tr>
<th>Hardware</th>
<td>PC
</td>
</tr>
<tr>
<th>OS</th>
<td>Linux
</td>
</tr>
<tr>
<th>Status</th>
<td>NEW
</td>
</tr>
<tr>
<th>Severity</th>
<td>enhancement
</td>
</tr>
<tr>
<th>Priority</th>
<td>P
</td>
</tr>
<tr>
<th>Component</th>
<td>Scalar Optimizations
</td>
</tr>
<tr>
<th>Assignee</th>
<td>unassignedbugs@nondot.org
</td>
</tr>
<tr>
<th>Reporter</th>
<td>mikael.holmen@ericsson.com
</td>
</tr>
<tr>
<th>CC</th>
<td>llvm-bugs@lists.llvm.org
</td>
</tr></table>
<p>
<div>
<pre>Created <span class=""><a href="attachment.cgi?id=21123" name="attach_21123" title="reproducer">attachment 21123</a> <a href="attachment.cgi?id=21123&action=edit" title="reproducer">[details]</a></span>
reproducer
The regression starts happening with r346397:
Return "[IndVars] Smart hard uses detection"
The patch has been reverted because it ended up prohibiting propagation
of a constant to exit value. For such values, we should skip all checks
related to hard uses because propagating a constant is always profitable.
Differential Revision: <a href="https://reviews.llvm.org/D53691">https://reviews.llvm.org/D53691</a>
git-svn-id: <a href="https://llvm.org/svn/llvm-project/llvm/trunk@346397">https://llvm.org/svn/llvm-project/llvm/trunk@346397</a>
91177308-0d34-0410-b5e6-96231b3b80d8
If I run
opt -S -o - foo2.ll -indvars
with the patch I get
define i16 @test10() {
entry:
br label %loop1
loop1: ; preds = %loop1, %entry
%l1 = phi i16 [ 0, %entry ], [ %l1.add, %loop1 ]
%l1.add = add nuw nsw i16 %l1, 1
%cmp1 = icmp ult i16 %l1.add, 2
br i1 %cmp1, label %loop1, label %loop2.preheader
loop2.preheader: ; preds = %loop1
br label %loop2
loop2: ; preds = %loop2,
%loop2.preheader
%k2 = phi i16 [ %k2.add, %loop2 ], [ 182, %loop2.preheader ]
%l2 = phi i16 [ %l2.add, %loop2 ], [ 0, %loop2.preheader ]
%l2.add = add nuw nsw i16 %l2, 1
tail call void @foo(i16 %k2)
%k2.add = add nuw nsw i16 %k2, 1
%cmp2 = icmp ult i16 %l2.add, 2
br i1 %cmp2, label %loop2, label %loop2.end
loop2.end: ; preds = %loop2
%k2.add.lcssa = phi i16 [ %k2.add, %loop2 ]
ret i16 %k2.add.lcssa
}
and without it:
define i16 @test10() {
entry:
br label %loop1
loop1: ; preds = %loop1, %entry
%l1 = phi i16 [ 0, %entry ], [ %l1.add, %loop1 ]
%l1.add = add nuw nsw i16 %l1, 1
%cmp1 = icmp ult i16 %l1.add, 2
br i1 %cmp1, label %loop1, label %loop2.preheader
loop2.preheader: ; preds = %loop1
br label %loop2
loop2: ; preds = %loop2,
%loop2.preheader
%k2 = phi i16 [ %k2.add, %loop2 ], [ 182, %loop2.preheader ]
%l2 = phi i16 [ %l2.add, %loop2 ], [ 0, %loop2.preheader ]
%l2.add = add nuw nsw i16 %l2, 1
tail call void @foo(i16 %k2)
%k2.add = add nuw nsw i16 %k2, 1
%cmp2 = icmp ult i16 %l2.add, 2
br i1 %cmp2, label %loop2, label %loop2.end
loop2.end: ; preds = %loop2
%0 = add i16 182, 2
ret i16 %0
}
Note the difference in how the return value is calculated.
I suppose that indvars doesn't consider this to be the constant case that
should always be allowed, but I think it's quite unfortunate if this
simplification isn't done.</pre>
</div>
</p>
<hr>
<span>You are receiving this mail because:</span>
<ul>
<li>You are on the CC list for the bug.</li>
</ul>
</body>
</html>