<html>
<head>
<base href="http://llvm.org/bugs/" />
</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 --- - compiler-rt/lib/builtins/muldi3.c: Infinite loop/stack overflow in __muldi3"
href="http://llvm.org/bugs/show_bug.cgi?id=21514">21514</a>
</td>
</tr>
<tr>
<th>Summary</th>
<td>compiler-rt/lib/builtins/muldi3.c: Infinite loop/stack overflow in __muldi3
</td>
</tr>
<tr>
<th>Product</th>
<td>compiler-rt
</td>
</tr>
<tr>
<th>Version</th>
<td>unspecified
</td>
</tr>
<tr>
<th>Hardware</th>
<td>PC
</td>
</tr>
<tr>
<th>OS</th>
<td>Windows NT
</td>
</tr>
<tr>
<th>Status</th>
<td>NEW
</td>
</tr>
<tr>
<th>Severity</th>
<td>normal
</td>
</tr>
<tr>
<th>Priority</th>
<td>P
</td>
</tr>
<tr>
<th>Component</th>
<td>compiler-rt
</td>
</tr>
<tr>
<th>Assignee</th>
<td>unassignedbugs@nondot.org
</td>
</tr>
<tr>
<th>Reporter</th>
<td>troshkovdanil@gmail.com
</td>
</tr>
<tr>
<th>CC</th>
<td>llvmbugs@cs.uiuc.edu
</td>
</tr>
<tr>
<th>Classification</th>
<td>Unclassified
</td>
</tr></table>
<p>
<div>
<pre>__muldi3(di_int a, di_int b)
{
dwords x;
x.all = a;
dwords y;
y.all = b;
dwords r;
r.all = __muldsi3(x.s.low, y.s.low);
r.s.high += x.s.high * y.s.low + x.s.low * y.s.high;
return r.all;
}
Function recursively uses itself (r.s.high += x.s.high * y.s.low + x.s.low *
y.s.high).
It occurs quite legally...
Dumps (-mllvm -print-after-all):
; Function Attrs: nounwind
define i64 @__muldi3(i64 %a, i64 %b) #0 {
entry:
%x.sroa.0.0.extract.trunc = trunc i64 %a to i32
%x.sroa.3.0.extract.shift = lshr i64 %a, 32
%x.sroa.3.0.extract.trunc = trunc i64 %x.sroa.3.0.extract.shift to i32
%y.sroa.0.0.extract.trunc = trunc i64 %b to i32
%y.sroa.3.0.extract.shift = lshr i64 %b, 32
%y.sroa.3.0.extract.trunc = trunc i64 %y.sroa.3.0.extract.shift to i32
%call = call fastcc i64 @__muldsi3(i32 %x.sroa.0.0.extract.trunc, i32
%y.sroa.0.0.extract.trunc)
%r.sroa.0.0.extract.trunc = trunc i64 %call to i32
%r.sroa.2.0.extract.shift = lshr i64 %call, 32
%r.sroa.2.0.extract.trunc = trunc i64 %r.sroa.2.0.extract.shift to i32
%mul = mul i32 %x.sroa.3.0.extract.trunc, %y.sroa.0.0.extract.trunc
%mul12 = mul i32 %x.sroa.0.0.extract.trunc, %y.sroa.3.0.extract.trunc
%add = add i32 %mul, %mul12
%add15 = add i32 %r.sroa.2.0.extract.trunc, %add
%r.sroa.2.0.insert.ext = zext i32 %add15 to i64
%r.sroa.2.0.insert.shift = shl i64 %r.sroa.2.0.insert.ext, 32
%r.sroa.0.0.insert.ext = zext i32 %r.sroa.0.0.extract.trunc to i64
%r.sroa.0.0.insert.mask = and i64 %r.sroa.2.0.insert.shift, -4294967296
%r.sroa.0.0.insert.insert = or i64 %r.sroa.0.0.insert.mask,
%r.sroa.0.0.insert.ext
ret i64 %r.sroa.0.0.insert.insert
}
.....
*** IR Dump After Combine redundant instructions ***
; Function Attrs: nounwind
define i64 @__muldi3(i64 %a, i64 %b) #0 {
entry:
%x.sroa.0.0.extract.trunc = trunc i64 %a to i32
%x.sroa.3.0.extract.shift = lshr i64 %a, 32
%y.sroa.0.0.extract.trunc = trunc i64 %b to i32
%y.sroa.3.0.extract.shift = lshr i64 %b, 32
%call = call fastcc i64 @__muldsi3(i32 %x.sroa.0.0.extract.trunc, i32
%y.sroa.0.0.extract.trunc)
%mul = mul i64 %x.sroa.3.0.extract.shift, %b
%mul12 = mul i64 %y.sroa.3.0.extract.shift, %a
%add = add i64 %mul, %mul12
%add1519 = shl i64 %add, 32
%r.sroa.2.0.extract.shift20 = add i64 %call, %add1519
ret i64 %r.sroa.2.0.extract.shift20
}
And (mul i64) is __muldi3...
I've looked for similar problems in the Internet and found something like this:
<a href="https://groups.google.com/forum/#!topic/llvm-dev/c-i8S5tOSYg">https://groups.google.com/forum/#!topic/llvm-dev/c-i8S5tOSYg</a>
So it looks like the solution is to explicitly use builtins:
r.all = __muldsi3(x.s.low, y.s.low);
- r.s.high += x.s.high * y.s.low + x.s.low * y.s.high;
+ r.s.high += __mulsi3(x.s.high, y.s.low) + __mulsi3(x.s.low, y.s.high);
return r.all;
But we have no __mulsi3...
Any ideas?</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>