<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 - Fold smax(shuffle(x,undef,m),shuffle(y,undef,m)) -> shuffle(smax(x,y),undef,m)"
href="https://bugs.llvm.org/show_bug.cgi?id=46894">46894</a>
</td>
</tr>
<tr>
<th>Summary</th>
<td>Fold smax(shuffle(x,undef,m),shuffle(y,undef,m)) -> shuffle(smax(x,y),undef,m)
</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>Windows NT
</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>llvm-dev@redking.me.uk
</td>
</tr>
<tr>
<th>CC</th>
<td>lebedev.ri@gmail.com, llvm-bugs@lists.llvm.org, spatel+llvm@rotateright.com
</td>
</tr></table>
<p>
<div>
<pre>For standard binary operators, instcombine optimizes:
define <2 x i64> @swap_mul(<2 x i64> %a0, <2 x i64> %a1) {
%a = shufflevector <2 x i64> %a0, <2 x i64> %a1, <2 x i32> <i32 1, i32 0>
%b = shufflevector <2 x i64> %a0, <2 x i64> %a1, <2 x i32> <i32 1, i32 0>
%r = mul <2 x i64> %a, %b
ret <2 x i64> %r
}
-->
define <2 x i64> @swap_mul(<2 x i64> %a0, <2 x i64> %a1) {
%1 = mul <2 x i64> %a0, %a0
%r = shufflevector <2 x i64> %1, <2 x i64> undef, <2 x i32> <i32 1, i32 0>
ret <2 x i64> %r
}
But we there's no way to do this for binop style intrinsics
(smax/smin/umax/umin):
define <2 x i64> @swap_smax(<2 x i64> %a0, <2 x i64> %a1) {
%a = shufflevector <2 x i64> %a0, <2 x i64> %a1, <2 x i32> <i32 1, i32 0>
%b = shufflevector <2 x i64> %a0, <2 x i64> %a1, <2 x i32> <i32 1, i32 0>
%r = call <2 x i64> @llvm.smax.v2i64(<2 x i64> %a, <2 x i64> %b)
ret <2 x i64> %r
}
declare <2 x i64> @llvm.smax.v2i64(<2 x i64>, <2 x i64>)
As a stretch goal we could do this for triop intrinsics, such as fshl/fshr, as
well:
define <2 x i64> @swap_fshl(<2 x i64> %a0, <2 x i64> %a1, <2 x i64> %a2) {
%a = shufflevector <2 x i64> %a0, <2 x i64> %a0, <2 x i32> <i32 1, i32 0>
%b = shufflevector <2 x i64> %a1, <2 x i64> %a1, <2 x i32> <i32 1, i32 0>
%c = shufflevector <2 x i64> %a2, <2 x i64> %a2, <2 x i32> <i32 1, i32 0>
%r = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> %a, <2 x i64> %b, <2 x i64>
%c)
ret <2 x i64> %r
}
declare <2 x i64> @llvm.fshl.v2i64(<2 x i64>, <2 x i64>, <2 x i64>)</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>