<html>
<head>
<base href="http://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 - [X86] Failure to use HADDPS for partial register result"
href="http://bugs.llvm.org/show_bug.cgi?id=32433">32433</a>
</td>
</tr>
<tr>
<th>Summary</th>
<td>[X86] Failure to use HADDPS for partial register result
</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>Backend: X86
</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>llvm-bugs@lists.llvm.org, spatel+llvm@rotateright.com
</td>
</tr></table>
<p>
<div>
<pre>While the 256-bit horizontal pair sums work fine (both on btver2 and btver1),
the 128-bit version completely fails:
#include <x86intrin.h>
void sum_pairs_128(__m128 f, float *p) {
p[0] = f[0] + f[1];
p[1] = f[2] + f[3];
}
void sum_pairs_256(__m256 f, float *p) {
p[0] = f[0] + f[1];
p[1] = f[2] + f[3];
p[2] = f[4] + f[5];
p[3] = f[6] + f[7];
}
clang -O3 -march=btver2
sum_pairs_128(float __vector(4), float*):
vmovshdup %xmm0, %xmm1 # xmm1 = xmm0[1,1,3,3]
vaddss %xmm1, %xmm0, %xmm1
vmovss %xmm1, (%rdi)
vpermilpd $1, %xmm0, %xmm1 # xmm1 = xmm0[1,0]
vpermilps $231, %xmm0, %xmm0 # xmm0 = xmm0[3,1,2,3]
vaddss %xmm0, %xmm1, %xmm0
vmovss %xmm0, 4(%rdi)
retq
sum_pairs_256(float __vector(8), float*):
vextractf128 $1, %ymm0, %xmm1
vhaddps %xmm1, %xmm0, %xmm0
vmovups %xmm0, (%rdi)
retq</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>