<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 - Use vpermps instead of vpermilps"
href="https://bugs.llvm.org/show_bug.cgi?id=47066">47066</a>
</td>
</tr>
<tr>
<th>Summary</th>
<td>Use vpermps instead of vpermilps
</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>Backend: X86
</td>
</tr>
<tr>
<th>Assignee</th>
<td>unassignedbugs@nondot.org
</td>
</tr>
<tr>
<th>Reporter</th>
<td>david.bolvansky@gmail.com
</td>
</tr>
<tr>
<th>CC</th>
<td>craig.topper@gmail.com, llvm-bugs@lists.llvm.org, llvm-dev@redking.me.uk, spatel+llvm@rotateright.com
</td>
</tr></table>
<p>
<div>
<pre>#define N 16
float f1[N] __attribute__ ((__aligned__(__BIGGEST_ALIGNMENT__)));
float f2[N] __attribute__ ((__aligned__(__BIGGEST_ALIGNMENT__)));
void foo ()
{
int j;
for (j=0; j<N; j++)
f1[j] += f2[N-j];
}
Clang -O3 -mavx512f:
foo(): # @foo()
vpermilps zmm0, zmmword ptr [rip + f2+4], 27 # zmm0 =
mem[3,2,1,0,7,6,5,4,11,10,9,8,15,14,13,12]
vshuff64x2 zmm0, zmm0, zmm0, 27 # zmm0 = zmm0[6,7,4,5,2,3,0,1]
vaddps zmm0, zmm0, zmmword ptr [rip + f1]
vmovaps zmmword ptr [rip + f1], zmm0
vzeroupper
ret
ICC -O3 -mavx512f:
foo():
vmovups zmm16, ZMMWORD PTR .L_2il0floatpacket.0[rip] #9.14
vpermps zmm17, zmm16, ZMMWORD PTR 4+f2[rip] #9.14
vaddps zmm18, zmm17, ZMMWORD PTR f1[rip] #9.5
vmovups ZMMWORD PTR f1[rip], zmm18 #9.5
vzeroupper #10.1
<a href="https://godbolt.org/z/7bToc7">https://godbolt.org/z/7bToc7</a>
Interestingly, if I changed float to double, Clang matches ICC:
<a href="https://godbolt.org/z/h13beq">https://godbolt.org/z/h13beq</a></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>