<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 - [X86][SSE] Failure to merge scalar float comparison results"
href="https://bugs.llvm.org/show_bug.cgi?id=51068">51068</a>
</td>
</tr>
<tr>
<th>Summary</th>
<td>[X86][SSE] Failure to merge scalar float comparison results
</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>craig.topper@gmail.com, llvm-bugs@lists.llvm.org, llvm-dev@redking.me.uk, pengfei.wang@intel.com, spatel+llvm@rotateright.com
</td>
</tr></table>
<p>
<div>
<pre><a href="https://simd.godbolt.org/z/vWb57hqzx">https://simd.godbolt.org/z/vWb57hqzx</a>
#include <x86intrin.h>
bool cmp2_eq(float x, float y, float z, float w) {
return (x == y) == (z == w);
}
define zeroext i1 @cmp2_eq(float %0, float %1, float %2, float %3) {
%5 = fcmp une float %0, %1
%6 = fcmp oeq float %2, %3
%7 = xor i1 %5, %6
ret i1 %7
}
cmp2:
ucomiss %xmm1, %xmm0
setp %al
setne %cl
orb %al, %cl
ucomiss %xmm3, %xmm2
setnp %dl
sete %al
andb %dl, %al
xorb %cl, %al
retq
We should be able to use cmpss (or cmpsd for doubles) and merge the results to
reduce fpu->gpr traffic and branch density.
Compare to vectors:
__v4si cmp2_eq(__v4sf x, __v4sf y, __v4sf z, __v4sf w) {
return (x == y) == (z == w);
}
define <4 x i32> @cmp2_(<4 x float> %0, <4 x float> %1, <4 x float> %2, <4 x
float> %3) {
%5 = fcmp une <4 x float> %0, %1
%6 = fcmp oeq <4 x float> %2, %3
%7 = xor <4 x i1> %6, %5
%8 = sext <4 x i1> %7 to <4 x i32>
ret <4 x i32> %8
}
cmp2_:
cmpneqps %xmm1, %xmm0
cmpeqps %xmm3, %xmm2
xorps %xmm2, %xmm0
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>