<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 - [AArch64] compare and branch with long immediate optimisation"
href="https://bugs.llvm.org/show_bug.cgi?id=52140">52140</a>
</td>
</tr>
<tr>
<th>Summary</th>
<td>[AArch64] compare and branch with long immediate optimisation
</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>Keywords</th>
<td>beginner
</td>
</tr>
<tr>
<th>Severity</th>
<td>enhancement
</td>
</tr>
<tr>
<th>Priority</th>
<td>P
</td>
</tr>
<tr>
<th>Component</th>
<td>Backend: AArch64
</td>
</tr>
<tr>
<th>Assignee</th>
<td>unassignedbugs@nondot.org
</td>
</tr>
<tr>
<th>Reporter</th>
<td>pavel.iliin@arm.com
</td>
</tr>
<tr>
<th>CC</th>
<td>arnaud.degrandmaison@arm.com, llvm-bugs@lists.llvm.org, smithp352@googlemail.com, Ties.Stuij@arm.com
</td>
</tr></table>
<p>
<div>
<pre>Сomparison with complex immediates followed by branch/cset can be improved.
Currently clang optimise
int feqi (int a)
{
return a == 0x111111;
}
long long feql (long long a)
{
return a == 0x222222;
}
int fnei (int a)
{
return a != 0x333333;
}
long long fnel (long long a)
{
return a != 0x444444;
}
into
feqi(int): // @feqi(int)
mov w8, #4369
movk w8, #17, lsl #16
cmp w0, w8
cset w0, eq
ret
feql(long long): // @feql(long long)
mov w8, #8738
movk w8, #34, lsl #16
cmp x0, x8
cset w0, eq
ret
fnei(int): // @fnei(int)
mov w8, #13107
movk w8, #51, lsl #16
cmp w0, w8
cset w0, ne
ret
fnel(long long): // @fnel(long long)
mov w8, #17476
movk w8, #68, lsl #16
cmp x0, x8
cset w0, ne
ret
However gcc performs better emitting sub+subs instead of mov+movk+cmp:
feqi(int):
sub w0, w0, #1118208
subs w0, w0, #273
cset w0, eq
ret
feql(long long):
sub x0, x0, #2236416
subs x0, x0, #546
cset x0, eq
ret
fnei(int):
sub w0, w0, #3354624
subs w0, w0, #819
cset w0, ne
ret
fnel(long long):
sub x0, x0, #4472832
subs x0, x0, #1092
cset x0, ne
ret
Details of GCC implementation:
<a href="https://gcc.gnu.org/legacy-ml/gcc-patches/2015-10/msg00800.html">https://gcc.gnu.org/legacy-ml/gcc-patches/2015-10/msg00800.html</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>