<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] Failure to use MOVMSK reduction result for branches"
href="https://bugs.llvm.org/show_bug.cgi?id=44565">44565</a>
</td>
</tr>
<tr>
<th>Summary</th>
<td>[X86] Failure to use MOVMSK reduction result for branches
</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, spatel+llvm@rotateright.com
</td>
</tr></table>
<p>
<div>
<pre><a href="https://godbolt.org/z/GLDirF">https://godbolt.org/z/GLDirF</a>
If we are selecting between 2 values using a comparison reduction then we use
the all_of/any_of MOVMSK result directly:
define i32 @select(<2 x double> %a0) {
%1 = fcmp ogt <2 x double> %a0, zeroinitializer
%2 = extractelement <2 x i1> %1, i32 0
%3 = extractelement <2 x i1> %1, i32 1
%4 = and i1 %2, %3
%5 = select i1 %4, i32 42, i32 88
ret i32 %5
}
select: # @select
vxorpd %xmm1, %xmm1, %xmm1
vcmpltpd %xmm0, %xmm1, %xmm0
vmovmskpd %xmm0, %eax
cmpb $3, %al
movl $42, %ecx
movl $88, %eax
cmovel %ecx, %eax
retq
But if we decided to create a branch instead then the MOVMSK bits are used
separately, meaning we end up with 2 compares+jmps instead:
define i32 @jmp(<2 x double> %a0) {
%1 = fcmp ogt <2 x double> %a0, zeroinitializer
%2 = extractelement <2 x i1> %1, i32 0
%3 = extractelement <2 x i1> %1, i32 1
%4 = and i1 %2, %3
br i1 %4, label %foo, label %bar
foo:
ret i32 42
bar:
ret i32 88
}
jmp: # @jmp
vxorpd %xmm1, %xmm1, %xmm1
vcmpltpd %xmm0, %xmm1, %xmm0
vmovmskpd %xmm0, %eax
testb $1, %al
je .LBB1_3
shrb %al
je .LBB1_3
movl $42, %eax
retq
.LBB1_3: # %bar
movl $88, %eax
retq
when I think it could be:
jmp: # @jmp
vxorpd %xmm1, %xmm1, %xmm1
vcmpltpd %xmm0, %xmm1, %xmm0
vmovmskpd %xmm0, %eax
cmp $3, %al
jne .LBB1_3
movl $42, %eax
retq
.LBB1_3: # %bar
movl $88, %eax
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>