<html>
<head>
<base href="http://llvm.org/bugs/" />
</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 --- - fast vector fneg not optimized like fast scalar fneg"
href="http://llvm.org/bugs/show_bug.cgi?id=20556">20556</a>
</td>
</tr>
<tr>
<th>Summary</th>
<td>fast vector fneg not optimized like fast scalar fneg
</td>
</tr>
<tr>
<th>Product</th>
<td>new-bugs
</td>
</tr>
<tr>
<th>Version</th>
<td>trunk
</td>
</tr>
<tr>
<th>Hardware</th>
<td>PC
</td>
</tr>
<tr>
<th>OS</th>
<td>All
</td>
</tr>
<tr>
<th>Status</th>
<td>NEW
</td>
</tr>
<tr>
<th>Severity</th>
<td>normal
</td>
</tr>
<tr>
<th>Priority</th>
<td>P
</td>
</tr>
<tr>
<th>Component</th>
<td>new bugs
</td>
</tr>
<tr>
<th>Assignee</th>
<td>unassignedbugs@nondot.org
</td>
</tr>
<tr>
<th>Reporter</th>
<td>spatel+llvm@rotateright.com
</td>
</tr>
<tr>
<th>CC</th>
<td>llvmbugs@cs.uiuc.edu
</td>
</tr>
<tr>
<th>Classification</th>
<td>Unclassified
</td>
</tr></table>
<p>
<div>
<pre>This is a test for optimizing the "wrong" definition of fneg (one that uses
+0.0 rather than -0.0) when enable-unsafe-fp-math is used.
Using llc built from r214897 on x86-64:
$ cat fast_fneg.ll
define float @fneg(float %x) #0 {
entry:
%sub = fsub float 0.000000e+00, %x
ret float %sub
}
define <4 x float> @fneg4(<4 x float> %x) #0 {
entry:
%sub = fsub <4 x float> zeroinitializer, %x
ret <4 x float> %sub
}
$ ./llc -enable-unsafe-fp-math fast_fneg.ll -o -
.section __TEXT,__text,regular,pure_instructions
.macosx_version_min 13, 3
.section __TEXT,__literal16,16byte_literals
.align 4
LCPI0_0:
.long 2147483648 ## float -0.000000e+00
.long 2147483648 ## float -0.000000e+00
.long 2147483648 ## float -0.000000e+00
.long 2147483648 ## float -0.000000e+00
.section __TEXT,__text,regular,pure_instructions
.globl _fneg
.align 4, 0x90
_fneg: ## @fneg
.cfi_startproc
## BB#0: ## %entry
xorps LCPI0_0(%rip), %xmm0
retq
.cfi_endproc
.globl _fneg4
.align 4, 0x90
_fneg4: ## @fneg4
.cfi_startproc
## BB#0: ## %entry
xorps %xmm1, %xmm1
subps %xmm0, %xmm1
movaps %xmm1, %xmm0
retq
.cfi_endproc
-------------------------------------
We optimized the scalar case to load a constant 0x80000000 to flip the sign bit
and avoid a subtraction. Shouldn't we do the same thing for the vector case?</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>