<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 --- - [x86] unnecessary shuffling/moves with unary AVX/SSE scalar math ops"
href="http://llvm.org/bugs/show_bug.cgi?id=21507">21507</a>
</td>
</tr>
<tr>
<th>Summary</th>
<td>[x86] unnecessary shuffling/moves with unary AVX/SSE scalar math ops
</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>Each of these functions should produce a single math instruction; no shuffling
or moves are needed:
$ cat unary_sse.c
#include <xmmintrin.h>
__m128 recip(__m128 x) {
return _mm_rcp_ss(x);
}
__m128 square_root(__m128 x) {
return _mm_sqrt_ss(x);
}
__m128 recip_square_root(__m128 x) {
return _mm_rsqrt_ss(x);
}
------------------------------------------------------------------------------
With an AVX subtarget, we always get a bonus shuffle:
$ ./clang unary_sse.c -O1 -fomit-frame-pointer -march=btver2 -S -o -
...
_recip: ## @recip
.cfi_startproc
## BB#0: ## %entry
vrcpss %xmm0, %xmm0, %xmm1
vblendps $1, %xmm1, %xmm0, %xmm0 ## xmm0 = xmm1[0],xmm0[1,2,3]
retq
.cfi_endproc
.globl _square_root
.align 4, 0x90
_square_root: ## @square_root
.cfi_startproc
## BB#0: ## %entry
vsqrtss %xmm0, %xmm0, %xmm1
vblendps $1, %xmm1, %xmm0, %xmm0 ## xmm0 = xmm1[0],xmm0[1,2,3]
retq
.cfi_endproc
.globl _recip_square_root
.align 4, 0x90
_recip_square_root: ## @recip_square_root
.cfi_startproc
## BB#0: ## %entry
vrsqrtss %xmm0, %xmm0, %xmm1
vblendps $1, %xmm1, %xmm0, %xmm0 ## xmm0 = xmm1[0],xmm0[1,2,3]
retq
With an SSE-only subtarget, we get bonus moves:
$ ./clang unary_sse.c -O1 -fomit-frame-pointer -march=core2 -S -o -
...
_recip: ## @recip
.cfi_startproc
## BB#0: ## %entry
movaps %xmm0, %xmm1
rcpss %xmm1, %xmm1
movss %xmm1, %xmm0
retq
.cfi_endproc
.globl _square_root
.align 4, 0x90
_square_root: ## @square_root
.cfi_startproc
## BB#0: ## %entry
sqrtss %xmm0, %xmm1
movss %xmm1, %xmm0
retq
.cfi_endproc
.globl _recip_square_root
.align 4, 0x90
_recip_square_root: ## @recip_square_root
.cfi_startproc
## BB#0: ## %entry
movaps %xmm0, %xmm1
rsqrtss %xmm1, %xmm1
movss %xmm1, %xmm0
retq
-----------------------------------------------------------------------------
$ ./clang -v
clang version 3.6.0 (221350)
Target: x86_64-apple-darwin13.4.0
Thread model: posix</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>