<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 - Suboptimal vectorization"
href="https://bugs.llvm.org/show_bug.cgi?id=47436">47436</a>
</td>
</tr>
<tr>
<th>Summary</th>
<td>Suboptimal vectorization
</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>Severity</th>
<td>enhancement
</td>
</tr>
<tr>
<th>Priority</th>
<td>P
</td>
</tr>
<tr>
<th>Component</th>
<td>Loop Optimizer
</td>
</tr>
<tr>
<th>Assignee</th>
<td>unassignedbugs@nondot.org
</td>
</tr>
<tr>
<th>Reporter</th>
<td>david.bolvansky@gmail.com
</td>
</tr>
<tr>
<th>CC</th>
<td>llvm-bugs@lists.llvm.org
</td>
</tr></table>
<p>
<div>
<pre>#include<stdint.h>
uint8_t data[16];
void test(unsigned i)
{
unsigned j;
for (j = 0; j < 16; j++)
data[j] = (i + j) >> 8;
}
Clang -O3 -march=haswell:
test(unsigned int): # @test(unsigned int)
lea eax, [rdi + 12]
vmovd xmm0, eax
lea eax, [rdi + 13]
vpinsrd xmm0, xmm0, eax, 1
lea eax, [rdi + 14]
vpinsrd xmm0, xmm0, eax, 2
lea eax, [rdi + 15]
vpinsrd xmm0, xmm0, eax, 3
lea eax, [rdi + 8]
vmovd xmm1, eax
lea eax, [rdi + 9]
vpinsrd xmm1, xmm1, eax, 1
lea eax, [rdi + 10]
vpinsrd xmm1, xmm1, eax, 2
lea eax, [rdi + 11]
vpinsrd xmm1, xmm1, eax, 3
lea eax, [rdi + 4]
vmovd xmm2, eax
lea eax, [rdi + 5]
vpinsrd xmm2, xmm2, eax, 1
lea eax, [rdi + 6]
vpinsrd xmm2, xmm2, eax, 2
lea eax, [rdi + 7]
vpinsrd xmm2, xmm2, eax, 3
lea eax, [rdi + 1]
vmovd xmm3, edi
vpinsrd xmm3, xmm3, eax, 1
lea eax, [rdi + 2]
vpinsrd xmm3, xmm3, eax, 2
lea eax, [rdi + 3]
vpinsrd xmm3, xmm3, eax, 3
vinserti128 ymm0, ymm1, xmm0, 1
vinserti128 ymm1, ymm3, xmm2, 1
vmovdqa ymm2, ymmword ptr [rip + .LCPI0_0] # ymm2 =
[1,2,5,6,9,10,13,14,9,10,13,14,13,14,15,128,17,18,21,22,25,26,29,30,25,26,29,30,29,30,31,128]
vpshufb ymm1, ymm1, ymm2
vpermq ymm1, ymm1, 232 # ymm1 = ymm1[0,2,2,3]
vmovdqa xmm3, xmmword ptr [rip + .LCPI0_1] # xmm3 =
[255,255,255,255,255,255,255,255]
vpand xmm1, xmm1, xmm3
vpshufb ymm0, ymm0, ymm2
vpermq ymm0, ymm0, 232 # ymm0 = ymm0[0,2,2,3]
vpand xmm0, xmm0, xmm3
vpackuswb xmm0, xmm1, xmm0
vmovdqa xmmword ptr [rip + data], xmm0
vzeroupper
ret
GCC -O3 -march=haswell:
test(unsigned int):
vmovd xmm0, edi
vmovdqa xmm2, XMMWORD PTR .LC2[rip]
vpbroadcastw xmm0, xmm0
vpaddw xmm1, xmm0, XMMWORD PTR .LC0[rip]
vpaddw xmm0, xmm0, XMMWORD PTR .LC1[rip]
vpsrlw xmm1, xmm1, 8
vpsrlw xmm0, xmm0, 8
vpand xmm1, xmm2, xmm1
vpand xmm2, xmm2, xmm0
vpackuswb xmm0, xmm1, xmm2
vmovdqa XMMWORD PTR data[rip], xmm0
ret
<a href="https://gcc.godbolt.org/z/zP4nWM">https://gcc.godbolt.org/z/zP4nWM</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>