boostorg/endian
1
0
Fork 1
mirror of https://github.com/boostorg/endian.git synced 2025-07-31 13:07:24 +02:00
Code Issues Packages Projects Releases Wiki Activity

Add timing tables.

Browse Source
This commit is contained in:
Beman
2013-05-28 08:02:56 -04:00
parent a55a44c67b
commit 01eba9b491
1 changed files with 71 additions and 104 deletions
Download Patch File Download Diff File Expand all files Collapse all files

175
doc/index.html
View File

@ -198,8 +198,7 @@ application concerns.</p>
</tr>
<tr>
<td valign="top">
<pre>
big_int32_t x;
<pre>big_int32_t x;
... read into x from a file ...
@ -241,8 +240,7 @@ generate exactly the same code for both.</p>
</tr>
<tr>
<td valign="top">
<pre>
big_int32_t x;
<pre>big_int32_t x;
... read into x from a file ...
@ -253,8 +251,7 @@ for (int32_t i = 0; i &lt; 1000000; ++i)
</pre>
</td>
<td>
<pre>
int32_t x;
<pre>int32_t x;
... read into x from a file ...
@ -290,121 +287,91 @@ stores, multiple instructions are required.</p>
<p>These tests were run against release builds on a circa 2012 4-core little endian X64 Intel Core i5-3570K
CPU @ 3.40GHz under Windows 7.</p>
<p>See <a href="../test/speed_test.cpp">speed_test.cpp</a>,
<a href="../test/speed_test_functions.hpp">speed_test_functions.hpp</a>,
<a href="../test/speed_test_functions.cpp">speed_test_functions.cpp</a>, and
<a href="../build/Jamfile.v2">Jamfile.v2</a> for the actual code and build. The timed functions are in a separate
compilation unit to prevent being optimized away.</p>
<p>Because the timings are anomalous, particularly for those high-lighted below
in yellow, the generated code from the GNU compiler was studied in detail. <b>
Exactly the same code is being generated for by-value conversion functions,
in-place conversion functions, and the endian types. Exactly the same code is
being generated whether intrinsics are used or not for 32 and 64-bit tests.</b>
<p>See <a href="../test/loop_time_test.cpp">loop_time_test.cpp</a> and
<a href="../build/Jamfile.v2">Jamfile.v2</a> for the actual code and build
setup.
(For GCC 4.7, there are no 16-bit intrinsics, so they are emulated by using
32-bit intrinsics.)</p>
<table border="1" cellpadding="5" cellspacing="0" style="border-collapse: collapse" bordercolor="#111111">
<tr>
<td bgcolor="#D7EEFF">
<p align="center"><b>Conclusions</b></p>
<p>The decision to use endian types or endian conversion functions should be
made based on application use cases, not assumptions about generated code
efficiency. Modern optimizers generate the same code for either approach,
and whether or not intrinsics are available.&nbsp; </td>
</tr>
</table>
<table border="1" cellpadding="5" cellspacing="0"style="border-collapse: collapse" bordercolor="#111111">
<tr><td colspan="6" align="center"><b>GNU g++ version 4.7.0</b></td></tr>
<tr><td colspan="6" align="center"><b> Iterations: 1,000,000,000, Intrinsics: __builtin_bswap16, etc.</b></td></tr>
<tr><td colspan="6" align="center"><b>GNU C++ version 4.7.0</b></td></tr>
<tr><td colspan="6" align="center"><b> Iterations: 1000000000, Intrinsics: __builtin_bswap16, etc.</b></td></tr>
<tr><td><b>Test Case</b></td>
<td align="center"><b>int<br>arg</b></td>
<td align="center"><b>int<br>value(arg)</b></td>
<td align="center"><b>int<br>in place(arg)</b></td>
<td align="center"><b>Endian<br>arg</b></td>
<td align="center"><b>Endian<br>type</b></td>
<td align="center"><b>Endian<br>conversion<br>function</b></td>
</tr>
<tr><td>16-bit aligned big endian</td><td align="right" bgcolor="#FFFFCC">2.71 s</td>
<td align="right">2.42 s</td><td align="right">2.42 s</td><td align="right">2.68 s</td></tr>
<tr><td>16-bit aligned little endian</td><td align="right">2.42 s</td>
<td align="right">2.40 s</td><td align="right">2.68 s</td><td align="right">2.45 s</td></tr>
<tr><td>32-bit aligned big endian</td><td align="right">2.68 s</td>
<td align="right">2.70 s</td><td align="right">2.70 s</td><td align="right">2.68 s</td></tr>
<tr><td>32-bit aligned little endian</td><td align="right">2.68 s</td>
<td align="right">2.68 s</td><td align="right">2.65 s</td><td align="right">2.68 s</td></tr>
<tr><td>64-bit aligned big endian</td><td align="right" bgcolor="#FFFFCC">2.96 s</td>
<td align="right" bgcolor="#FFFFCC">2.95 s</td>
<td align="right" bgcolor="#FFFFCC">2.95 s</td>
<td align="right" bgcolor="#FFFFCC">2.95 s</td></tr>
<tr><td>64-bit aligned little endian</td><td align="right">2.42 s</td>
<td align="right">2.40 s</td><td align="right">2.70 s</td><td align="right">2.42 s</td></tr>
<tr><td colspan="6" align="center"><b> Iterations: 1,000,000,000, Intrinsics: no byte swap intrinsics</b></td></tr>
<tr><td>16-bit aligned big endian</td><td align="right">1.37 s</td><td align="right">0.81 s</td></tr>
<tr><td>16-bit aligned little endian</td><td align="right">0.83 s</td><td align="right">0.81 s</td></tr>
<tr><td>16-bit unaligned big endian</td><td align="right">1.09 s</td><td align="right">0.83 s</td></tr>
<tr><td>16-bit unaligned little endian</td><td align="right">1.09 s</td><td align="right">0.81 s</td></tr>
<tr><td>32-bit aligned big endian</td><td align="right">0.98 s</td><td align="right">0.27 s</td></tr>
<tr><td>32-bit aligned little endian</td><td align="right">0.28 s</td><td align="right">0.27 s</td></tr>
<tr><td>32-bit unaligned big endian</td><td align="right">3.82 s</td><td align="right">0.27 s</td></tr>
<tr><td>32-bit unaligned little endian</td><td align="right">3.82 s</td><td align="right">0.27 s</td></tr>
<tr><td>64-bit aligned big endian</td><td align="right">1.65 s</td><td align="right">0.41 s</td></tr>
<tr><td>64-bit aligned little endian</td><td align="right">0.41 s</td><td align="right">0.41 s</td></tr>
<tr><td>64-bit unaligned big endian</td><td align="right">17.53 s</td><td align="right">0.41 s</td></tr>
<tr><td>64-bit unaligned little endian</td><td align="right">17.52 s</td><td align="right">0.41 s</td></tr>
<tr><td colspan="6" align="center"><b> Iterations: 1000000000, Intrinsics: no byte swap intrinsics</b></td></tr>
<tr><td><b>Test Case</b></td>
<td align="center"><b>int<br>arg</b></td>
<td align="center"><b>int<br>value(arg)</b></td>
<td align="center"><b>int<br>in place(arg)</b></td>
<td align="center"><b>Endian<br>arg</b></td>
<td align="center"><b>Endian<br>type</b></td>
<td align="center"><b>Endian<br>conversion<br>function</b></td>
</tr>
<tr><td>16-bit aligned big endian</td><td align="right" bgcolor="#FFFFCC">2.71 s</td>
<td align="right">2.42 s</td><td align="right">2.42 s</td><td align="right">2.68 s</td></tr>
<tr><td>16-bit aligned little endian</td><td align="right">2.42 s</td>
<td align="right">2.40 s</td><td align="right">2.68 s</td><td align="right">2.42 s</td></tr>
<tr><td>32-bit aligned big endian</td><td align="right">2.68 s</td>
<td align="right">2.70 s</td><td align="right">2.67 s</td><td align="right">2.70 s</td></tr>
<tr><td>32-bit aligned little endian</td><td align="right">2.68 s</td>
<td align="right">2.67 s</td><td align="right">2.70 s</td><td align="right">2.67 s</td></tr>
<tr><td>64-bit aligned big endian</td><td align="right" bgcolor="#FFFFCC">2.96 s</td>
<td align="right" bgcolor="#FFFFCC">2.95 s</td>
<td align="right" bgcolor="#FFFFCC">2.95 s</td>
<td align="right" bgcolor="#FFFFCC">2.93 s</td></tr>
<tr><td>64-bit aligned little endian</td><td align="right">2.42 s</td>
<td align="right">2.42 s</td><td align="right">2.67 s</td><td align="right">2.40 s</td></tr>
<tr><td>16-bit aligned big endian</td><td align="right">1.95 s</td><td align="right">0.81 s</td></tr>
<tr><td>16-bit aligned little endian</td><td align="right">0.83 s</td><td align="right">0.81 s</td></tr>
<tr><td>16-bit unaligned big endian</td><td align="right">1.19 s</td><td align="right">0.81 s</td></tr>
<tr><td>16-bit unaligned little endian</td><td align="right">1.20 s</td><td align="right">0.81 s</td></tr>
<tr><td>32-bit aligned big endian</td><td align="right">0.97 s</td><td align="right">0.28 s</td></tr>
<tr><td>32-bit aligned little endian</td><td align="right">0.27 s</td><td align="right">0.28 s</td></tr>
<tr><td>32-bit unaligned big endian</td><td align="right">4.10 s</td><td align="right">0.27 s</td></tr>
<tr><td>32-bit unaligned little endian</td><td align="right">4.10 s</td><td align="right">0.27 s</td></tr>
<tr><td>64-bit aligned big endian</td><td align="right">1.64 s</td><td align="right">0.42 s</td></tr>
<tr><td>64-bit aligned little endian</td><td align="right">0.41 s</td><td align="right">0.41 s</td></tr>
<tr><td>64-bit unaligned big endian</td><td align="right">17.52 s</td><td align="right">0.42 s</td></tr>
<tr><td>64-bit unaligned little endian</td><td align="right">17.52 s</td><td align="right">0.41 s</td></tr>
</table>
<p></p>
<table border="1" cellpadding="5" cellspacing="0"style="border-collapse: collapse" bordercolor="#111111">
<tr><td colspan="6" align="center"><b>Microsoft Visual C++ version 11.0</b></td></tr>
<tr><td colspan="6" align="center"><b> Iterations: 1,000,000,000, Intrinsics: cstdlib _byteswap_ushort, etc.</b></td></tr>
<tr><td colspan="6" align="center"><b> Iterations: 1000000000, Intrinsics: cstdlib _byteswap_ushort, etc.</b></td></tr>
<tr><td><b>Test Case</b></td>
<td align="center"><b>int<br>arg</b></td>
<td align="center"><b>int<br>value(arg)</b></td>
<td align="center"><b>int<br>in place(arg)</b></td>
<td align="center"><b>Endian<br>arg</b></td>
<td align="center"><b>Endian<br>type</b></td>
<td align="center"><b>Endian<br>conversion<br>function</b></td>
</tr>
<tr><td>16-bit aligned big endian</td><td align="right">1.90 s</td>
<td align="right">1.87 s</td><td align="right">1.89 s</td><td align="right">1.87 s</td></tr>
<tr><td>16-bit aligned little endian</td><td align="right">1.89 s</td>
<td align="right">1.87 s</td><td align="right">1.89 s</td><td align="right">1.87 s</td></tr>
<tr><td>32-bit aligned big endian</td><td align="right">1.89 s</td>
<td align="right">1.87 s</td><td align="right">1.89 s</td><td align="right">1.87 s</td></tr>
<tr><td>32-bit aligned little endian</td><td align="right">1.89 s</td>
<td align="right">1.87 s</td><td align="right">1.87 s</td><td align="right">1.89 s</td></tr>
<tr><td>64-bit aligned big endian</td><td align="right">1.87 s</td>
<td align="right">1.89 s</td><td align="right">1.87 s</td><td align="right">1.89 s</td></tr>
<tr><td>64-bit aligned little endian</td><td align="right">1.87 s</td>
<td align="right">1.87 s</td><td align="right">1.87 s</td><td align="right">1.89 s</td></tr>
<tr><td colspan="6" align="center"><b> Iterations: 1,000,000,000, Intrinsics: no byte swap intrinsics</b></td></tr>
<tr><td>16-bit aligned big endian</td><td align="right">2.18 s</td><td align="right">0.83 s</td></tr>
<tr><td>16-bit aligned little endian</td><td align="right">0.81 s</td><td align="right">0.83 s</td></tr>
<tr><td>16-bit unaligned big endian</td><td align="right">1.64 s</td><td align="right">0.83 s</td></tr>
<tr><td>16-bit unaligned little endian</td><td align="right">1.64 s</td><td align="right">0.83 s</td></tr>
<tr><td>32-bit aligned big endian</td><td align="right">0.83 s</td><td align="right">0.81 s</td></tr>
<tr><td>32-bit aligned little endian</td><td align="right">0.83 s</td><td align="right">0.81 s</td></tr>
<tr><td>32-bit unaligned big endian</td><td align="right">3.01 s</td><td align="right">0.83 s</td></tr>
<tr><td>32-bit unaligned little endian</td><td align="right">3.01 s</td><td align="right">0.81 s</td></tr>
<tr><td>64-bit aligned big endian</td><td align="right">1.09 s</td><td align="right">1.05 s</td></tr>
<tr><td>64-bit aligned little endian</td><td align="right">0.83 s</td><td align="right">1.03 s</td></tr>
<tr><td>64-bit unaligned big endian</td><td align="right">12.64 s</td><td align="right">1.01 s</td></tr>
<tr><td>64-bit unaligned little endian</td><td align="right">8.41 s</td><td align="right">0.83 s</td></tr>
<tr><td colspan="6" align="center"><b> Iterations: 1000000000, Intrinsics: no byte swap intrinsics</b></td></tr>
<tr><td><b>Test Case</b></td>
<td align="center"><b>int<br>arg</b></td>
<td align="center"><b>int<br>value(arg)</b></td>
<td align="center"><b>int<br>in place(arg)</b></td>
<td align="center"><b>Endian<br>arg</b></td>
<td align="center"><b>Endian<br>type</b></td>
<td align="center"><b>Endian<br>conversion<br>function</b></td>
</tr>
<tr><td>16-bit aligned big endian</td><td align="right">1.90 s</td>
<td align="right">1.89 s</td><td align="right">1.87 s</td><td align="right">1.87 s</td></tr>
<tr><td>16-bit aligned little endian</td><td align="right">1.89 s</td>
<td align="right">1.87 s</td><td align="right">1.89 s</td><td align="right">1.87 s</td></tr>
<tr><td>32-bit aligned big endian</td><td align="right">1.89 s</td>
<td align="right">1.87 s</td><td align="right">1.87 s</td><td align="right">1.89 s</td></tr>
<tr><td>32-bit aligned little endian</td><td align="right">1.87 s</td>
<td align="right">1.89 s</td><td align="right">1.87 s</td><td align="right">1.89 s</td></tr>
<tr><td>64-bit aligned big endian</td><td align="right" bgcolor="#FFFFCC">2.32 s</td>
<td align="right" bgcolor="#FFFFCC">2.46 s</td>
<td align="right" bgcolor="#FFFFCC">2.45 s</td>
<td align="right" bgcolor="#FFFFCC">2.34 s</td></tr>
<tr><td>64-bit aligned little endian</td><td align="right">1.87 s</td>
<td align="right">1.87 s</td><td align="right">1.89 s</td><td align="right">1.87 s</td></tr>
<tr><td>16-bit aligned big endian</td><td align="right">0.84 s</td><td align="right">0.81 s</td></tr>
<tr><td>16-bit aligned little endian</td><td align="right">0.83 s</td><td align="right">0.81 s</td></tr>
<tr><td>16-bit unaligned big endian</td><td align="right">1.65 s</td><td align="right">0.81 s</td></tr>
<tr><td>16-bit unaligned little endian</td><td align="right">1.65 s</td><td align="right">0.83 s</td></tr>
<tr><td>32-bit aligned big endian</td><td align="right">3.46 s</td><td align="right">0.83 s</td></tr>
<tr><td>32-bit aligned little endian</td><td align="right">0.81 s</td><td align="right">0.83 s</td></tr>
<tr><td>32-bit unaligned big endian</td><td align="right">3.01 s</td><td align="right">0.81 s</td></tr>
<tr><td>32-bit unaligned little endian</td><td align="right">3.01 s</td><td align="right">0.81 s</td></tr>
<tr><td>64-bit aligned big endian</td><td align="right">10.50 s</td><td align="right">0.83 s</td></tr>
<tr><td>64-bit aligned little endian</td><td align="right">0.83 s</td><td align="right">0.97 s</td></tr>
<tr><td>64-bit unaligned big endian</td><td align="right">12.62 s</td><td align="right">0.81 s</td></tr>
<tr><td>64-bit unaligned little endian</td><td align="right">8.42 s</td><td align="right">0.81 s</td></tr>
</table>
@ -458,7 +425,7 @@ Tim Blechmann, Tim Moore, tymofey, Tomas Puverle, Vincente Botet, Yuval Ronen
and Vitaly Budovski,.</p>
<hr>
<p>Last revised:
<!--webbot bot="Timestamp" s-type="EDITED" s-format="%d %B, %Y" startspan -->26 May, 2013<!--webbot bot="Timestamp" endspan i-checksum="13988" --></p>
<!--webbot bot="Timestamp" s-type="EDITED" s-format="%d %B, %Y" startspan -->28 May, 2013<!--webbot bot="Timestamp" endspan i-checksum="13992" --></p>
<p><EFBFBD> Copyright Beman Dawes, 2011, 2013</p>
<p>Distributed under the Boost Software License, Version 1.0. See
<a href="http://www.boost.org/LICENSE_1_0.txt">www.boost.org/ LICENSE_1_0.txt</a></p>