1 <?xml version="1.0" encoding="utf-8" ?>
2 <!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN" "http://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd">
3 <html xmlns="http://www.w3.org/1999/xhtml" xml:lang="en" lang="en">
5 <meta http-equiv="Content-Type" content="text/html; charset=utf-8" />
6 <meta name="generator" content="Docutils 0.12: http://docutils.sourceforge.net/" />
7 <title>LBM Benchmark Kernels Documentation</title>
8 <style type="text/css">
11 :Author: David Goodger (goodger@python.org)
12 :Id: $Id: html4css1.css 7614 2013-02-21 15:55:51Z milde $
13 :Copyright: This stylesheet has been placed in the public domain.
15 Default cascading style sheet for the HTML output of Docutils.
17 See http://docutils.sf.net/docs/howto/html-stylesheets.html for how to
18 customize this style sheet.
21 /* used to remove borders from tables and images */
22 .borderless, table.borderless td, table.borderless th {
25 table.borderless td, table.borderless th {
26 /* Override padding for "table.docutils td" with "! important".
27 The right padding separates the table cells. */
28 padding: 0 0.5em 0 0 ! important }
31 /* Override more specific margin styles with "! important". */
32 margin-top: 0 ! important }
34 .last, .with-subtitle {
35 margin-bottom: 0 ! important }
41 text-decoration: none ;
48 margin-bottom: 0.5em }
50 object[type="image/svg+xml"], object[type="application/x-shockwave-flash"] {
54 /* Uncomment (and remove this text!) to get bold-faced definition list terms
62 div.abstract p.topic-title {
66 div.admonition, div.attention, div.caution, div.danger, div.error,
67 div.hint, div.important, div.note, div.tip, div.warning {
69 border: medium outset ;
72 div.admonition p.admonition-title, div.hint p.admonition-title,
73 div.important p.admonition-title, div.note p.admonition-title,
74 div.tip p.admonition-title {
76 font-family: sans-serif }
78 div.attention p.admonition-title, div.caution p.admonition-title,
79 div.danger p.admonition-title, div.error p.admonition-title,
80 div.warning p.admonition-title, .code .error {
83 font-family: sans-serif }
85 /* Uncomment (and remove this text!) to get reduced vertical space in
87 div.compound .compound-first, div.compound .compound-middle {
88 margin-bottom: 0.5em }
90 div.compound .compound-last, div.compound .compound-middle {
99 div.dedication p.topic-title {
107 div.footer, div.header {
116 div.line-block div.line-block {
122 margin: 0 0 0.5em 1em ;
123 border: medium outset ;
125 background-color: #ffffee ;
130 div.sidebar p.rubric {
131 font-family: sans-serif ;
134 div.system-messages {
137 div.system-messages h1 {
141 border: medium outset ;
144 div.system-message p.system-message-title {
151 h1.section-subtitle, h2.section-subtitle, h3.section-subtitle,
152 h4.section-subtitle, h5.section-subtitle, h6.section-subtitle {
164 img.align-left, .figure.align-left, object.align-left {
169 img.align-right, .figure.align-right, object.align-right {
174 img.align-center, .figure.align-center, object.align-center {
190 /* reset inner alignment in figures */
192 text-align: inherit }
194 /* div.align-center * { */
195 /* text-align: left } */
197 ol.simple, ul.simple {
201 list-style: decimal }
204 list-style: lower-alpha }
207 list-style: upper-alpha }
210 list-style: lower-roman }
213 list-style: upper-roman }
227 white-space: nowrap }
236 font-family: sans-serif ;
241 font-family: sans-serif ;
252 pre.literal-block, pre.doctest-block, pre.math, pre.code {
256 pre.code .ln { color: grey; } /* line numbers */
257 pre.code, code { background-color: #eeeeee }
258 pre.code .comment, code .comment { color: #5C6576 }
259 pre.code .keyword, code .keyword { color: #3B0D06; font-weight: bold }
260 pre.code .literal.string, code .literal.string { color: #0C5404 }
261 pre.code .name.builtin, code .name.builtin { color: #352B84 }
262 pre.code .deleted, code .deleted { background-color: #DEB0A1}
263 pre.code .inserted, code .inserted { background-color: #A3D289}
266 font-family: sans-serif ;
267 font-style: oblique }
269 span.classifier-delimiter {
270 font-family: sans-serif ;
274 font-family: sans-serif }
277 white-space: nowrap }
285 span.section-subtitle {
286 /* font-size relative to parent (h1..h6 element) */
290 border-left: solid 1px gray;
298 margin-bottom: 0.5em }
301 border-left: solid 1px black;
304 table.docutils td, table.docutils th,
305 table.docinfo td, table.docinfo th {
306 padding-left: 0.5em ;
307 padding-right: 0.5em ;
308 vertical-align: top }
310 table.docutils th.field-name, table.docinfo th.docinfo-name {
313 white-space: nowrap ;
316 /* "booktabs" style (no vertical lines) */
317 table.docutils.booktabs {
319 border-top: 2px solid;
320 border-bottom: 2px solid;
321 border-collapse: collapse;
323 table.docutils.booktabs * {
326 table.docutils.booktabs th {
327 border-bottom: thin solid;
331 h1 tt.docutils, h2 tt.docutils, h3 tt.docutils,
332 h4 tt.docutils, h5 tt.docutils, h6 tt.docutils {
336 list-style-type: none }
339 <style type="text/css">
342 h1, h2, h3, h4, h5, h6 {
343 font-family: sans-serif;
345 background-color: #dcdcdc;
349 background-color: gray;
365 table, table.docutils, td, th {
369 table.citation, table.footnote {
374 background-color: lavender ;
378 xxbackground-color: aliceblue;
379 background-color: white;
382 xxbackground-color: lavender;
383 background-color: whitesmoke;
391 <div class="document" id="lbm-benchmark-kernels-documentation">
392 <h1 class="title">LBM Benchmark Kernels Documentation</h1>
394 <!-- # - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
397 # Markus Wittmann, 2016-2017
398 # RRZE, University of Erlangen-Nuremberg, Germany
399 # markus.wittmann -at- fau.de or hpc -at- rrze.fau.de
402 # LSS, University of Erlangen-Nuremberg, Germany
404 # This file is part of the Lattice Boltzmann Benchmark Kernels (LbmBenchKernels).
406 # LbmBenchKernels is free software: you can redistribute it and/or modify
407 # it under the terms of the GNU General Public License as published by
408 # the Free Software Foundation, either version 3 of the License, or
409 # (at your option) any later version.
411 # LbmBenchKernels is distributed in the hope that it will be useful,
412 # but WITHOUT ANY WARRANTY; without even the implied warranty of
413 # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
414 # GNU General Public License for more details.
416 # You should have received a copy of the GNU General Public License
417 # along with LbmBenchKernels. If not, see <http://www.gnu.org/licenses/>.
419 # - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -->
420 <div class="contents topic" id="contents">
421 <p class="topic-title first">Contents</p>
422 <ul class="auto-toc simple">
423 <li><a class="reference internal" href="#compilation" id="id2">1 Compilation</a><ul class="auto-toc">
424 <li><a class="reference internal" href="#debug-and-verification" id="id3">1.1 Debug and Verification</a></li>
425 <li><a class="reference internal" href="#benchmarking" id="id4">1.2 Benchmarking</a></li>
426 <li><a class="reference internal" href="#release-and-verification" id="id5">1.3 Release and Verification</a></li>
427 <li><a class="reference internal" href="#compilers" id="id6">1.4 Compilers</a></li>
428 <li><a class="reference internal" href="#cleaning" id="id7">1.5 Cleaning</a></li>
429 <li><a class="reference internal" href="#options-summary" id="id8">1.6 Options Summary</a></li>
432 <li><a class="reference internal" href="#invocation" id="id9">2 Invocation</a><ul class="auto-toc">
433 <li><a class="reference internal" href="#command-line-parameters" id="id10">2.1 Command Line Parameters</a></li>
434 <li><a class="reference internal" href="#kernels" id="id11">2.2 Kernels</a></li>
437 <li><a class="reference internal" href="#id1" id="id12">3 Benchmarking</a><ul class="auto-toc">
438 <li><a class="reference internal" href="#padding" id="id13">3.1 Padding</a></li>
441 <li><a class="reference internal" href="#geometries" id="id14">4 Geometries</a></li>
442 <li><a class="reference internal" href="#results" id="id15">5 Results</a></li>
443 <li><a class="reference internal" href="#licence" id="id16">6 Licence</a></li>
444 <li><a class="reference internal" href="#acknowledgements" id="id17">7 Acknowledgements</a></li>
447 <div class="section" id="compilation">
448 <h1><a class="toc-backref" href="#id2">1 Compilation</a></h1>
449 <p>The benchmark framework currently supports only Linux systems and the GCC and
450 Intel compilers. Every other configuration probably requires adjustment inside
451 the code and the makefiles. Further some code might be platform or at least
453 <p>The benchmark can be build via <tt class="docutils literal">make</tt> from the <tt class="docutils literal">src</tt> subdirectory. This will
454 generate one binary which hosts all implemented benchmark kernels.</p>
455 <p>Binaries are located under the <tt class="docutils literal">bin</tt> subdirectory and will have different names
456 depending on compiler and build configuration.</p>
457 <div class="section" id="debug-and-verification">
458 <h2><a class="toc-backref" href="#id3">1.1 Debug and Verification</a></h2>
459 <pre class="literal-block">
460 make BUILD=debug BENCHMARK=off
462 <p>Running <tt class="docutils literal">make</tt> with <tt class="docutils literal">BUILD=debug</tt> builds the debug version of
463 the benchmark kernels, where no optimizations are performed, line numbers and
464 debug symbols are included as well as <tt class="docutils literal">DEBUG</tt> will be defined. The resulting
465 binary will be found in the <tt class="docutils literal">bin</tt> subdirectory and named
466 <tt class="docutils literal"><span class="pre">lbmbenchk-linux-<compiler>-debug</span></tt>.</p>
467 <p>Specifying <tt class="docutils literal">BENCHMARK=off</tt> turns on verification
468 (<tt class="docutils literal">VERIFICATION=on</tt>), statistics (<tt class="docutils literal">STATISTICS=on</tt>), and VTK output
469 (<tt class="docutils literal">VTK_OUTPUT=on</tt>) enabled.</p>
470 <p>Please note that the generated binary will therefore
471 exhibit a poor performance.</p>
473 <div class="section" id="benchmarking">
474 <h2><a class="toc-backref" href="#id4">1.2 Benchmarking</a></h2>
475 <p>To generate a binary for benchmarking run make with</p>
476 <pre class="literal-block">
479 <p>As default <tt class="docutils literal">BENCHMARK=on</tt> and <tt class="docutils literal">BUILD=release</tt> is set, where
480 BUILD=release turns optimizations on and <tt class="docutils literal">BENCHMARK=on</tt> disables
481 verfification, statistics, and VTK output.</p>
483 <div class="section" id="release-and-verification">
484 <h2><a class="toc-backref" href="#id5">1.3 Release and Verification</a></h2>
485 <p>Verification with the debug builds can be extremely slow. Hence verification
486 capabilities can be build with release builds:</p>
487 <pre class="literal-block">
491 <div class="section" id="compilers">
492 <h2><a class="toc-backref" href="#id6">1.4 Compilers</a></h2>
493 <p>Currently only the GCC and Intel compiler under Linux are supported. Between
494 both configuration can be chosen via <tt class="docutils literal"><span class="pre">CONFIG=linux-gcc</span></tt> or
495 <tt class="docutils literal"><span class="pre">CONFIG=linux-intel</span></tt>.</p>
497 <div class="section" id="cleaning">
498 <h2><a class="toc-backref" href="#id7">1.5 Cleaning</a></h2>
499 <p>For each configuration and build (debug/release) a subdirectory under the
500 <tt class="docutils literal">src/obj</tt> directory is created where the dependency and object files are
503 <pre class="literal-block">
504 make CONFIG=... BUILD=... clean
506 <p>a specific combination is select and cleaned, whereas with</p>
507 <pre class="literal-block">
510 <p>all object and dependency files are deleted.</p>
512 <div class="section" id="options-summary">
513 <h2><a class="toc-backref" href="#id8">1.6 Options Summary</a></h2>
514 <p>Options that can be specified when building the framework with make:</p>
515 <table border="1" class="docutils">
528 <tr><td>BENCHMARK</td>
531 <td>If enabled, disables VERIFICATION, STATISTICS, VTK_OUTPUT. If disabled enables the three former options.</td>
534 <td>debug, release</td>
536 <td>No optimization, debug symbols, DEBUG defined.</td>
539 <td>linux-gcc, linux-intel</td>
541 <td>Select GCC or Intel compiler.</td>
546 <td>Determines which ISA extension is used for macro definitions. This is <em>not</em> the architecture the compiler generates code for.</td>
551 <td>OpenMP, i.,e.. threading support.</td>
553 <tr><td>STATISTICS</td>
556 <td>View statistics, like density etc, during simulation.</td>
561 <td>Via TARCH the architecture the compiler generates code for can be overridden. The value depends on the chosen compiler.</td>
563 <tr><td>VERIFICATION</td>
566 <td>Turn verification on/off.</td>
568 <tr><td>VTK_OUTPUT</td>
571 <td>Enable/Disable VTK file output.</td>
577 <div class="section" id="invocation">
578 <h1><a class="toc-backref" href="#id9">2 Invocation</a></h1>
579 <p>Running the binary will print among the GPL licence header a line like the following:</p>
580 <pre class="literal-block">
581 LBM Benchmark Kernels 0.1, compiled Jul 5 2017 21:59:22, type: verification
583 <p>if verfication was enabled during compilation or</p>
584 <pre class="literal-block">
585 LBM Benchmark Kernels 0.1, compiled Jul 5 2017 21:59:22, type: benchmark
587 <p>if verfication was disabled during compilation.</p>
588 <div class="section" id="command-line-parameters">
589 <h2><a class="toc-backref" href="#id10">2.1 Command Line Parameters</a></h2>
590 <p>Running the binary with <tt class="docutils literal"><span class="pre">-h</span></tt> list all available parameters:</p>
591 <pre class="literal-block">
595 [-dims XxYyZ] [-geometry box|channel|pipe|blocks[-<block size>]] [-iterations <iterations>] [-lattice-dump-ascii]
596 [-rho-in <density>] [-rho-out <density] [-omega <omega>] [-kernel <kernel>]
598 [-t <number of threads>]
601 -- <kernel specific parameters>
603 -list List available kernels.
605 -dims XxYxZ Specify geometry dimensions.
607 -geometry blocks-<block size>
608 Geometetry with blocks of size <block size> regularily layout out.
610 <p>If an option is specified multiple times the last one overrides previous ones.
611 This holds also true for <tt class="docutils literal"><span class="pre">-verify</span></tt> which sets geometry dimensions,
612 iterations, etc, which can afterward be override, e.g.:</p>
613 <pre class="literal-block">
614 $ bin/lbmbenchk-linux-intel-release -verfiy -dims 32x32x32
616 <p>Kernel specific parameters can be opatained via selecting the specific kernel
617 and passing <tt class="docutils literal"><span class="pre">-h</span></tt> as parameter:</p>
618 <pre class="literal-block">
619 $ bin/lbmbenchk-linux-intel-release -kernel kernel-name -- -h
622 [-blk <n>] [-blk-[xyz] <n>]
624 <p>A list of all available kernels can be obtained via <tt class="docutils literal"><span class="pre">-list</span></tt>:</p>
625 <pre class="literal-block">
626 $ ../bin/lbmbenchk-linux-gcc-debug -list
627 Lattice Boltzmann Benchmark Kernels (LbmBenchKernels) Copyright (C) 2016, 2017 LSS, RRZE
628 This program comes with ABSOLUTELY NO WARRANTY; for details see LICENSE.
629 This is free software, and you are welcome to redistribute it under certain conditions.
631 LBM Benchmark Kernels 0.1, compiled Jul 5 2017 21:59:22, type: verification
632 Available kernels to benchmark:
637 list-pull-split-nt-1s-soa
638 list-pull-split-nt-2s-soa
653 <div class="section" id="kernels">
654 <h2><a class="toc-backref" href="#id11">2.2 Kernels</a></h2>
655 <p>The following list shortly describes available kernels:</p>
657 <li>push-soa/push-aos/pull-soa/pull-aos:
658 Unoptimized kernels (but stream/collide are already fused) using two grids as
659 source and destination. Implement push/pull semantics as well structure of
660 arrays (soa) or array of structures (aos) layout.</li>
661 <li>blk-push-soa/blk-push-aos/blk-pull-soa/blk-pull-aos:
662 The same as the unoptimized kernels without the blk prefix, except that they support
663 spatial blocking, i.e. loop blocking of the three loops used to iterate over
664 the lattice. Here manual work sharing for OpenMP is used.</li>
665 <li>list-push-soa/list-push-aos/list-pull-soa/list-pull-aos:
666 The same as the unoptimized kernels without the list prefix, but for indirect addressing.
667 Here only a 1D vector of is used to store the fluid nodes, omitting the
668 obstacles. An adjacency list is used to recover the neighborhood associations.</li>
669 <li>list-pull-split-nt-1s-soa/list-pull-split-nt-2s-soa:
670 Optimized variant of list-pull-soa. Chunks of the lattice are processed as
671 once. Postcollision values are written back via nontemporal stores in 18 (1s)
672 or 9 (2s) loops.</li>
673 <li>list-aa-aos/list-aa-soa:
674 Unoptimized implementation of the AA pattern for the 1D vector with adjacency
675 list. Supported are array of structures (aos) and structure of arrays (soa)
676 data layout is supported.</li>
678 Implementation of AA pattern with intrinsics for the 1D vector with adjacency
679 list. Furthermore it contains a vectorized even time step and run length
680 coding to reduce the loop balance of the odd time step.</li>
682 All optimizations of list-aa-ria-soa. Additional with partial vectorization
683 of the odd time step.</li>
685 <p>Note that all array of structures (aos) kernels might require blocking
686 (depending on the domain size) to reach the performance of their structure of
687 arrays (soa) counter parts.</p>
688 <p>The following table summarizes the properties of the kernels. Here <strong>D</strong> means
689 direct addressing, i.e. full array, <strong>I</strong> means indirect addressing, i.e. 1D
690 vector with adjacency list, <strong>x</strong> means supported, whereas <strong>--</strong> means unsupported.
691 The loop balance B_l is computed for D3Q19 model with double precision floating
692 point for PDFs (8 byte) and 4 byte integers for the index (adjacency list).
693 As list-aa-ria-soa and list-aa-pv-soa support run length coding their effective
694 loop balance depends on the geometry. The effective loop balance is printed
696 <table border="1" class="docutils">
706 <thead valign="bottom">
707 <tr><th class="head">kernel name</th>
708 <th class="head">prop. step</th>
709 <th class="head">data layout</th>
710 <th class="head">addr.</th>
711 <th class="head">parallel</th>
712 <th class="head">blocking</th>
713 <th class="head">B_l [B/FLUP]</th>
717 <tr><td>push-soa</td>
725 <tr><td>push-aos</td>
733 <tr><td>pull-soa</td>
741 <tr><td>pull-aos</td>
749 <tr><td>blk-push-soa</td>
757 <tr><td>blk-push-aos</td>
765 <tr><td>blk-pull-soa</td>
773 <tr><td>blk-pull-aos</td>
781 <tr><td>list-push-soa</td>
789 <tr><td>list-push-aos</td>
797 <tr><td>list-pull-soa</td>
805 <tr><td>list-pull-aos</td>
813 <tr><td>list-pull-split-nt-1s</td>
821 <tr><td>list-pull-split-nt-2s</td>
829 <tr><td>list-aa-soa</td>
837 <tr><td>list-aa-aos</td>
845 <tr><td>list-aa-ria-soa</td>
853 <tr><td>list-aa-pv-soa</td>
865 <div class="section" id="id1">
866 <h1><a class="toc-backref" href="#id12">3 Benchmarking</a></h1>
867 <p>Correct benchmarking is a nontrivial task. Whenever benchmark results should be
868 created make sure the binary was compiled with:</p>
870 <li><tt class="docutils literal">BENCHMARK=on</tt> (default if not overriden) and</li>
871 <li><tt class="docutils literal">BUILD=release</tt> (default if not overriden) and</li>
872 <li>the correct ISA for macros is used, selected via <tt class="docutils literal">ISA</tt> and</li>
873 <li>use <tt class="docutils literal">TARCH</tt> to specify the architecture the compiler generates code for.</li>
875 <p>During benchmarking pinning should be used via the <tt class="docutils literal"><span class="pre">-pin</span></tt> parameter. Running
876 a benchmark with 10 threads an pin them to the first 10 cores works like</p>
877 <pre class="literal-block">
878 $ bin/lbmbenchk-linux-intel-release ... -t 10 -pin $(seq -s , 0 9)
880 <p>Things the binary does nor check or controll:</p>
882 <li>transparent huge pages: when allocating memory small 4 KiB pages might be
883 replaced with larger ones. This is in general a good thing, but if this is
884 really the case, depends on the system settings (check e.g. the status of
885 <tt class="docutils literal">/sys/kernel/mm/transparent_hugepage/enabled</tt>).
886 Currently <tt class="docutils literal">madvise(MADV_HUGEPAGE)</tt> is used for allocations which are aligned to
887 a 4 KiB page, which should be the case for the lattices.
888 This should result in huge pages except THP is disabled on the machine.
889 (NOTE: madvise() is used if <tt class="docutils literal">HAVE_HUGE_PAGES</tt> is defined, which is currently
890 hard coded defined in <tt class="docutils literal">Memory.c</tt>).</li>
891 <li>CPU/core frequency: For reproducible results the frequency of all cores
892 should be fixed.</li>
893 <li>NUMA placement policy: The benchmark assumes a first touch policy, which
894 means the memory will be placed at the NUMA domain the touching core is
895 associated with. If a different policy is in place or the NUMA domain to be
896 used is already full memory might be allocated in a remote domain. Accesses
897 to remote domains typically have a higher latency and lower bandwidth.</li>
898 <li>System load: interference with other application, espcially on desktop
899 systems should be avoided.</li>
900 <li>Padding: For SoA based kernels the number of (fluid) nodes is automatically
901 adjusted so that no cache or TLB thrashing should occur. The parameters are
902 optimized for current Intel based systems. For more details look into the
903 padding section.</li>
904 <li>CPU dispatcher function: the compiler might add different versions of a
905 function for different ISA extensions. Make sure the code you might think is
906 executed is actually the code which is executed.</li>
908 <div class="section" id="padding">
909 <h2><a class="toc-backref" href="#id13">3.1 Padding</a></h2>
910 <p>With correct padding cache and TLB thrashing can be avoided. Therefore the
911 number of (fluid) nodes used in the data layout is artificially increased.</p>
912 <p>Currently automatic padding is active for kernels which support it. It can be
913 controlled via the kernel parameter (i.e. parameter after the <tt class="docutils literal"><span class="pre">--</span></tt>)
914 <tt class="docutils literal"><span class="pre">-pad</span></tt>. Supported values are <tt class="docutils literal">auto</tt> (default), <tt class="docutils literal">no</tt> (to disable padding),
915 or a manual padding.</p>
916 <p>Automatic padding tries to avoid cache and TLB thrashing and pads for a 32
917 entry (huge pages) TLB with 8 sets and a 512 set (L2) cache. This reflects the
918 parameters of current Intel based processors.</p>
919 <p>Manual padding is done via a padding string and has the format
920 <tt class="docutils literal"><span class="pre">mod_1+offset_1(,mod_n+offset_n)</span></tt>, which specifies numbers of bytes.
921 SoA data layouts can exhibit TLB thrashing. Therefore we want to distribute the
922 19 pages with one lattice (36 with two lattices) we are concurrently accessing
923 over as much sets in the TLB as possible.
924 This is controlled by the distance between the accessed pages, which is the
925 number of (fluid) nodes in between them and can be adjusted by adding further
927 We want the distance d (in bytes) between two accessed pages to be e.g.
928 <strong>d % (PAGE_SIZE * TLB_SETS) = PAGE_SIZE</strong>.
929 This would distribute the pages evenly over the sets. Hereby <strong>PAGE_SIZE * TLB_SETS</strong>
930 would be our <tt class="docutils literal">mod_1</tt> and <strong>PAGE_SIZE</strong> (after the =) our <tt class="docutils literal">offset_1</tt>.
931 Measurements show that with only a quarter of half of a page size as offset
932 higher performance is achieved, which is done by automatic padding.
933 On top of this padding more paddings can be added. They are just added to the
934 padding string and are separated by commas.</p>
935 <p>A zero modulus in the padding string has a special meaning. Here the
936 corresponding offset is just added to the number of nodes. A padding string
937 like <tt class="docutils literal"><span class="pre">-pad</span> 0+16</tt> would at a static padding of two nodes (one node = 8 b).</p>
940 <div class="section" id="geometries">
941 <h1><a class="toc-backref" href="#id14">4 Geometries</a></h1>
942 <p>TODO: supported geometries: channel, pipe, blocks</p>
944 <div class="section" id="results">
945 <h1><a class="toc-backref" href="#id15">5 Results</a></h1>
948 <div class="section" id="licence">
949 <h1><a class="toc-backref" href="#id16">6 Licence</a></h1>
950 <p>The Lattice Boltzmann Benchmark Kernels are licensed under GPLv3.</p>
952 <div class="section" id="acknowledgements">
953 <h1><a class="toc-backref" href="#id17">7 Acknowledgements</a></h1>
954 <p>This work was funded by BMBF, grant no. 01IH15003A (project SKAMPY).</p>
955 <p>This work was funded by KONWHIR project OMI4PAPS.</p>
956 <p>Document was generated at 2017-11-02 15:33.</p>