/usr/share/doc/python-pyopencl-doc/html/algorithm.html is in python-pyopencl-doc 2016.1+git20161130-1.
This file is owned by root:root, with mode 0o644.
The actual contents of the file can be viewed below.
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 | <!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN"
"http://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd">
<html xmlns="http://www.w3.org/1999/xhtml">
<head>
<meta http-equiv="Content-Type" content="text/html; charset=utf-8" />
<title>Parallel Algorithms — PyOpenCL 2016.2 documentation</title>
<link rel="stylesheet" href="_static/alabaster.css" type="text/css" />
<link rel="stylesheet" href="_static/pygments.css" type="text/css" />
<script type="text/javascript">
var DOCUMENTATION_OPTIONS = {
URL_ROOT: './',
VERSION: '2016.2',
COLLAPSE_INDEX: false,
FILE_SUFFIX: '.html',
HAS_SOURCE: false
};
</script>
<script type="text/javascript" src="_static/jquery.js"></script>
<script type="text/javascript" src="_static/underscore.js"></script>
<script type="text/javascript" src="_static/doctools.js"></script>
<link rel="index" title="Index" href="genindex.html" />
<link rel="search" title="Search" href="search.html" />
<link rel="top" title="PyOpenCL 2016.2 documentation" href="index.html" />
<link rel="next" title="How-tos" href="howto.html" />
<link rel="prev" title="Multi-dimensional arrays" href="array.html" />
<link rel="stylesheet" href="_static/custom.css" type="text/css" />
<meta name="viewport" content="width=device-width, initial-scale=0.9, maximum-scale=0.9" />
</head>
<body role="document">
<div class="document">
<div class="documentwrapper">
<div class="bodywrapper">
<div class="body" role="main">
<div class="section" id="parallel-algorithms">
<h1>Parallel Algorithms<a class="headerlink" href="#parallel-algorithms" title="Permalink to this headline">¶</a></h1>
<div class="section" id="module-pyopencl.elementwise">
<span id="element-wise-expression-evalution-map"></span><h2>Element-wise expression evalution (“map”)<a class="headerlink" href="#module-pyopencl.elementwise" title="Permalink to this headline">¶</a></h2>
<p>Evaluating involved expressions on <a class="reference internal" href="array.html#pyopencl.array.Array" title="pyopencl.array.Array"><code class="xref py py-class docutils literal"><span class="pre">pyopencl.array.Array</span></code></a> instances by
using overloaded operators can be somewhat inefficient, because a new temporary
is created for each intermediate result. The functionality in the module
<a class="reference internal" href="#module-pyopencl.elementwise" title="pyopencl.elementwise"><code class="xref py py-mod docutils literal"><span class="pre">pyopencl.elementwise</span></code></a> contains tools to help generate kernels that
evaluate multi-stage expressions on one or several operands in a single pass.</p>
<dl class="class">
<dt id="pyopencl.elementwise.ElementwiseKernel">
<em class="property">class </em><code class="descclassname">pyopencl.elementwise.</code><code class="descname">ElementwiseKernel</code><span class="sig-paren">(</span><em>context</em>, <em>arguments</em>, <em>operation</em>, <em>name="kernel"</em>, <em>preamble=""</em>, <em>options=[]</em><span class="sig-paren">)</span><a class="headerlink" href="#pyopencl.elementwise.ElementwiseKernel" title="Permalink to this definition">¶</a></dt>
<dd><p>A kernel that takes a number of scalar or vector <em>arguments</em> and performs
an <em>operation</em> specified as a snippet of C on these arguments.</p>
<table class="docutils field-list" frame="void" rules="none">
<col class="field-name" />
<col class="field-body" />
<tbody valign="top">
<tr class="field-odd field"><th class="field-name">Parameters:</th><td class="field-body"><ul class="first last simple">
<li><strong>arguments</strong> – a string formatted as a C argument list.</li>
<li><strong>operation</strong> – a snippet of C that carries out the desired ‘map’
operation. The current index is available as the variable <em>i</em>.
<em>operation</em> may contain the statement <code class="docutils literal"><span class="pre">PYOPENCL_ELWISE_CONTINUE</span></code>,
which will terminate processing for the current element.</li>
<li><strong>name</strong> – the function name as which the kernel is compiled</li>
<li><strong>options</strong> – passed unmodified to <a class="reference internal" href="runtime_program.html#pyopencl.Program.build" title="pyopencl.Program.build"><code class="xref py py-meth docutils literal"><span class="pre">pyopencl.Program.build()</span></code></a>.</li>
<li><strong>preamble</strong> – a piece of C source code that gets inserted outside of the
function context in the elementwise operation’s kernel source code.</li>
</ul>
</td>
</tr>
</tbody>
</table>
<div class="admonition warning">
<p class="first admonition-title">Warning</p>
<p class="last">Using a <cite>return</cite> statement in <em>operation</em> will lead to
incorrect results, as some elements may never get processed. Use
<code class="docutils literal"><span class="pre">PYOPENCL_ELWISE_CONTINUE</span></code> instead.</p>
</div>
<div class="versionchanged">
<p><span class="versionmodified">Changed in version 2013.1: </span>Added <code class="docutils literal"><span class="pre">PYOPENCL_ELWISE_CONTINUE</span></code>.</p>
</div>
<dl class="method">
<dt id="pyopencl.elementwise.ElementwiseKernel.__call__">
<code class="descname">__call__</code><span class="sig-paren">(</span><em>*args</em>, <em>wait_for=None</em><span class="sig-paren">)</span><a class="headerlink" href="#pyopencl.elementwise.ElementwiseKernel.__call__" title="Permalink to this definition">¶</a></dt>
<dd><p>Invoke the generated scalar kernel. The arguments may either be scalars or
<code class="xref py py-class docutils literal"><span class="pre">GPUArray</span></code> instances.</p>
<p>Returns a new <a class="reference internal" href="runtime_queue.html#pyopencl.Event" title="pyopencl.Event"><code class="xref py py-class docutils literal"><span class="pre">pyopencl.Event</span></code></a>. <em>wait_for</em>
may either be <em>None</em> or a list of <a class="reference internal" href="runtime_queue.html#pyopencl.Event" title="pyopencl.Event"><code class="xref py py-class docutils literal"><span class="pre">pyopencl.Event</span></code></a> instances for
whose completion this command waits before starting exeuction.</p>
</dd></dl>
</dd></dl>
<p>Here’s a usage example:</p>
<div class="highlight-default"><div class="highlight"><pre><span></span><span class="o">..</span> <span class="n">literalinclude</span><span class="p">::</span> <span class="o">../</span><span class="n">examples</span><span class="o">/</span><span class="n">demo_elementwise</span><span class="o">.</span><span class="n">py</span>
</pre></div>
</div>
<p>(You can find this example as
<a class="reference download internal" href="_downloads/demo_elementwise.py" download=""><code class="xref download docutils literal"><span class="pre">examples/demo_elementwise.py</span></code></a>
in the PyOpenCL distribution.)</p>
</div>
<div class="section" id="module-pyopencl.reduction">
<span id="sums-and-counts-reduce"></span><span id="custom-reductions"></span><h2>Sums and counts (“reduce”)<a class="headerlink" href="#module-pyopencl.reduction" title="Permalink to this headline">¶</a></h2>
<dl class="class">
<dt id="pyopencl.reduction.ReductionKernel">
<em class="property">class </em><code class="descclassname">pyopencl.reduction.</code><code class="descname">ReductionKernel</code><span class="sig-paren">(</span><em>ctx</em>, <em>dtype_out</em>, <em>neutral</em>, <em>reduce_expr</em>, <em>map_expr=None</em>, <em>arguments=None</em>, <em>name="reduce_kernel"</em>, <em>options=[]</em>, <em>preamble=""</em><span class="sig-paren">)</span><a class="headerlink" href="#pyopencl.reduction.ReductionKernel" title="Permalink to this definition">¶</a></dt>
<dd><p>Generate a kernel that takes a number of scalar or vector <em>arguments</em>
(at least one vector argument), performs the <em>map_expr</em> on each entry of
the vector argument and then the <em>reduce_expr</em> on the outcome of that.
<em>neutral</em> serves as an initial value. <em>preamble</em> offers the possibility
to add preprocessor directives and other code (such as helper functions)
to be added before the actual reduction kernel code.</p>
<p>Vectors in <em>map_expr</em> should be indexed by the variable <em>i</em>. <em>reduce_expr</em>
uses the formal values “a” and “b” to indicate two operands of a binary
reduction operation. If you do not specify a <em>map_expr</em>, <code class="docutils literal"><span class="pre">in[i]</span></code> is
automatically assumed and treated as the only one input argument.</p>
<p><em>dtype_out</em> specifies the <a class="reference external" href="file:///usr/share/doc/python-numpy-doc/html/reference/generated/numpy.dtype.html#numpy.dtype" title="(in NumPy v1.11)"><code class="xref py py-class docutils literal"><span class="pre">numpy.dtype</span></code></a> in which the reduction is
performed and in which the result is returned. <em>neutral</em> is specified as
float or integer formatted as string. <em>reduce_expr</em> and <em>map_expr</em> are
specified as string formatted operations and <em>arguments</em> is specified as a
string formatted as a C argument list. <em>name</em> specifies the name as which
the kernel is compiled. <em>options</em> are passed unmodified to
<a class="reference internal" href="runtime_program.html#pyopencl.Program.build" title="pyopencl.Program.build"><code class="xref py py-meth docutils literal"><span class="pre">pyopencl.Program.build()</span></code></a>. <em>preamble</em> specifies a string of code that
is inserted before the actual kernels.</p>
<dl class="method">
<dt id="pyopencl.reduction.ReductionKernel.__call__">
<code class="descname">__call__</code><span class="sig-paren">(</span><em>*args</em>, <em>queue=None</em>, <em>wait_for=None</em>, <em>return_event=False</em>, <em>out=None</em><span class="sig-paren">)</span><a class="headerlink" href="#pyopencl.reduction.ReductionKernel.__call__" title="Permalink to this definition">¶</a></dt>
<dd><p><em>wait_for</em>
may either be <em>None</em> or a list of <a class="reference internal" href="runtime_queue.html#pyopencl.Event" title="pyopencl.Event"><code class="xref py py-class docutils literal"><span class="pre">pyopencl.Event</span></code></a> instances for
whose completion this command waits before starting exeuction.</p>
<p>With <em>out</em> the resulting single-entry <a class="reference internal" href="array.html#pyopencl.array.Array" title="pyopencl.array.Array"><code class="xref py py-class docutils literal"><span class="pre">pyopencl.array.Array</span></code></a> can
be specified. Because offsets are supported one can store results
anywhere (e.g. <code class="docutils literal"><span class="pre">out=a[3]</span></code>).</p>
<table class="docutils field-list" frame="void" rules="none">
<col class="field-name" />
<col class="field-body" />
<tbody valign="top">
<tr class="field-odd field"><th class="field-name">Returns:</th><td class="field-body">the resulting scalar as a single-entry <a class="reference internal" href="array.html#pyopencl.array.Array" title="pyopencl.array.Array"><code class="xref py py-class docutils literal"><span class="pre">pyopencl.array.Array</span></code></a>
if <em>return_event</em> is <em>False</em>, otherwise a tuple <code class="docutils literal"><span class="pre">(scalar_array,</span> <span class="pre">event)</span></code>.</td>
</tr>
</tbody>
</table>
<div class="admonition note">
<p class="first admonition-title">Note</p>
<p class="last">The returned <a class="reference internal" href="runtime_queue.html#pyopencl.Event" title="pyopencl.Event"><code class="xref py py-class docutils literal"><span class="pre">pyopencl.Event</span></code></a> corresponds only to part of the
execution of the reduction. It is not suitable for profiling.</p>
</div>
</dd></dl>
<div class="versionadded">
<p><span class="versionmodified">New in version 2011.1.</span></p>
</div>
<div class="versionchanged">
<p><span class="versionmodified">Changed in version 2014.2: </span>Added <em>out</em> parameter.</p>
</div>
</dd></dl>
<p>Here’s a usage example:</p>
<div class="highlight-default"><div class="highlight"><pre><span></span><span class="n">a</span> <span class="o">=</span> <span class="n">pyopencl</span><span class="o">.</span><span class="n">array</span><span class="o">.</span><span class="n">arange</span><span class="p">(</span><span class="n">queue</span><span class="p">,</span> <span class="mi">400</span><span class="p">,</span> <span class="n">dtype</span><span class="o">=</span><span class="n">numpy</span><span class="o">.</span><span class="n">float32</span><span class="p">)</span>
<span class="n">b</span> <span class="o">=</span> <span class="n">pyopencl</span><span class="o">.</span><span class="n">array</span><span class="o">.</span><span class="n">arange</span><span class="p">(</span><span class="n">queue</span><span class="p">,</span> <span class="mi">400</span><span class="p">,</span> <span class="n">dtype</span><span class="o">=</span><span class="n">numpy</span><span class="o">.</span><span class="n">float32</span><span class="p">)</span>
<span class="n">krnl</span> <span class="o">=</span> <span class="n">ReductionKernel</span><span class="p">(</span><span class="n">ctx</span><span class="p">,</span> <span class="n">numpy</span><span class="o">.</span><span class="n">float32</span><span class="p">,</span> <span class="n">neutral</span><span class="o">=</span><span class="s2">"0"</span><span class="p">,</span>
<span class="n">reduce_expr</span><span class="o">=</span><span class="s2">"a+b"</span><span class="p">,</span> <span class="n">map_expr</span><span class="o">=</span><span class="s2">"x[i]*y[i]"</span><span class="p">,</span>
<span class="n">arguments</span><span class="o">=</span><span class="s2">"__global float *x, __global float *y"</span><span class="p">)</span>
<span class="n">my_dot_prod</span> <span class="o">=</span> <span class="n">krnl</span><span class="p">(</span><span class="n">a</span><span class="p">,</span> <span class="n">b</span><span class="p">)</span><span class="o">.</span><span class="n">get</span><span class="p">()</span>
</pre></div>
</div>
</div>
<div class="section" id="module-pyopencl.scan">
<span id="prefix-sums-scan"></span><span id="custom-scan"></span><h2>Prefix Sums (“scan”)<a class="headerlink" href="#module-pyopencl.scan" title="Permalink to this headline">¶</a></h2>
<p>A prefix sum is a running sum of an array, as provided by
e.g. <code class="xref py py-mod docutils literal"><span class="pre">numpy.cumsum</span></code>:</p>
<div class="highlight-default"><div class="highlight"><pre><span></span><span class="gp">>>> </span><span class="kn">import</span> <span class="nn">numpy</span> <span class="k">as</span> <span class="nn">np</span>
<span class="gp">>>> </span><span class="n">a</span> <span class="o">=</span> <span class="p">[</span><span class="mi">1</span><span class="p">,</span><span class="mi">1</span><span class="p">,</span><span class="mi">1</span><span class="p">,</span><span class="mi">1</span><span class="p">,</span><span class="mi">1</span><span class="p">,</span><span class="mi">2</span><span class="p">,</span><span class="mi">2</span><span class="p">,</span><span class="mi">2</span><span class="p">,</span><span class="mi">2</span><span class="p">,</span><span class="mi">2</span><span class="p">]</span>
<span class="gp">>>> </span><span class="n">np</span><span class="o">.</span><span class="n">cumsum</span><span class="p">(</span><span class="n">a</span><span class="p">)</span>
<span class="go">array([ 1, 2, 3, 4, 5, 7, 9, 11, 13, 15])</span>
</pre></div>
</div>
<p>This is a very simple example of what a scan can do. It turns out that scans
are significantly more versatile. They are a basic building block of many
non-trivial parallel algorithms. Many of the operations enabled by scans seem
difficult to parallelize because of loop-carried dependencies.</p>
<div class="admonition seealso">
<p class="first admonition-title">See also</p>
<dl class="last docutils">
<dt><a class="reference external" href="http://citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.128.6230">Prefix sums and their applications</a>, by Guy Blelloch.</dt>
<dd>This article gives an overview of some surprising applications of scans.</dd>
<dt><a class="reference internal" href="#predefined-scans"><span class="std std-ref">Simple / Legacy Interface</span></a></dt>
<dd>These operations built into PyOpenCL are realized using <a class="reference internal" href="#pyopencl.scan.GenericScanKernel" title="pyopencl.scan.GenericScanKernel"><code class="xref py py-class docutils literal"><span class="pre">GenericScanKernel</span></code></a>.</dd>
</dl>
</div>
<div class="section" id="usage-example">
<h3>Usage Example<a class="headerlink" href="#usage-example" title="Permalink to this headline">¶</a></h3>
<p>This example illustrates the implementation of a simplified version of
<a class="reference internal" href="#pyopencl.algorithm.copy_if" title="pyopencl.algorithm.copy_if"><code class="xref py py-func docutils literal"><span class="pre">pyopencl.algorithm.copy_if()</span></code></a>,
which copies integers from an array into the (variable-size) output if they are
greater than 300:</p>
<div class="highlight-default"><div class="highlight"><pre><span></span><span class="n">knl</span> <span class="o">=</span> <span class="n">GenericScanKernel</span><span class="p">(</span>
<span class="n">ctx</span><span class="p">,</span> <span class="n">np</span><span class="o">.</span><span class="n">int32</span><span class="p">,</span>
<span class="n">arguments</span><span class="o">=</span><span class="s2">"__global int *ary, __global int *out"</span><span class="p">,</span>
<span class="n">input_expr</span><span class="o">=</span><span class="s2">"(ary[i] > 300) ? 1 : 0"</span><span class="p">,</span>
<span class="n">scan_expr</span><span class="o">=</span><span class="s2">"a+b"</span><span class="p">,</span> <span class="n">neutral</span><span class="o">=</span><span class="s2">"0"</span><span class="p">,</span>
<span class="n">output_statement</span><span class="o">=</span><span class="s2">"""</span>
<span class="s2"> if (prev_item != item) out[item-1] = ary[i];</span>
<span class="s2"> """</span><span class="p">)</span>
<span class="n">out</span> <span class="o">=</span> <span class="n">a</span><span class="o">.</span><span class="n">copy</span><span class="p">()</span>
<span class="n">knl</span><span class="p">(</span><span class="n">a</span><span class="p">,</span> <span class="n">out</span><span class="p">)</span>
<span class="n">a_host</span> <span class="o">=</span> <span class="n">a</span><span class="o">.</span><span class="n">get</span><span class="p">()</span>
<span class="n">out_host</span> <span class="o">=</span> <span class="n">a_host</span><span class="p">[</span><span class="n">a_host</span> <span class="o">></span> <span class="mi">300</span><span class="p">]</span>
<span class="k">assert</span> <span class="p">(</span><span class="n">out_host</span> <span class="o">==</span> <span class="n">out</span><span class="o">.</span><span class="n">get</span><span class="p">()[:</span><span class="nb">len</span><span class="p">(</span><span class="n">out_host</span><span class="p">)])</span><span class="o">.</span><span class="n">all</span><span class="p">()</span>
</pre></div>
</div>
<p>The value being scanned over is a number of flags indicating whether each array
element is greater than 300. These flags are computed by <em>input_expr</em>. The
prefix sum over this array gives a running count of array items greater than
300. The <em>output_statement</em> the compares <cite>prev_item</cite> (the previous item’s scan
result, i.e. index) to <cite>item</cite> (the current item’s scan result, i.e.
index). If they differ, i.e. if the predicate was satisfied at this
position, then the item is stored in the output at the computed index.</p>
<p>This example does not make use of the following advanced features also available
in PyOpenCL:</p>
<ul class="simple">
<li>Segmented scans</li>
<li>Access to the previous item in <em>input_expr</em> (e.g. for comparisons)
See the <a class="reference external" href="https://github.com/pyopencl/pyopencl/blob/master/pyopencl/scan.py#L1353">implementation</a> of <code class="xref py py-func docutils literal"><span class="pre">unique()</span></code> for an example.</li>
</ul>
</div>
<div class="section" id="making-custom-scan-kernels">
<h3>Making Custom Scan Kernels<a class="headerlink" href="#making-custom-scan-kernels" title="Permalink to this headline">¶</a></h3>
<div class="versionadded">
<p><span class="versionmodified">New in version 2013.1.</span></p>
</div>
<dl class="class">
<dt id="pyopencl.scan.GenericScanKernel">
<em class="property">class </em><code class="descclassname">pyopencl.scan.</code><code class="descname">GenericScanKernel</code><span class="sig-paren">(</span><em>ctx</em>, <em>dtype</em>, <em>arguments</em>, <em>input_expr</em>, <em>scan_expr</em>, <em>neutral</em>, <em>output_statement</em>, <em>is_segment_start_expr=None</em>, <em>input_fetch_exprs=[]</em>, <em>index_dtype=<class 'numpy.int32'></em>, <em>name_prefix='scan'</em>, <em>options=[]</em>, <em>preamble=''</em>, <em>devices=None</em><span class="sig-paren">)</span><a class="headerlink" href="#pyopencl.scan.GenericScanKernel" title="Permalink to this definition">¶</a></dt>
<dd><p>Generates and executes code that performs prefix sums (“scans”) on
arbitrary types, with many possible tweaks.</p>
<p>Usage example:</p>
<div class="highlight-default"><div class="highlight"><pre><span></span><span class="kn">from</span> <span class="nn">pyopencl.scan</span> <span class="k">import</span> <span class="n">GenericScanKernel</span>
<span class="n">knl</span> <span class="o">=</span> <span class="n">GenericScanKernel</span><span class="p">(</span>
<span class="n">context</span><span class="p">,</span> <span class="n">np</span><span class="o">.</span><span class="n">int32</span><span class="p">,</span>
<span class="n">arguments</span><span class="o">=</span><span class="s2">"__global int *ary"</span><span class="p">,</span>
<span class="n">input_expr</span><span class="o">=</span><span class="s2">"ary[i]"</span><span class="p">,</span>
<span class="n">scan_expr</span><span class="o">=</span><span class="s2">"a+b"</span><span class="p">,</span> <span class="n">neutral</span><span class="o">=</span><span class="s2">"0"</span><span class="p">,</span>
<span class="n">output_statement</span><span class="o">=</span><span class="s2">"ary[i+1] = item;"</span><span class="p">)</span>
<span class="n">a</span> <span class="o">=</span> <span class="n">cl</span><span class="o">.</span><span class="n">array</span><span class="o">.</span><span class="n">arange</span><span class="p">(</span><span class="n">queue</span><span class="p">,</span> <span class="mi">10000</span><span class="p">,</span> <span class="n">dtype</span><span class="o">=</span><span class="n">np</span><span class="o">.</span><span class="n">int32</span><span class="p">)</span>
<span class="n">scan_kernel</span><span class="p">(</span><span class="n">a</span><span class="p">,</span> <span class="n">queue</span><span class="o">=</span><span class="n">queue</span><span class="p">)</span>
</pre></div>
</div>
<table class="docutils field-list" frame="void" rules="none">
<col class="field-name" />
<col class="field-body" />
<tbody valign="top">
<tr class="field-odd field"><th class="field-name">Parameters:</th><td class="field-body"><ul class="first last simple">
<li><strong>ctx</strong> – a <a class="reference internal" href="runtime_platform.html#pyopencl.Context" title="pyopencl.Context"><code class="xref py py-class docutils literal"><span class="pre">pyopencl.Context</span></code></a> within which the code
for this scan kernel will be generated.</li>
<li><strong>dtype</strong> – the <a class="reference external" href="file:///usr/share/doc/python-numpy-doc/html/reference/generated/numpy.dtype.html#numpy.dtype" title="(in NumPy v1.11)"><code class="xref py py-class docutils literal"><span class="pre">numpy.dtype</span></code></a> with which the scan will
be performed. May be a structured type if that type was registered
through <a class="reference internal" href="array.html#pyopencl.tools.get_or_register_dtype" title="pyopencl.tools.get_or_register_dtype"><code class="xref py py-func docutils literal"><span class="pre">pyopencl.tools.get_or_register_dtype()</span></code></a>.</li>
<li><strong>arguments</strong> – A string of comma-separated C argument declarations.
If <em>arguments</em> is specified, then <em>input_expr</em> must also be
specified. All types used here must be known to PyOpenCL.
(see <a class="reference internal" href="array.html#pyopencl.tools.get_or_register_dtype" title="pyopencl.tools.get_or_register_dtype"><code class="xref py py-func docutils literal"><span class="pre">pyopencl.tools.get_or_register_dtype()</span></code></a>).</li>
<li><strong>scan_expr</strong> – <p>The associative, binary operation carrying out the scan,
represented as a C string. Its two arguments are available as <cite>a</cite>
and <cite>b</cite> when it is evaluated. <cite>b</cite> is guaranteed to be the
‘element being updated’, and <cite>a</cite> is the increment. Thus,
if some data is supposed to just propagate along without being
modified by the scan, it should live in <cite>b</cite>.</p>
<p>This expression may call functions given in the <em>preamble</em>.</p>
<p>Another value available to this expression is <cite>across_seg_boundary</cite>,
a C <cite>bool</cite> indicating whether this scan update is crossing a
segment boundary, as defined by <cite>is_segment_start_expr</cite>.
The scan routine does not implement segmentation
semantics on its own. It relies on <cite>scan_expr</cite> to do this.
This value is available (but always <cite>false</cite>) even for a
non-segmented scan.</p>
<div class="admonition note">
<p class="first admonition-title">Note</p>
<p class="last">In early pre-releases of the segmented scan,
segmentation semantics were implemented <em>without</em>
relying on <cite>scan_expr</cite>.</p>
</div>
</li>
<li><strong>input_expr</strong> – <p>A C expression, encoded as a string, resulting
in the values to which the scan is applied. This may be used
to apply a mapping to values stored in <em>arguments</em> before being
scanned. The result of this expression must match <em>dtype</em>.
The index intended to be mapped is available as <cite>i</cite> in this
expression. This expression may also use the variables defined
by <em>input_fetch_expr</em>.</p>
<p>This expression may also call functions given in the <em>preamble</em>.</p>
</li>
<li><strong>output_statement</strong> – <p>a C statement that writes
the output of the scan. It has access to the scan result as <cite>item</cite>,
the preceding scan result item as <cite>prev_item</cite>, and the current index
as <cite>i</cite>. <cite>prev_item</cite> in a segmented scan will be the neutral element
at a segment boundary, not the immediately preceding item.</p>
<p>Using <em>prev_item</em> in output statement has a small run-time cost.
<cite>prev_item</cite> enables the construction of an exclusive scan.</p>
<p>For non-segmented scans, <em>output_statement</em> may also reference
<cite>last_item</cite>, which evaluates to the scan result of the last
array entry.</p>
</li>
<li><strong>is_segment_start_expr</strong> – <p>A C expression, encoded as a string,
resulting in a C <cite>bool</cite> value that determines whether a new
scan segments starts at index <em>i</em>. If given, makes the scan a
segmented scan. Has access to the current index <cite>i</cite>, the result
of <em>input_expr</em> as a, and in addition may use <em>arguments</em> and
<em>input_fetch_expr</em> variables just like <em>input_expr</em>.</p>
<p>If it returns true, then previous sums will not spill over into the
item with index <em>i</em> or subsequent items.</p>
</li>
<li><strong>input_fetch_exprs</strong> – <p>a list of tuples <em>(NAME, ARG_NAME, OFFSET)</em>.
An entry here has the effect of doing the equivalent of the following
before input_expr:</p>
<div class="highlight-default"><div class="highlight"><pre><span></span><span class="n">ARG_NAME_TYPE</span> <span class="n">NAME</span> <span class="o">=</span> <span class="n">ARG_NAME</span><span class="p">[</span><span class="n">i</span><span class="o">+</span><span class="n">OFFSET</span><span class="p">];</span>
</pre></div>
</div>
<p><cite>OFFSET</cite> is allowed to be 0 or -1, and <cite>ARG_NAME_TYPE</cite> is the type
of <cite>ARG_NAME</cite>.</p>
</li>
<li><strong>preamble</strong> – A snippet of C that is inserted into the compiled kernel
before the actual kernel function. May be used for, e.g. type definitions
or include statements.</li>
</ul>
</td>
</tr>
</tbody>
</table>
<p>The first array in the argument list determines the size of the index
space over which the scan is carried out, and thus the values over
which the index <em>i</em> occurring in a number of code fragments in
arguments above will vary.</p>
<p>All code fragments further have access to N, the number of elements
being processed in the scan.</p>
<dl class="method">
<dt id="pyopencl.scan.GenericScanKernel.__call__">
<code class="descname">__call__</code><span class="sig-paren">(</span><em>*args</em>, <em>allocator=None</em>, <em>queue=None</em>, <em>size=None</em>, <em>wait_for=None</em><span class="sig-paren">)</span><a class="headerlink" href="#pyopencl.scan.GenericScanKernel.__call__" title="Permalink to this definition">¶</a></dt>
<dd><p><em>queue</em> and <em>allocator</em> default to the ones provided on the first
<a class="reference internal" href="array.html#pyopencl.array.Array" title="pyopencl.array.Array"><code class="xref py py-class docutils literal"><span class="pre">pyopencl.array.Array</span></code></a> in <em>args</em>. <em>size</em> may specify the
length of the scan to be carried out. If not given, this length
is inferred from the first array argument passed.</p>
<p>Returns a new <a class="reference internal" href="runtime_queue.html#pyopencl.Event" title="pyopencl.Event"><code class="xref py py-class docutils literal"><span class="pre">pyopencl.Event</span></code></a>. <em>wait_for</em>
may either be <em>None</em> or a list of <a class="reference internal" href="runtime_queue.html#pyopencl.Event" title="pyopencl.Event"><code class="xref py py-class docutils literal"><span class="pre">pyopencl.Event</span></code></a> instances for
whose completion this command waits before starting exeuction.</p>
<div class="admonition note">
<p class="first admonition-title">Note</p>
<p class="last">The returned <a class="reference internal" href="runtime_queue.html#pyopencl.Event" title="pyopencl.Event"><code class="xref py py-class docutils literal"><span class="pre">pyopencl.Event</span></code></a> corresponds only to part of the
execution of the scan. It is not suitable for profiling.</p>
</div>
</dd></dl>
</dd></dl>
<div class="section" id="debugging-aids">
<h4>Debugging aids<a class="headerlink" href="#debugging-aids" title="Permalink to this headline">¶</a></h4>
<dl class="class">
<dt id="pyopencl.scan.GenericDebugScanKernel">
<em class="property">class </em><code class="descclassname">pyopencl.scan.</code><code class="descname">GenericDebugScanKernel</code><a class="headerlink" href="#pyopencl.scan.GenericDebugScanKernel" title="Permalink to this definition">¶</a></dt>
<dd><p>Performs the same function and has the same interface as
<a class="reference internal" href="#pyopencl.scan.GenericScanKernel" title="pyopencl.scan.GenericScanKernel"><code class="xref py py-class docutils literal"><span class="pre">GenericScanKernel</span></code></a>, but uses a dead-simple, sequential scan. Works
best on CPU platforms, and helps isolate bugs in scans by removing the
potential for issues originating in parallel execution.</p>
</dd></dl>
</div>
</div>
<div class="section" id="simple-legacy-interface">
<span id="predefined-scans"></span><h3>Simple / Legacy Interface<a class="headerlink" href="#simple-legacy-interface" title="Permalink to this headline">¶</a></h3>
<dl class="class">
<dt id="pyopencl.scan.ExclusiveScanKernel">
<em class="property">class </em><code class="descclassname">pyopencl.scan.</code><code class="descname">ExclusiveScanKernel</code><span class="sig-paren">(</span><em>ctx</em>, <em>dtype</em>, <em>scan_expr</em>, <em>neutral</em>, <em>name_prefix="scan"</em>, <em>options=[]</em>, <em>preamble=""</em>, <em>devices=None</em><span class="sig-paren">)</span><a class="headerlink" href="#pyopencl.scan.ExclusiveScanKernel" title="Permalink to this definition">¶</a></dt>
<dd><p>Generates a kernel that can compute a <a class="reference external" href="https://secure.wikimedia.org/wikipedia/en/wiki/Prefix_sum">prefix sum</a>
using any associative operation given as <em>scan_expr</em>.
<em>scan_expr</em> uses the formal values “a” and “b” to indicate two operands of
an associative binary operation. <em>neutral</em> is the neutral element
of <em>scan_expr</em>, obeying <em>scan_expr(a, neutral) == a</em>.</p>
<p><em>dtype</em> specifies the type of the arrays being operated on.
<em>name_prefix</em> is used for kernel names to ensure recognizability
in profiles and logs. <em>options</em> is a list of compiler options to use
when building. <em>preamble</em> specifies a string of code that is
inserted before the actual kernels. <em>devices</em> may be used to restrict
the set of devices on which the kernel is meant to run. (defaults
to all devices in the context <em>ctx</em>.</p>
<dl class="method">
<dt id="pyopencl.scan.ExclusiveScanKernel.__call__">
<code class="descname">__call__</code><span class="sig-paren">(</span><em>self</em>, <em>input_ary</em>, <em>output_ary=None</em>, <em>allocator=None</em>, <em>queue=None</em><span class="sig-paren">)</span><a class="headerlink" href="#pyopencl.scan.ExclusiveScanKernel.__call__" title="Permalink to this definition">¶</a></dt>
<dd></dd></dl>
</dd></dl>
<dl class="class">
<dt id="pyopencl.scan.InclusiveScanKernel">
<em class="property">class </em><code class="descclassname">pyopencl.scan.</code><code class="descname">InclusiveScanKernel</code><span class="sig-paren">(</span><em>dtype</em>, <em>scan_expr</em>, <em>neutral=None</em>, <em>name_prefix="scan"</em>, <em>options=[]</em>, <em>preamble=""</em>, <em>devices=None</em><span class="sig-paren">)</span><a class="headerlink" href="#pyopencl.scan.InclusiveScanKernel" title="Permalink to this definition">¶</a></dt>
<dd><p>Works like <a class="reference internal" href="#pyopencl.scan.ExclusiveScanKernel" title="pyopencl.scan.ExclusiveScanKernel"><code class="xref py py-class docutils literal"><span class="pre">ExclusiveScanKernel</span></code></a>.</p>
<div class="versionchanged">
<p><span class="versionmodified">Changed in version 2013.1: </span><em>neutral</em> is now always required.</p>
</div>
</dd></dl>
<p>For the array <cite>[1,2,3]</cite>, inclusive scan results in <cite>[1,3,6]</cite>, and exclusive
scan results in <cite>[0,1,3]</cite>.</p>
<p>Here’s a usage example:</p>
<div class="highlight-default"><div class="highlight"><pre><span></span><span class="n">knl</span> <span class="o">=</span> <span class="n">InclusiveScanKernel</span><span class="p">(</span><span class="n">context</span><span class="p">,</span> <span class="n">np</span><span class="o">.</span><span class="n">int32</span><span class="p">,</span> <span class="s2">"a+b"</span><span class="p">)</span>
<span class="n">n</span> <span class="o">=</span> <span class="mi">2</span><span class="o">**</span><span class="mi">20</span><span class="o">-</span><span class="mi">2</span><span class="o">**</span><span class="mi">18</span><span class="o">+</span><span class="mi">5</span>
<span class="n">host_data</span> <span class="o">=</span> <span class="n">np</span><span class="o">.</span><span class="n">random</span><span class="o">.</span><span class="n">randint</span><span class="p">(</span><span class="mi">0</span><span class="p">,</span> <span class="mi">10</span><span class="p">,</span> <span class="n">n</span><span class="p">)</span><span class="o">.</span><span class="n">astype</span><span class="p">(</span><span class="n">np</span><span class="o">.</span><span class="n">int32</span><span class="p">)</span>
<span class="n">dev_data</span> <span class="o">=</span> <span class="n">cl_array</span><span class="o">.</span><span class="n">to_device</span><span class="p">(</span><span class="n">queue</span><span class="p">,</span> <span class="n">host_data</span><span class="p">)</span>
<span class="n">knl</span><span class="p">(</span><span class="n">dev_data</span><span class="p">)</span>
<span class="k">assert</span> <span class="p">(</span><span class="n">dev_data</span><span class="o">.</span><span class="n">get</span><span class="p">()</span> <span class="o">==</span> <span class="n">np</span><span class="o">.</span><span class="n">cumsum</span><span class="p">(</span><span class="n">host_data</span><span class="p">,</span> <span class="n">axis</span><span class="o">=</span><span class="mi">0</span><span class="p">))</span><span class="o">.</span><span class="n">all</span><span class="p">()</span>
</pre></div>
</div>
</div>
</div>
<div class="section" id="module-pyopencl.algorithm">
<span id="predicated-copies-partition-unique"></span><h2>Predicated copies (“partition”, “unique”, ...)<a class="headerlink" href="#module-pyopencl.algorithm" title="Permalink to this headline">¶</a></h2>
<dl class="function">
<dt id="pyopencl.algorithm.copy_if">
<code class="descclassname">pyopencl.algorithm.</code><code class="descname">copy_if</code><span class="sig-paren">(</span><em>ary</em>, <em>predicate</em>, <em>extra_args=[]</em>, <em>preamble=''</em>, <em>queue=None</em>, <em>wait_for=None</em><span class="sig-paren">)</span><a class="headerlink" href="#pyopencl.algorithm.copy_if" title="Permalink to this definition">¶</a></dt>
<dd><p>Copy the elements of <em>ary</em> satisfying <em>predicate</em> to an output array.</p>
<table class="docutils field-list" frame="void" rules="none">
<col class="field-name" />
<col class="field-body" />
<tbody valign="top">
<tr class="field-odd field"><th class="field-name">Parameters:</th><td class="field-body"><ul class="first simple">
<li><strong>predicate</strong> – a C expression evaluating to a <cite>bool</cite>, represented as a string.
The value to test is available as <cite>ary[i]</cite>, and if the expression evaluates
to <cite>true</cite>, then this value ends up in the output.</li>
<li><strong>extra_args</strong> – a list of tuples <em>(name, value)</em> specifying
extra arguments to pass to the scan procedure. For version 2013.1,
<em>value</em> must be a of a <a class="reference external" href="file:///usr/share/doc/python-numpy-doc/html/reference/index.html#module-numpy" title="(in NumPy v1.11)"><code class="xref py py-mod docutils literal"><span class="pre">numpy</span></code></a> sized scalar type. As of version 2013.2,
<em>value</em> may also be a <a class="reference internal" href="array.html#pyopencl.array.Array" title="pyopencl.array.Array"><code class="xref py py-class docutils literal"><span class="pre">pyopencl.array.Array</span></code></a>.</li>
<li><strong>preamble</strong> – A snippet of C that is inserted into the compiled kernel
before the actual kernel function. May be used for, e.g. type definitions
or include statements.</li>
<li><strong>wait_for</strong> – <em>wait_for</em>
may either be <em>None</em> or a list of <a class="reference internal" href="runtime_queue.html#pyopencl.Event" title="pyopencl.Event"><code class="xref py py-class docutils literal"><span class="pre">pyopencl.Event</span></code></a> instances for
whose completion this command waits before starting exeuction.</li>
</ul>
</td>
</tr>
<tr class="field-even field"><th class="field-name">Returns:</th><td class="field-body"><p class="first last">a tuple <em>(out, count, event)</em> where <em>out</em> is the output array, <em>count</em>
is an on-device scalar (fetch to host with <cite>count.get()</cite>) indicating
how many elements satisfied <em>predicate</em>, and <em>event</em> is a
<a class="reference internal" href="runtime_queue.html#pyopencl.Event" title="pyopencl.Event"><code class="xref py py-class docutils literal"><span class="pre">pyopencl.Event</span></code></a> for dependency management. <em>out</em> is allocated
to the same length as <em>ary</em>, but only the first <em>count</em> entries carry
meaning.</p>
</td>
</tr>
</tbody>
</table>
<div class="versionadded">
<p><span class="versionmodified">New in version 2013.1.</span></p>
</div>
</dd></dl>
<dl class="function">
<dt id="pyopencl.algorithm.remove_if">
<code class="descclassname">pyopencl.algorithm.</code><code class="descname">remove_if</code><span class="sig-paren">(</span><em>ary</em>, <em>predicate</em>, <em>extra_args=[]</em>, <em>preamble=''</em>, <em>queue=None</em>, <em>wait_for=None</em><span class="sig-paren">)</span><a class="headerlink" href="#pyopencl.algorithm.remove_if" title="Permalink to this definition">¶</a></dt>
<dd><p>Copy the elements of <em>ary</em> not satisfying <em>predicate</em> to an output array.</p>
<table class="docutils field-list" frame="void" rules="none">
<col class="field-name" />
<col class="field-body" />
<tbody valign="top">
<tr class="field-odd field"><th class="field-name">Parameters:</th><td class="field-body"><ul class="first simple">
<li><strong>predicate</strong> – a C expression evaluating to a <cite>bool</cite>, represented as a string.
The value to test is available as <cite>ary[i]</cite>, and if the expression evaluates
to <cite>false</cite>, then this value ends up in the output.</li>
<li><strong>extra_args</strong> – a list of tuples <em>(name, value)</em> specifying
extra arguments to pass to the scan procedure. For version 2013.1,
<em>value</em> must be a of a <a class="reference external" href="file:///usr/share/doc/python-numpy-doc/html/reference/index.html#module-numpy" title="(in NumPy v1.11)"><code class="xref py py-mod docutils literal"><span class="pre">numpy</span></code></a> sized scalar type. As of version 2013.2,
<em>value</em> may also be a <a class="reference internal" href="array.html#pyopencl.array.Array" title="pyopencl.array.Array"><code class="xref py py-class docutils literal"><span class="pre">pyopencl.array.Array</span></code></a>.</li>
<li><strong>preamble</strong> – A snippet of C that is inserted into the compiled kernel
before the actual kernel function. May be used for, e.g. type definitions
or include statements.</li>
<li><strong>wait_for</strong> – <em>wait_for</em>
may either be <em>None</em> or a list of <a class="reference internal" href="runtime_queue.html#pyopencl.Event" title="pyopencl.Event"><code class="xref py py-class docutils literal"><span class="pre">pyopencl.Event</span></code></a> instances for
whose completion this command waits before starting exeuction.</li>
</ul>
</td>
</tr>
<tr class="field-even field"><th class="field-name">Returns:</th><td class="field-body"><p class="first last">a tuple <em>(out, count, event)</em> where <em>out</em> is the output array, <em>count</em>
is an on-device scalar (fetch to host with <cite>count.get()</cite>) indicating
how many elements did not satisfy <em>predicate</em>, and <em>event</em> is a
<a class="reference internal" href="runtime_queue.html#pyopencl.Event" title="pyopencl.Event"><code class="xref py py-class docutils literal"><span class="pre">pyopencl.Event</span></code></a> for dependency management.</p>
</td>
</tr>
</tbody>
</table>
<div class="versionadded">
<p><span class="versionmodified">New in version 2013.1.</span></p>
</div>
</dd></dl>
<dl class="function">
<dt id="pyopencl.algorithm.partition">
<code class="descclassname">pyopencl.algorithm.</code><code class="descname">partition</code><span class="sig-paren">(</span><em>ary</em>, <em>predicate</em>, <em>extra_args=[]</em>, <em>preamble=''</em>, <em>queue=None</em>, <em>wait_for=None</em><span class="sig-paren">)</span><a class="headerlink" href="#pyopencl.algorithm.partition" title="Permalink to this definition">¶</a></dt>
<dd><p>Copy the elements of <em>ary</em> into one of two arrays depending on whether
they satisfy <em>predicate</em>.</p>
<table class="docutils field-list" frame="void" rules="none">
<col class="field-name" />
<col class="field-body" />
<tbody valign="top">
<tr class="field-odd field"><th class="field-name">Parameters:</th><td class="field-body"><ul class="first simple">
<li><strong>predicate</strong> – a C expression evaluating to a <cite>bool</cite>, represented as a string.
The value to test is available as <cite>ary[i]</cite>.</li>
<li><strong>extra_args</strong> – a list of tuples <em>(name, value)</em> specifying
extra arguments to pass to the scan procedure. For version 2013.1,
<em>value</em> must be a of a <a class="reference external" href="file:///usr/share/doc/python-numpy-doc/html/reference/index.html#module-numpy" title="(in NumPy v1.11)"><code class="xref py py-mod docutils literal"><span class="pre">numpy</span></code></a> sized scalar type. As of version 2013.2,
<em>value</em> may also be a <a class="reference internal" href="array.html#pyopencl.array.Array" title="pyopencl.array.Array"><code class="xref py py-class docutils literal"><span class="pre">pyopencl.array.Array</span></code></a>.</li>
<li><strong>preamble</strong> – A snippet of C that is inserted into the compiled kernel
before the actual kernel function. May be used for, e.g. type definitions
or include statements.</li>
<li><strong>wait_for</strong> – <em>wait_for</em>
may either be <em>None</em> or a list of <a class="reference internal" href="runtime_queue.html#pyopencl.Event" title="pyopencl.Event"><code class="xref py py-class docutils literal"><span class="pre">pyopencl.Event</span></code></a> instances for
whose completion this command waits before starting exeuction.</li>
</ul>
</td>
</tr>
<tr class="field-even field"><th class="field-name">Returns:</th><td class="field-body"><p class="first last">a tuple <em>(out_true, out_false, count, event)</em> where <em>count</em>
is an on-device scalar (fetch to host with <cite>count.get()</cite>) indicating
how many elements satisfied the predicate, and <em>event</em> is a
<a class="reference internal" href="runtime_queue.html#pyopencl.Event" title="pyopencl.Event"><code class="xref py py-class docutils literal"><span class="pre">pyopencl.Event</span></code></a> for dependency management.</p>
</td>
</tr>
</tbody>
</table>
<div class="versionadded">
<p><span class="versionmodified">New in version 2013.1.</span></p>
</div>
</dd></dl>
<dl class="function">
<dt id="pyopencl.algorithm.unique">
<code class="descclassname">pyopencl.algorithm.</code><code class="descname">unique</code><span class="sig-paren">(</span><em>ary</em>, <em>is_equal_expr='a == b'</em>, <em>extra_args=[]</em>, <em>preamble=''</em>, <em>queue=None</em>, <em>wait_for=None</em><span class="sig-paren">)</span><a class="headerlink" href="#pyopencl.algorithm.unique" title="Permalink to this definition">¶</a></dt>
<dd><p>Copy the elements of <em>ary</em> into the output if <em>is_equal_expr</em>, applied to the
array element and its predecessor, yields false.</p>
<p>Works like the UNIX command <strong class="program">uniq</strong>, with a potentially custom
comparison. This operation is often used on sorted sequences.</p>
<table class="docutils field-list" frame="void" rules="none">
<col class="field-name" />
<col class="field-body" />
<tbody valign="top">
<tr class="field-odd field"><th class="field-name">Parameters:</th><td class="field-body"><ul class="first simple">
<li><strong>is_equal_expr</strong> – a C expression evaluating to a <cite>bool</cite>,
represented as a string. The elements being compared are
available as <cite>a</cite> and <cite>b</cite>. If this expression yields <cite>false</cite>, the
two are considered distinct.</li>
<li><strong>extra_args</strong> – a list of tuples <em>(name, value)</em> specifying
extra arguments to pass to the scan procedure. For version 2013.1,
<em>value</em> must be a of a <a class="reference external" href="file:///usr/share/doc/python-numpy-doc/html/reference/index.html#module-numpy" title="(in NumPy v1.11)"><code class="xref py py-mod docutils literal"><span class="pre">numpy</span></code></a> sized scalar type. As of version 2013.2,
<em>value</em> may also be a <a class="reference internal" href="array.html#pyopencl.array.Array" title="pyopencl.array.Array"><code class="xref py py-class docutils literal"><span class="pre">pyopencl.array.Array</span></code></a>.</li>
<li><strong>preamble</strong> – A snippet of C that is inserted into the compiled kernel
before the actual kernel function. May be used for, e.g. type definitions
or include statements.</li>
<li><strong>wait_for</strong> – <em>wait_for</em>
may either be <em>None</em> or a list of <a class="reference internal" href="runtime_queue.html#pyopencl.Event" title="pyopencl.Event"><code class="xref py py-class docutils literal"><span class="pre">pyopencl.Event</span></code></a> instances for
whose completion this command waits before starting exeuction.</li>
</ul>
</td>
</tr>
<tr class="field-even field"><th class="field-name">Returns:</th><td class="field-body"><p class="first last">a tuple <em>(out, count, event)</em> where <em>out</em> is the output array, <em>count</em>
is an on-device scalar (fetch to host with <cite>count.get()</cite>) indicating
how many elements satisfied the predicate, and <em>event</em> is a
<a class="reference internal" href="runtime_queue.html#pyopencl.Event" title="pyopencl.Event"><code class="xref py py-class docutils literal"><span class="pre">pyopencl.Event</span></code></a> for dependency management.</p>
</td>
</tr>
</tbody>
</table>
<div class="versionadded">
<p><span class="versionmodified">New in version 2013.1.</span></p>
</div>
</dd></dl>
</div>
<div class="section" id="sorting-radix-sort">
<h2>Sorting (radix sort)<a class="headerlink" href="#sorting-radix-sort" title="Permalink to this headline">¶</a></h2>
<dl class="class">
<dt id="pyopencl.algorithm.RadixSort">
<em class="property">class </em><code class="descclassname">pyopencl.algorithm.</code><code class="descname">RadixSort</code><span class="sig-paren">(</span><em>context</em>, <em>arguments</em>, <em>key_expr</em>, <em>sort_arg_names</em>, <em>bits_at_a_time=2</em>, <em>index_dtype=<class 'numpy.int32'></em>, <em>key_dtype=<class 'numpy.uint32'></em>, <em>scan_kernel=<class 'pyopencl.scan.GenericScanKernel'></em>, <em>options=[]</em><span class="sig-paren">)</span><a class="headerlink" href="#pyopencl.algorithm.RadixSort" title="Permalink to this definition">¶</a></dt>
<dd><p>Provides a general <a class="reference external" href="https://en.wikipedia.org/wiki/Radix_sort">radix sort</a>
on the compute device.</p>
<div class="admonition seealso">
<p class="first admonition-title">See also</p>
<p class="last"><code class="xref py py-class docutils literal"><span class="pre">pyopencl.algorithm.BitonicSort</span></code></p>
</div>
<div class="versionadded">
<p><span class="versionmodified">New in version 2013.1.</span></p>
</div>
<table class="docutils field-list" frame="void" rules="none">
<col class="field-name" />
<col class="field-body" />
<tbody valign="top">
<tr class="field-odd field"><th class="field-name">Parameters:</th><td class="field-body"><ul class="first last simple">
<li><strong>arguments</strong> – A string of comma-separated C argument declarations.
If <em>arguments</em> is specified, then <em>input_expr</em> must also be
specified. All types used here must be known to PyOpenCL.
(see <a class="reference internal" href="array.html#pyopencl.tools.get_or_register_dtype" title="pyopencl.tools.get_or_register_dtype"><code class="xref py py-func docutils literal"><span class="pre">pyopencl.tools.get_or_register_dtype()</span></code></a>).</li>
<li><strong>key_expr</strong> – An integer-valued C expression returning the
key based on which the sort is performed. The array index
for which the key is to be computed is available as <cite>i</cite>.
The expression may refer to any of the <em>arguments</em>.</li>
<li><strong>sort_arg_names</strong> – A list of argument names whose corresponding
array arguments will be sorted according to <em>key_expr</em>.</li>
</ul>
</td>
</tr>
</tbody>
</table>
<dl class="method">
<dt id="pyopencl.algorithm.RadixSort.__call__">
<code class="descname">__call__</code><span class="sig-paren">(</span><em>*args</em>, <em>**kwargs</em><span class="sig-paren">)</span><a class="headerlink" href="#pyopencl.algorithm.RadixSort.__call__" title="Permalink to this definition">¶</a></dt>
<dd><p>Run the radix sort. In addition to <em>args</em> which must match the
<em>arguments</em> specification on the constructor, the following
keyword arguments are supported:</p>
<table class="docutils field-list" frame="void" rules="none">
<col class="field-name" />
<col class="field-body" />
<tbody valign="top">
<tr class="field-odd field"><th class="field-name">Parameters:</th><td class="field-body"><ul class="first simple">
<li><strong>key_bits</strong> – specify how many bits (starting from least-significant)
there are in the key.</li>
<li><strong>allocator</strong> – See the <em>allocator</em> argument of <a class="reference internal" href="array.html#pyopencl.array.empty" title="pyopencl.array.empty"><code class="xref py py-func docutils literal"><span class="pre">pyopencl.array.empty()</span></code></a>.</li>
<li><strong>queue</strong> – A <a class="reference internal" href="runtime_queue.html#pyopencl.CommandQueue" title="pyopencl.CommandQueue"><code class="xref py py-class docutils literal"><span class="pre">pyopencl.CommandQueue</span></code></a>, defaulting to the
one from the first argument array.</li>
<li><strong>wait_for</strong> – <em>wait_for</em>
may either be <em>None</em> or a list of <a class="reference internal" href="runtime_queue.html#pyopencl.Event" title="pyopencl.Event"><code class="xref py py-class docutils literal"><span class="pre">pyopencl.Event</span></code></a> instances for
whose completion this command waits before starting exeuction.</li>
</ul>
</td>
</tr>
<tr class="field-even field"><th class="field-name">Returns:</th><td class="field-body"><p class="first last">A tuple <code class="docutils literal"><span class="pre">(sorted,</span> <span class="pre">event)</span></code>. <em>sorted</em> consists of sorted
copies of the arrays named in <em>sorted_args</em>, in the order of that
list. <em>event</em> is a <a class="reference internal" href="runtime_queue.html#pyopencl.Event" title="pyopencl.Event"><code class="xref py py-class docutils literal"><span class="pre">pyopencl.Event</span></code></a> for dependency management.</p>
</td>
</tr>
</tbody>
</table>
</dd></dl>
</dd></dl>
</div>
<div class="section" id="building-many-variable-size-lists">
<h2>Building many variable-size lists<a class="headerlink" href="#building-many-variable-size-lists" title="Permalink to this headline">¶</a></h2>
<dl class="class">
<dt id="pyopencl.algorithm.ListOfListsBuilder">
<em class="property">class </em><code class="descclassname">pyopencl.algorithm.</code><code class="descname">ListOfListsBuilder</code><span class="sig-paren">(</span><em>context</em>, <em>list_names_and_dtypes</em>, <em>generate_template</em>, <em>arg_decls</em>, <em>count_sharing=None</em>, <em>devices=None</em>, <em>name_prefix='plb_build_list'</em>, <em>options=[]</em>, <em>preamble=''</em>, <em>debug=False</em>, <em>complex_kernel=False</em><span class="sig-paren">)</span><a class="headerlink" href="#pyopencl.algorithm.ListOfListsBuilder" title="Permalink to this definition">¶</a></dt>
<dd><p>Generates and executes code to produce a large number of variable-size
lists, simply.</p>
<div class="admonition note">
<p class="first admonition-title">Note</p>
<p class="last">This functionality is provided as a preview. Its interface
is subject to change until this notice is removed.</p>
</div>
<div class="versionadded">
<p><span class="versionmodified">New in version 2013.1.</span></p>
</div>
<p>Here’s a usage example:</p>
<div class="highlight-default"><div class="highlight"><pre><span></span><span class="kn">from</span> <span class="nn">pyopencl.algorithm</span> <span class="k">import</span> <span class="n">ListOfListsBuilder</span>
<span class="n">builder</span> <span class="o">=</span> <span class="n">ListOfListsBuilder</span><span class="p">(</span><span class="n">context</span><span class="p">,</span> <span class="p">[(</span><span class="s2">"mylist"</span><span class="p">,</span> <span class="n">np</span><span class="o">.</span><span class="n">int32</span><span class="p">)],</span> <span class="s2">"""</span>
<span class="s2"> void generate(LIST_ARG_DECL USER_ARG_DECL index_type i)</span>
<span class="s2"> {</span>
<span class="s2"> int count = i % 4;</span>
<span class="s2"> for (int j = 0; j < count; ++j)</span>
<span class="s2"> {</span>
<span class="s2"> APPEND_mylist(count);</span>
<span class="s2"> }</span>
<span class="s2"> }</span>
<span class="s2"> """</span><span class="p">,</span> <span class="n">arg_decls</span><span class="o">=</span><span class="p">[])</span>
<span class="n">result</span><span class="p">,</span> <span class="n">event</span> <span class="o">=</span> <span class="n">builder</span><span class="p">(</span><span class="n">queue</span><span class="p">,</span> <span class="mi">2000</span><span class="p">)</span>
<span class="n">inf</span> <span class="o">=</span> <span class="n">result</span><span class="p">[</span><span class="s2">"mylist"</span><span class="p">]</span>
<span class="k">assert</span> <span class="n">inf</span><span class="o">.</span><span class="n">count</span> <span class="o">==</span> <span class="mi">3000</span>
<span class="k">assert</span> <span class="p">(</span><span class="n">inf</span><span class="o">.</span><span class="n">list</span><span class="o">.</span><span class="n">get</span><span class="p">()[</span><span class="o">-</span><span class="mi">6</span><span class="p">:]</span> <span class="o">==</span> <span class="p">[</span><span class="mi">1</span><span class="p">,</span> <span class="mi">2</span><span class="p">,</span> <span class="mi">2</span><span class="p">,</span> <span class="mi">3</span><span class="p">,</span> <span class="mi">3</span><span class="p">,</span> <span class="mi">3</span><span class="p">])</span><span class="o">.</span><span class="n">all</span><span class="p">()</span>
</pre></div>
</div>
<p>The function <cite>generate</cite> above is called once for each “input object”.
Each input object can then generate zero or more list entries.
The number of these input objects is given to <a class="reference internal" href="#pyopencl.algorithm.ListOfListsBuilder.__call__" title="pyopencl.algorithm.ListOfListsBuilder.__call__"><code class="xref py py-meth docutils literal"><span class="pre">__call__()</span></code></a> as <em>n_objects</em>.
List entries are generated by calls to <cite>APPEND_<list name>(value)</cite>.
Multiple lists may be generated at once.</p>
<dl class="method">
<dt id="pyopencl.algorithm.ListOfListsBuilder.__init__">
<code class="descname">__init__</code><span class="sig-paren">(</span><em>context</em>, <em>list_names_and_dtypes</em>, <em>generate_template</em>, <em>arg_decls</em>, <em>count_sharing=None</em>, <em>devices=None</em>, <em>name_prefix='plb_build_list'</em>, <em>options=[]</em>, <em>preamble=''</em>, <em>debug=False</em>, <em>complex_kernel=False</em><span class="sig-paren">)</span><a class="headerlink" href="#pyopencl.algorithm.ListOfListsBuilder.__init__" title="Permalink to this definition">¶</a></dt>
<dd><table class="docutils field-list" frame="void" rules="none">
<col class="field-name" />
<col class="field-body" />
<tbody valign="top">
<tr class="field-odd field"><th class="field-name">Parameters:</th><td class="field-body"><ul class="first last simple">
<li><strong>context</strong> – A <a class="reference internal" href="runtime_platform.html#pyopencl.Context" title="pyopencl.Context"><code class="xref py py-class docutils literal"><span class="pre">pyopencl.Context</span></code></a>.</li>
<li><strong>list_names_and_dtypes</strong> – a list of <cite>(name, dtype)</cite> tuples
indicating the lists to be built.</li>
<li><strong>generate_template</strong> – a snippet of C as described below</li>
<li><strong>arg_decls</strong> – A string of comma-separated C argument declarations.</li>
<li><strong>count_sharing</strong> – A mapping consisting of <cite>(child, mother)</cite>
indicating that <cite>mother</cite> and <cite>child</cite> will always have the
same number of indices, and the <cite>APPEND</cite> to <cite>mother</cite>
will always happen <em>before</em> the <cite>APPEND</cite> to the child.</li>
<li><strong>name_prefix</strong> – the name prefix to use for the compiled kernels</li>
<li><strong>options</strong> – OpenCL compilation options for kernels using
<em>generate_template</em>.</li>
<li><strong>complex_kernel</strong> – If <cite>True</cite>, prevents vectorization on CPUs.</li>
</ul>
</td>
</tr>
</tbody>
</table>
<p><em>generate_template</em> may use the following C macros/identifiers:</p>
<ul class="simple">
<li><cite>index_type</cite>: expands to C identifier for the index type used
for the calculation</li>
<li><cite>USER_ARG_DECL</cite>: expands to the C declarator for <cite>arg_decls</cite></li>
<li><cite>USER_ARGS</cite>: a list of C argument values corresponding to
<cite>user_arg_decl</cite></li>
<li><cite>LIST_ARG_DECL</cite>: expands to a C argument list representing the
data for the output lists. These are escaped prefixed with
<cite>“plg_”</cite> so as to not interfere with user-provided names.</li>
<li><cite>LIST_ARGS</cite>: a list of C argument values corresponding to
<cite>LIST_ARG_DECL</cite></li>
<li><cite>APPEND_name(entry)</cite>: inserts <cite>entry</cite> into the list <cite>name</cite>.
<em>entry</em> must be a valid C expression of the correct type.</li>
</ul>
<p>All argument-list related macros have a trailing comma included
if they are non-empty.</p>
<p><em>generate_template</em> must supply a function:</p>
<div class="highlight-c"><div class="highlight"><pre><span></span><span class="kt">void</span> <span class="nf">generate</span><span class="p">(</span><span class="n">USER_ARG_DECL</span> <span class="n">LIST_ARG_DECL</span> <span class="n">index_type</span> <span class="n">i</span><span class="p">)</span>
<span class="p">{</span>
<span class="n">APPEND_mylist</span><span class="p">(</span><span class="mi">5</span><span class="p">);</span>
<span class="p">}</span>
</pre></div>
</div>
<p>Internally, the <cite>kernel_template</cite> is expanded (at least) twice. Once,
for a ‘counting’ stage where the size of all the lists is determined,
and a second time, for a ‘generation’ stage where the lists are
actually filled. A <cite>generate</cite> function that has side effects beyond
calling <cite>append</cite> is therefore ill-formed.</p>
</dd></dl>
<dl class="method">
<dt id="pyopencl.algorithm.ListOfListsBuilder.__call__">
<code class="descname">__call__</code><span class="sig-paren">(</span><em>queue</em>, <em>n_objects</em>, <em>*args</em>, <em>**kwargs</em><span class="sig-paren">)</span><a class="headerlink" href="#pyopencl.algorithm.ListOfListsBuilder.__call__" title="Permalink to this definition">¶</a></dt>
<dd><table class="docutils field-list" frame="void" rules="none">
<col class="field-name" />
<col class="field-body" />
<tbody valign="top">
<tr class="field-odd field"><th class="field-name">Parameters:</th><td class="field-body"><ul class="first simple">
<li><strong>args</strong> – arguments corresponding to arg_decls in the constructor.
<a class="reference internal" href="array.html#pyopencl.array.Array" title="pyopencl.array.Array"><code class="xref py py-class docutils literal"><span class="pre">pyopencl.array.Array</span></code></a> are not allowed directly and should
be passed as their <a class="reference internal" href="array.html#pyopencl.array.Array.data" title="pyopencl.array.Array.data"><code class="xref py py-attr docutils literal"><span class="pre">pyopencl.array.Array.data</span></code></a> attribute instead.</li>
<li><strong>allocator</strong> – optionally, the allocator to use to allocate new
arrays.</li>
<li><strong>omit_lists</strong> – An iterable of list names that should <em>not</em> be built
with this invocation. The kernel code may <em>not</em> call <code class="docutils literal"><span class="pre">APPEND_name</span></code>
for these omitted lists. If it does, undefined behavior will result.
The returned <em>lists</em> dictionary will not contain an entry for names
in <em>omit_lists</em>.</li>
<li><strong>wait_for</strong> – <em>wait_for</em>
may either be <em>None</em> or a list of <a class="reference internal" href="runtime_queue.html#pyopencl.Event" title="pyopencl.Event"><code class="xref py py-class docutils literal"><span class="pre">pyopencl.Event</span></code></a> instances for
whose completion this command waits before starting exeuction.</li>
</ul>
</td>
</tr>
<tr class="field-even field"><th class="field-name">Returns:</th><td class="field-body"><p class="first last"><p>a tuple <code class="docutils literal"><span class="pre">(lists,</span> <span class="pre">event)</span></code>, where
<em>lists</em> a mapping from (built) list names to objects which
have attributes</p>
<ul>
<li><p class="first"><code class="docutils literal"><span class="pre">count</span></code> for the total number of entries in all lists combined</p>
</li>
<li><p class="first"><code class="docutils literal"><span class="pre">lists</span></code> for the array containing all lists.</p>
</li>
<li><p class="first"><code class="docutils literal"><span class="pre">starts</span></code> for the array of starting indices in <cite>lists</cite>.
<cite>starts</cite> is built so that it has n+1 entries, so that
the <em>i</em>‘th entry is the start of the <em>i</em>‘th list, and the
<em>i</em>‘th entry is the index one past the <em>i</em>‘th list’s end,
even for the last list.</p>
<p>This implies that all lists are contiguous.</p>
<p><em>event</em> is a <a class="reference internal" href="runtime_queue.html#pyopencl.Event" title="pyopencl.Event"><code class="xref py py-class docutils literal"><span class="pre">pyopencl.Event</span></code></a> for dependency management.</p>
</li>
</ul>
</p>
</td>
</tr>
</tbody>
</table>
<div class="versionchanged">
<p><span class="versionmodified">Changed in version 2016.2: </span>Added omit_lists.</p>
</div>
</dd></dl>
<table class="docutils field-list" frame="void" rules="none">
<col class="field-name" />
<col class="field-body" />
<tbody valign="top">
<tr class="field-odd field"><th class="field-name">Parameters:</th><td class="field-body"><ul class="first last simple">
<li><strong>context</strong> – A <a class="reference internal" href="runtime_platform.html#pyopencl.Context" title="pyopencl.Context"><code class="xref py py-class docutils literal"><span class="pre">pyopencl.Context</span></code></a>.</li>
<li><strong>list_names_and_dtypes</strong> – a list of <cite>(name, dtype)</cite> tuples
indicating the lists to be built.</li>
<li><strong>generate_template</strong> – a snippet of C as described below</li>
<li><strong>arg_decls</strong> – A string of comma-separated C argument declarations.</li>
<li><strong>count_sharing</strong> – A mapping consisting of <cite>(child, mother)</cite>
indicating that <cite>mother</cite> and <cite>child</cite> will always have the
same number of indices, and the <cite>APPEND</cite> to <cite>mother</cite>
will always happen <em>before</em> the <cite>APPEND</cite> to the child.</li>
<li><strong>name_prefix</strong> – the name prefix to use for the compiled kernels</li>
<li><strong>options</strong> – OpenCL compilation options for kernels using
<em>generate_template</em>.</li>
<li><strong>complex_kernel</strong> – If <cite>True</cite>, prevents vectorization on CPUs.</li>
</ul>
</td>
</tr>
</tbody>
</table>
<p><em>generate_template</em> may use the following C macros/identifiers:</p>
<ul class="simple">
<li><cite>index_type</cite>: expands to C identifier for the index type used
for the calculation</li>
<li><cite>USER_ARG_DECL</cite>: expands to the C declarator for <cite>arg_decls</cite></li>
<li><cite>USER_ARGS</cite>: a list of C argument values corresponding to
<cite>user_arg_decl</cite></li>
<li><cite>LIST_ARG_DECL</cite>: expands to a C argument list representing the
data for the output lists. These are escaped prefixed with
<cite>“plg_”</cite> so as to not interfere with user-provided names.</li>
<li><cite>LIST_ARGS</cite>: a list of C argument values corresponding to
<cite>LIST_ARG_DECL</cite></li>
<li><cite>APPEND_name(entry)</cite>: inserts <cite>entry</cite> into the list <cite>name</cite>.
<em>entry</em> must be a valid C expression of the correct type.</li>
</ul>
<p>All argument-list related macros have a trailing comma included
if they are non-empty.</p>
<p><em>generate_template</em> must supply a function:</p>
<div class="highlight-c"><div class="highlight"><pre><span></span><span class="kt">void</span> <span class="nf">generate</span><span class="p">(</span><span class="n">USER_ARG_DECL</span> <span class="n">LIST_ARG_DECL</span> <span class="n">index_type</span> <span class="n">i</span><span class="p">)</span>
<span class="p">{</span>
<span class="n">APPEND_mylist</span><span class="p">(</span><span class="mi">5</span><span class="p">);</span>
<span class="p">}</span>
</pre></div>
</div>
<p>Internally, the <cite>kernel_template</cite> is expanded (at least) twice. Once,
for a ‘counting’ stage where the size of all the lists is determined,
and a second time, for a ‘generation’ stage where the lists are
actually filled. A <cite>generate</cite> function that has side effects beyond
calling <cite>append</cite> is therefore ill-formed.</p>
</dd></dl>
</div>
<div class="section" id="module-pyopencl.bitonic_sort">
<span id="bitonic-sort"></span><h2>Bitonic Sort<a class="headerlink" href="#module-pyopencl.bitonic_sort" title="Permalink to this headline">¶</a></h2>
<dl class="class">
<dt id="pyopencl.bitonic_sort.BitonicSort">
<em class="property">class </em><code class="descclassname">pyopencl.bitonic_sort.</code><code class="descname">BitonicSort</code><span class="sig-paren">(</span><em>context</em><span class="sig-paren">)</span><a class="headerlink" href="#pyopencl.bitonic_sort.BitonicSort" title="Permalink to this definition">¶</a></dt>
<dd><p>Sort an array (or one axis of one) using a sorting network.</p>
<p>Will only work if the axis of the array to be sorted has a length
that is a power of 2.</p>
<div class="versionadded">
<p><span class="versionmodified">New in version 2015.2.</span></p>
</div>
<div class="admonition seealso">
<p class="first admonition-title">See also</p>
<p class="last"><a class="reference internal" href="#pyopencl.algorithm.RadixSort" title="pyopencl.algorithm.RadixSort"><code class="xref py py-class docutils literal"><span class="pre">pyopencl.algorithm.RadixSort</span></code></a></p>
</div>
<dl class="method">
<dt id="pyopencl.bitonic_sort.BitonicSort.__call__">
<code class="descname">__call__</code><span class="sig-paren">(</span><em>arr</em>, <em>idx=None</em>, <em>queue=None</em>, <em>wait_for=None</em>, <em>axis=0</em><span class="sig-paren">)</span><a class="headerlink" href="#pyopencl.bitonic_sort.BitonicSort.__call__" title="Permalink to this definition">¶</a></dt>
<dd><table class="docutils field-list" frame="void" rules="none">
<col class="field-name" />
<col class="field-body" />
<tbody valign="top">
<tr class="field-odd field"><th class="field-name">Parameters:</th><td class="field-body"><ul class="first simple">
<li><strong>arr</strong> – the array to be sorted. Will be overwritten with the sorted array.</li>
<li><strong>idx</strong> – an array of indices to be tracked along with the sorting of <em>arr</em></li>
<li><strong>queue</strong> – a <a class="reference internal" href="runtime_queue.html#pyopencl.CommandQueue" title="pyopencl.CommandQueue"><code class="xref py py-class docutils literal"><span class="pre">pyopencl.CommandQueue</span></code></a>, defaults to the array’s queue
if None</li>
<li><strong>wait_for</strong> – a list of <a class="reference internal" href="runtime_queue.html#pyopencl.Event" title="pyopencl.Event"><code class="xref py py-class docutils literal"><span class="pre">pyopencl.Event</span></code></a> instances or None</li>
<li><strong>axis</strong> – the axis of the array by which to sort</li>
</ul>
</td>
</tr>
<tr class="field-even field"><th class="field-name">Returns:</th><td class="field-body"><p class="first last">a tuple (sorted_array, event)</p>
</td>
</tr>
</tbody>
</table>
</dd></dl>
</dd></dl>
</div>
</div>
</div>
</div>
</div>
<div class="sphinxsidebar" role="navigation" aria-label="main navigation">
<div class="sphinxsidebarwrapper">
<h1 class="logo"><a href="index.html">PyOpenCL</a></h1>
<h3>Navigation</h3>
<ul class="current">
<li class="toctree-l1"><a class="reference internal" href="runtime.html">OpenCL Runtime: Basics</a></li>
<li class="toctree-l1"><a class="reference internal" href="runtime_const.html">OpenCL Runtime: Constants</a></li>
<li class="toctree-l1"><a class="reference internal" href="runtime_platform.html">OpenCL Runtime: Platforms, Devices and Contexts</a></li>
<li class="toctree-l1"><a class="reference internal" href="runtime_queue.html">OpenCL Runtime: Command Queues and Events</a></li>
<li class="toctree-l1"><a class="reference internal" href="runtime_memory.html">OpenCL Runtime: Memory</a></li>
<li class="toctree-l1"><a class="reference internal" href="runtime_program.html">OpenCL Runtime: Programs and Kernels</a></li>
<li class="toctree-l1"><a class="reference internal" href="runtime_gl.html">OpenCL Runtime: OpenGL Interoperability</a></li>
<li class="toctree-l1"><a class="reference internal" href="array.html">Multi-dimensional arrays</a></li>
<li class="toctree-l1 current"><a class="current reference internal" href="#">Parallel Algorithms</a><ul>
<li class="toctree-l2"><a class="reference internal" href="#module-pyopencl.elementwise">Element-wise expression evalution (“map”)</a></li>
<li class="toctree-l2"><a class="reference internal" href="#module-pyopencl.reduction">Sums and counts (“reduce”)</a></li>
<li class="toctree-l2"><a class="reference internal" href="#module-pyopencl.scan">Prefix Sums (“scan”)</a></li>
<li class="toctree-l2"><a class="reference internal" href="#module-pyopencl.algorithm">Predicated copies (“partition”, “unique”, ...)</a></li>
<li class="toctree-l2"><a class="reference internal" href="#sorting-radix-sort">Sorting (radix sort)</a></li>
<li class="toctree-l2"><a class="reference internal" href="#building-many-variable-size-lists">Building many variable-size lists</a></li>
<li class="toctree-l2"><a class="reference internal" href="#module-pyopencl.bitonic_sort">Bitonic Sort</a></li>
</ul>
</li>
<li class="toctree-l1"><a class="reference internal" href="howto.html">How-tos</a></li>
<li class="toctree-l1"><a class="reference internal" href="tools.html">Built-in Utilities</a></li>
<li class="toctree-l1"><a class="reference internal" href="misc.html">Installation</a></li>
<li class="toctree-l1"><a class="reference internal" href="misc.html#tips">Tips</a></li>
<li class="toctree-l1"><a class="reference internal" href="misc.html#guidelines">Guidelines</a></li>
<li class="toctree-l1"><a class="reference internal" href="misc.html#user-visible-changes">User-visible Changes</a></li>
<li class="toctree-l1"><a class="reference internal" href="misc.html#license">License</a></li>
<li class="toctree-l1"><a class="reference internal" href="misc.html#frequently-asked-questions">Frequently Asked Questions</a></li>
<li class="toctree-l1"><a class="reference internal" href="misc.html#citing-pyopencl">Citing PyOpenCL</a></li>
<li class="toctree-l1"><a class="reference internal" href="misc.html#acknowledgments">Acknowledgments</a></li>
</ul>
<hr />
<ul>
<li class="toctree-l1"><a href="https://github.com/pyopencl/pyopencl">🚀 Github</a></li>
<li class="toctree-l1"><a href="https://pypi.python.org/pypi/pyopencl">💾 Download Releases</a></li>
</ul>
<div class="relations">
<h3>Related Topics</h3>
<ul>
<li><a href="index.html">Documentation overview</a><ul>
<li>Previous: <a href="array.html" title="previous chapter">Multi-dimensional arrays</a></li>
<li>Next: <a href="howto.html" title="next chapter">How-tos</a></li>
</ul></li>
</ul>
</div>
<div id="searchbox" style="display: none" role="search">
<h3>Quick search</h3>
<form class="search" action="search.html" method="get">
<div><input type="text" name="q" /></div>
<div><input type="submit" value="Go" /></div>
<input type="hidden" name="check_keywords" value="yes" />
<input type="hidden" name="area" value="default" />
</form>
</div>
<script type="text/javascript">$('#searchbox').show(0);</script>
</div>
</div>
<div class="clearer"></div>
</div>
<div class="footer">
©2016, Andreas Kloeckner.
|
Powered by <a href="http://sphinx-doc.org/">Sphinx 1.4.9</a>
& <a href="https://github.com/bitprophet/alabaster">Alabaster 0.7.8</a>
</div>
</body>
</html>
|