<?xml version="1.0" encoding="UTF-8"?>
<rss xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:rdf="http://www.w3.org/1999/02/22-rdf-syntax-ns#" xmlns:taxo="http://purl.org/rss/1.0/modules/taxonomy/" version="2.0">
  <channel>
    <title>topic MPI Broadcast lossless? in Intel® MPI Library</title>
    <link>https://community.intel.com/t5/Intel-MPI-Library/MPI-Broadcast-lossless/m-p/850651#M1462</link>
    <description>All collective operations (at the final stage) are implemented as pt2pt communication using different algorithms. To implement multi-cast it should be supported in hardware.&lt;BR /&gt;&lt;BR /&gt;There are 2 options you could try:&lt;BR /&gt;1. Play with different algorithms using I_MPI_ADJUST_BCAST environment variable - see Reference Manual.&lt;BR /&gt;2. OFA Fabric in the Intel MPI Library supports multi-rail feature. Set &lt;META http-equiv="Content-Type" content="text/html; charset=utf-8" /&gt;&lt;META name="ProgId" content="PowerPoint.Slide" /&gt;&lt;META name="Generator" content="Microsoft PowerPoint 12" /&gt;I_MPI_FABRICS=shm:ofa, I_MPI_OFA_NUM_ADAPTERS=&lt;N&gt; e.g. 2
(1 by default), I_MPI_OFA_NUM_PORTS=&lt;N&gt;. If your nodes have more than 1 interconnect (or multi-port interconnects) you can try this feature.&lt;BR /&gt;&lt;BR /&gt;Regards!&lt;BR /&gt; Dmitry&lt;BR /&gt;&lt;BR /&gt;&lt;/N&gt;&lt;/N&gt;</description>
    <pubDate>Sat, 17 Apr 2010 13:32:09 GMT</pubDate>
    <dc:creator>Dmitry_K_Intel2</dc:creator>
    <dc:date>2010-04-17T13:32:09Z</dc:date>
    <item>
      <title>MPI Broadcast lossless?</title>
      <link>https://community.intel.com/t5/Intel-MPI-Library/MPI-Broadcast-lossless/m-p/850646#M1457</link>
      <description>Hi,&lt;BR /&gt;&lt;BR /&gt; I will like to know if broadcast feature in MPI is 100%lossless. Do we need to handle cases where it is not lossless.&lt;BR /&gt;&lt;BR /&gt;Thanks.</description>
      <pubDate>Thu, 08 Apr 2010 13:16:47 GMT</pubDate>
      <guid>https://community.intel.com/t5/Intel-MPI-Library/MPI-Broadcast-lossless/m-p/850646#M1457</guid>
      <dc:creator>jimmy82</dc:creator>
      <dc:date>2010-04-08T13:16:47Z</dc:date>
    </item>
    <item>
      <title>MPI Broadcast lossless?</title>
      <link>https://community.intel.com/t5/Intel-MPI-Library/MPI-Broadcast-lossless/m-p/850647#M1458</link>
      <description>&lt;P&gt;Hi jimmy,&lt;/P&gt;&lt;P&gt;Are you referring to the new fault-tolerant behavior in Intel MPI Library 4.0? The FT support is limited and does not cover collectives. Feel free to check out the &lt;A target="_blank" href="http://software.intel.com/en-us/articles/intel-mpi-library-documentation/"&gt;Reference Manual&lt;/A&gt; for details on provided functionality.&lt;/P&gt;&lt;P&gt;Regards,&lt;BR /&gt;~Gergana&lt;/P&gt;</description>
      <pubDate>Fri, 09 Apr 2010 22:31:40 GMT</pubDate>
      <guid>https://community.intel.com/t5/Intel-MPI-Library/MPI-Broadcast-lossless/m-p/850647#M1458</guid>
      <dc:creator>Gergana_S_Intel</dc:creator>
      <dc:date>2010-04-09T22:31:40Z</dc:date>
    </item>
    <item>
      <title>MPI Broadcast lossless?</title>
      <link>https://community.intel.com/t5/Intel-MPI-Library/MPI-Broadcast-lossless/m-p/850648#M1459</link>
      <description>I am referring not referring to fault tolerance. I need to check when doing a broadcast, is it guarnteed that all receipents will receive the broadcast data.</description>
      <pubDate>Sun, 11 Apr 2010 17:19:51 GMT</pubDate>
      <guid>https://community.intel.com/t5/Intel-MPI-Library/MPI-Broadcast-lossless/m-p/850648#M1459</guid>
      <dc:creator>jimmy82</dc:creator>
      <dc:date>2010-04-11T17:19:51Z</dc:date>
    </item>
    <item>
      <title>MPI Broadcast lossless?</title>
      <link>https://community.intel.com/t5/Intel-MPI-Library/MPI-Broadcast-lossless/m-p/850649#M1460</link>
      <description>&lt;P&gt;Intel MPI Library works according to MPI standard: completion
of MPI_Bcast on the root only says that it complete sending, completion on
non-root guarantee that we got all data (without loses). There are no
additional confirmations to root.&lt;/P&gt;Regards!&lt;BR /&gt; Dmitry&lt;BR /&gt;</description>
      <pubDate>Fri, 16 Apr 2010 06:06:03 GMT</pubDate>
      <guid>https://community.intel.com/t5/Intel-MPI-Library/MPI-Broadcast-lossless/m-p/850649#M1460</guid>
      <dc:creator>Dmitry_K_Intel2</dc:creator>
      <dc:date>2010-04-16T06:06:03Z</dc:date>
    </item>
    <item>
      <title>MPI Broadcast lossless?</title>
      <link>https://community.intel.com/t5/Intel-MPI-Library/MPI-Broadcast-lossless/m-p/850650#M1461</link>
      <description>Is it possible to force MPI_Bcast to perform a multi-cast?&lt;BR /&gt;&lt;BR /&gt;I realise the time to broadcast 200MB of data increases signficant with increased nodes.&lt;BR /&gt;Say if I were to broadcast to 1 node, it takes &amp;lt; 2s to complete.&lt;BR /&gt;Broadcast to 2 nodes takes &amp;lt;6s to complete.&lt;BR /&gt;Broadcast to 3 nodes takes &amp;lt;10s to complete.&lt;BR /&gt;&lt;BR /&gt;Forcing a multicast will allow me to complete transmitting to all nodes in the shortest time.</description>
      <pubDate>Fri, 16 Apr 2010 08:58:27 GMT</pubDate>
      <guid>https://community.intel.com/t5/Intel-MPI-Library/MPI-Broadcast-lossless/m-p/850650#M1461</guid>
      <dc:creator>Jimmy821</dc:creator>
      <dc:date>2010-04-16T08:58:27Z</dc:date>
    </item>
    <item>
      <title>MPI Broadcast lossless?</title>
      <link>https://community.intel.com/t5/Intel-MPI-Library/MPI-Broadcast-lossless/m-p/850651#M1462</link>
      <description>All collective operations (at the final stage) are implemented as pt2pt communication using different algorithms. To implement multi-cast it should be supported in hardware.&lt;BR /&gt;&lt;BR /&gt;There are 2 options you could try:&lt;BR /&gt;1. Play with different algorithms using I_MPI_ADJUST_BCAST environment variable - see Reference Manual.&lt;BR /&gt;2. OFA Fabric in the Intel MPI Library supports multi-rail feature. Set &lt;META http-equiv="Content-Type" content="text/html; charset=utf-8" /&gt;&lt;META name="ProgId" content="PowerPoint.Slide" /&gt;&lt;META name="Generator" content="Microsoft PowerPoint 12" /&gt;I_MPI_FABRICS=shm:ofa, I_MPI_OFA_NUM_ADAPTERS=&lt;N&gt; e.g. 2
(1 by default), I_MPI_OFA_NUM_PORTS=&lt;N&gt;. If your nodes have more than 1 interconnect (or multi-port interconnects) you can try this feature.&lt;BR /&gt;&lt;BR /&gt;Regards!&lt;BR /&gt; Dmitry&lt;BR /&gt;&lt;BR /&gt;&lt;/N&gt;&lt;/N&gt;</description>
      <pubDate>Sat, 17 Apr 2010 13:32:09 GMT</pubDate>
      <guid>https://community.intel.com/t5/Intel-MPI-Library/MPI-Broadcast-lossless/m-p/850651#M1462</guid>
      <dc:creator>Dmitry_K_Intel2</dc:creator>
      <dc:date>2010-04-17T13:32:09Z</dc:date>
    </item>
    <item>
      <title>MPI Broadcast lossless?</title>
      <link>https://community.intel.com/t5/Intel-MPI-Library/MPI-Broadcast-lossless/m-p/850652#M1463</link>
      <description>How do you invoke these? I tried but was unsuccessful. Using a configuration file and invoked with mpiexec.exe .</description>
      <pubDate>Mon, 24 May 2010 15:49:51 GMT</pubDate>
      <guid>https://community.intel.com/t5/Intel-MPI-Library/MPI-Broadcast-lossless/m-p/850652#M1463</guid>
      <dc:creator>jimmy82</dc:creator>
      <dc:date>2010-05-24T15:49:51Z</dc:date>
    </item>
    <item>
      <title>MPI Broadcast lossless?</title>
      <link>https://community.intel.com/t5/Intel-MPI-Library/MPI-Broadcast-lossless/m-p/850653#M1464</link>
      <description>Jimmy,&lt;BR /&gt;&lt;BR /&gt;Are you working on Windows?&lt;BR /&gt;OFA module is not supported on Windows platform! Sorry. And do not expect it in the nearest future. It means that you cannot use multi-rail feature either.&lt;BR /&gt;&lt;BR /&gt;Regards!&lt;BR /&gt; Dmitry&lt;BR /&gt;</description>
      <pubDate>Tue, 25 May 2010 06:06:12 GMT</pubDate>
      <guid>https://community.intel.com/t5/Intel-MPI-Library/MPI-Broadcast-lossless/m-p/850653#M1464</guid>
      <dc:creator>Dmitry_K_Intel2</dc:creator>
      <dc:date>2010-05-25T06:06:12Z</dc:date>
    </item>
    <item>
      <title>MPI Broadcast lossless?</title>
      <link>https://community.intel.com/t5/Intel-MPI-Library/MPI-Broadcast-lossless/m-p/850654#M1465</link>
      <description>I am referring to the I_MPI_ADJUST_BCAST parameter. How do I set this parameter? Can I have some examples?&lt;BR /&gt;&lt;BR /&gt;Thanks.</description>
      <pubDate>Wed, 26 May 2010 05:54:14 GMT</pubDate>
      <guid>https://community.intel.com/t5/Intel-MPI-Library/MPI-Broadcast-lossless/m-p/850654#M1465</guid>
      <dc:creator>jimmy82</dc:creator>
      <dc:date>2010-05-26T05:54:14Z</dc:date>
    </item>
    <item>
      <title>MPI Broadcast lossless?</title>
      <link>https://community.intel.com/t5/Intel-MPI-Library/MPI-Broadcast-lossless/m-p/850655#M1466</link>
      <description>&amp;gt;I am referring to the I_MPI_ADJUST_BCAST parameter. How do I set this parameter? Can I have some examples?&lt;BR /&gt;&lt;BR /&gt;Yeah, sure:&lt;BR /&gt; -genv I_MPI_ADJUST_BCAST '1:4-16;2:17-128;3:129-4096;7:4097-4000000'&lt;BR /&gt;Means that alrorithm 1 (Binominal) will be used for message from 4 to 16 bytes long, algorithm 2 (Recoursive doubling) for messages from 17 to 128 bytes long, algorithm 3 (Ring) for messages from 129 to 4K bytes long, algorithm 7 (Shumilin's) for large messages.&lt;BR /&gt;&lt;BR /&gt;BTW: Intel MPI library doesn't support multi-cast communication.&lt;BR /&gt;&lt;BR /&gt;Regards!&lt;BR /&gt; Dmitry&lt;BR /&gt;</description>
      <pubDate>Wed, 26 May 2010 08:08:43 GMT</pubDate>
      <guid>https://community.intel.com/t5/Intel-MPI-Library/MPI-Broadcast-lossless/m-p/850655#M1466</guid>
      <dc:creator>Dmitry_K_Intel2</dc:creator>
      <dc:date>2010-05-26T08:08:43Z</dc:date>
    </item>
  </channel>
</rss>

