<?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 Re: Questions about HW performance conters on P4 in Intel® MPI Library</title>
    <link>https://community.intel.com/t5/Intel-MPI-Library/Questions-about-HW-performance-conters-on-P4/m-p/950824#M2916</link>
    <description>&lt;DIV&gt;Georg,&lt;/DIV&gt;
&lt;DIV&gt;&lt;/DIV&gt;
&lt;DIV&gt;Did you try posting this question to the Vtune forums?&lt;/DIV&gt;
&lt;DIV&gt;&lt;/DIV&gt;
&lt;DIV&gt;&lt;A href="http://softwareforums.intel.com/ids/board?board.id=14" target="_blank"&gt;http://softwareforums.intel.com/ids/board?board.id=14&lt;/A&gt;&lt;/DIV&gt;
&lt;DIV&gt;&lt;/DIV&gt;
&lt;DIV&gt;Thanks.&lt;/DIV&gt;
&lt;DIV&gt;Vikram&lt;/DIV&gt;</description>
    <pubDate>Sat, 10 Jul 2004 03:15:03 GMT</pubDate>
    <dc:creator>Vikram_C_Intel</dc:creator>
    <dc:date>2004-07-10T03:15:03Z</dc:date>
    <item>
      <title>Questions about HW performance conters on P4</title>
      <link>https://community.intel.com/t5/Intel-MPI-Library/Questions-about-HW-performance-conters-on-P4/m-p/950823#M2915</link>
      <description>Hi!&lt;BR /&gt;&lt;BR /&gt;We've had some problems in understanding performance counter results&lt;BR /&gt;on P4/Xeon (collected with VTune 2.0 on Linux). Maybe this is just a&lt;BR /&gt;series of misunderstandings with the documentation, but anyway:&lt;BR /&gt;&lt;BR /&gt;1) In the IA32 Architecture Optimization document it is said that P4's&lt;BR /&gt;hardware counter "2nd Level Cache Read Misses" has bugs that can cause&lt;BR /&gt;miscounting by a factor of two. Since the measurements for same code with&lt;BR /&gt;same data size delivers reproducable counting results with vtune, this&lt;BR /&gt;bug has to occur under specific circumstances. Is anything known about&lt;BR /&gt;under which circumstances this bug occurs? There are some algorithms that&lt;BR /&gt;seem to result in reliable counts, other algorithms are obviously&lt;BR /&gt;miscounted. It would be great if a correct result could be drawn out of the&lt;BR /&gt;measurements and some assumptions or estimations.&lt;BR /&gt;&lt;BR /&gt;2) If data is loaded that is not in 2nd Level cache, the cache loads two&lt;BR /&gt;cache lines from memory. Is that counted as one or two events&lt;BR /&gt;for 2nd level Cache Read Misses? And which counters count L2 cache &lt;BR /&gt;store misses?&lt;BR /&gt;&lt;BR /&gt;3) As "2nd Level Cache Load Misses Retired" counts the Loads from L2 Cache,&lt;BR /&gt;which caused a cache miss, and "2nd Level Cache Read Misses" counts the&lt;BR /&gt;memory load misses as seen by the bus queue (VTune Reference), can it be&lt;BR /&gt;assumed, that - including some error concerning instruction loads&lt;BR /&gt;a.s.o - the difference of both are a measure for 2nd Level Cache Write&lt;BR /&gt;Misses? If not, how else could write misses be determined?&lt;BR /&gt;&lt;BR /&gt;&lt;BR /&gt;4) In the P4 Architecture Optimization document it is noted that &lt;BR /&gt;P4's event "2nd Level Cache Load Misses Retired" 'is known to undercount&lt;BR /&gt;when loads are apart'. Could you explain why/when this occurs, &lt;BR /&gt;and give an estimation of the factor by which it is undercounting&lt;BR /&gt;for some specific code example?&lt;BR /&gt;&lt;BR /&gt;Can anyone help us?&lt;BR /&gt;&lt;BR /&gt;Bye,&lt;BR /&gt;Georg.</description>
      <pubDate>Thu, 01 Jul 2004 16:07:49 GMT</pubDate>
      <guid>https://community.intel.com/t5/Intel-MPI-Library/Questions-about-HW-performance-conters-on-P4/m-p/950823#M2915</guid>
      <dc:creator>schorscherl</dc:creator>
      <dc:date>2004-07-01T16:07:49Z</dc:date>
    </item>
    <item>
      <title>Re: Questions about HW performance conters on P4</title>
      <link>https://community.intel.com/t5/Intel-MPI-Library/Questions-about-HW-performance-conters-on-P4/m-p/950824#M2916</link>
      <description>&lt;DIV&gt;Georg,&lt;/DIV&gt;
&lt;DIV&gt;&lt;/DIV&gt;
&lt;DIV&gt;Did you try posting this question to the Vtune forums?&lt;/DIV&gt;
&lt;DIV&gt;&lt;/DIV&gt;
&lt;DIV&gt;&lt;A href="http://softwareforums.intel.com/ids/board?board.id=14" target="_blank"&gt;http://softwareforums.intel.com/ids/board?board.id=14&lt;/A&gt;&lt;/DIV&gt;
&lt;DIV&gt;&lt;/DIV&gt;
&lt;DIV&gt;Thanks.&lt;/DIV&gt;
&lt;DIV&gt;Vikram&lt;/DIV&gt;</description>
      <pubDate>Sat, 10 Jul 2004 03:15:03 GMT</pubDate>
      <guid>https://community.intel.com/t5/Intel-MPI-Library/Questions-about-HW-performance-conters-on-P4/m-p/950824#M2916</guid>
      <dc:creator>Vikram_C_Intel</dc:creator>
      <dc:date>2004-07-10T03:15:03Z</dc:date>
    </item>
  </channel>
</rss>

