<?xml version="1.0" encoding="UTF-8"?>
<!-- Generated on Fri, 10 Feb 2012 02:37:22 -0800 -->
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
  <channel>
    <atom:link href="http://software.intel.com/en-us/articles/parallelism-iq-challenge/feed/" rel="self" type="application/rss+xml" />
    <title>Intel Software Network Comments Feed</title>
    <link>http://software.intel.com/en-us/articles/parallelism-iq-challenge</link>
    <description></description>
    <language>en-us</language>
    <item>
      <title>By David Wohlford</title>
      <description><![CDATA[ 
Regarding question #5, Amdahl's Law for speedup is typically expressed as T_old/T_new.  For discussions of serial vs. parallel systems it is typically expressed as T_serial/T_parallel, with the provision that the serial version is the best.  

In other words, you're making the quad-core processors look pretty bad if the answer is (c).
 ]]></description>
      <link>http://software.intel.com/en-us/articles/parallelism-iq-challenge/#comment-21</link>
      <pubDate>Wed, 25 Jul 2007 14:48:29 -0700</pubDate>
      <guid isPermaLink="true">http://software.intel.com/en-us/articles/parallelism-iq-challenge/#comment-21</guid>
    </item>
    <item>
      <title>By James Reinders</title>
      <description><![CDATA[ 
David - precisely right.  The right answer should have the terms flipped on (c): &quot;T(1)/T(4), but T(1) must be for best serial version of the application&quot;.  We used this quiz for conversation at our OSCON booth - and this error on the final copy for the quiz stirred up more conversation.

A serial (single core) running a program in 600 seconds, which sees time go to only 200 seconds on a quad-core... would definitely say that they saw a 3X speed-up (600/200).

Thank you for pointing this out!
 ]]></description>
      <link>http://software.intel.com/en-us/articles/parallelism-iq-challenge/#comment-22</link>
      <pubDate>Sat, 28 Jul 2007 07:20:11 -0700</pubDate>
      <guid isPermaLink="true">http://software.intel.com/en-us/articles/parallelism-iq-challenge/#comment-22</guid>
    </item>
    <item>
      <title>By Dmitriy V&amp;#39;jukov</title>
      <description><![CDATA[ Question 4.
If own pool is FIFO then thread accesses BOTH ends of it (head - to push, tail - to pop). So it&#39;s actually doesn&#39;t matter how other threads will access it - clashes will be.
This is why no framework uses this strange policy (option b). If own pool is FIFO, so it&#39;s better to use FIFO for stealing too, because it at least gives substantial degree of fairness.
 ]]></description>
      <link>http://software.intel.com/en-us/articles/parallelism-iq-challenge/#comment-8411</link>
      <pubDate>Mon, 20 Oct 2008 02:25:03 -0700</pubDate>
      <guid isPermaLink="true">http://software.intel.com/en-us/articles/parallelism-iq-challenge/#comment-8411</guid>
    </item>
  </channel></rss>
