<html>
  <head>
    <meta content="text/html; charset=UTF-8" http-equiv="Content-Type">
  </head>
  <body text="#000000" bgcolor="#FFFFFF">
    Hi Stefan,<br>
    <br>
    Nice change, looks really good, but I have a few small comments.<br>
    <br>
    In workgroup.hpp:<br>
     177   GangTaskDispatcher* dispatcher()const {<br>
    Add a space before const.<br>
    ---<br>
    <br>
    In workgroup.cpp:<br>
     122       // Limit the semaphore value to the number of workers.<br>
     123       _start_semaphore(new Semaphore()),<br>
     124       _end_semaphore(new Semaphore())<br>
    Seems like the limit has been lost. I would prefer to have it added,
    but if that's somehow problematic just remove the comment.<br>
    <br>
     140     // Wait for the last worker to signal the coordinator.<br>
     141     _end_semaphore->wait();<br>
     142 <br>
     143     // No workers are allowed to read the state variables after
    the coordinator has been signaled.<br>
     144     _task         = NULL;<br>
     145     _started      = 0;<br>
     146     _not_finished = 0;<br>
    Do we need to set _not_finished to 0 or could we instead assert that
    it already is 0 since that's when _end_semaphore should have been
    signaled. <br>
    ---<br>
    <br>
    I don't need to see a new webrev for the above changes. Reviewed.<br>
    <br>
    Thanks,<br>
    Stefan<br>
    <br>
    <div class="moz-cite-prefix">On 2015-08-13 19:50, Stefan Karlsson
      wrote:<br>
    </div>
    <blockquote cite="mid:55CCD8D5.1070509@oracle.com" type="cite">
      <meta content="text/html; charset=UTF-8" http-equiv="Content-Type">
      <div class="moz-cite-prefix">Hi all,<br>
        <br>
        Could I get a second review for the patch below (plus the
        suggested removal)?<br>
        <br>
        Thanks,<br>
        StefanK<br>
        <br>
        On 2015-07-02 19:58, Stefan Karlsson wrote:<br>
      </div>
      <blockquote cite="mid:55957BCD.9020604@oracle.com" type="cite">
        <meta content="text/html; charset=UTF-8"
          http-equiv="Content-Type">
        <div class="moz-cite-prefix">On 2015-07-02 19:43, Jon Masamitsu
          wrote:<br>
        </div>
        <blockquote cite="mid:5595784A.9050903@oracle.com" type="cite">
          <meta content="text/html; charset=UTF-8"
            http-equiv="Content-Type">
          <br>
          <br>
          <div class="moz-cite-prefix">On 6/29/2015 2:38 AM, Stefan
            Karlsson wrote:<br>
          </div>
          <blockquote cite="mid:5591122F.2000704@oracle.com" type="cite">
            <meta content="text/html; charset=UTF-8"
              http-equiv="Content-Type">
            Hi all,<br>
            <br>
            "8087322: Implement a Semaphore utility class" has now been
            pushed, so I've updated the patch to reflect the changes.<br>
            <br>
            <a moz-do-not-send="true" class="moz-txt-link-freetext"
              href="http://cr.openjdk.java.net/%7Estefank/8087324/webrev.01.delta">http://cr.openjdk.java.net/~stefank/8087324/webrev.01.delta</a><br>
            <a moz-do-not-send="true" class="moz-txt-link-freetext"
              href="http://cr.openjdk.java.net/%7Estefank/8087324/webrev.01">http://cr.openjdk.java.net/~stefank/8087324/webrev.01</a><br>
          </blockquote>
          <a moz-do-not-send="true" class="moz-txt-link-freetext"
href="http://cr.openjdk.java.net/%7Estefank/8087324/webrev.01/src/share/vm/gc/shared/workgroup.hpp.frames.html">http://cr.openjdk.java.net/~stefank/8087324/webrev.01/src/share/vm/gc/shared/workgroup.hpp.frames.html</a><br>
          <br>
          Are these used?<br>
          <br>
          <pre><span class="changed"> 194   void print_worker_started_task(AbstractGangTask* task, uint worker_id);</span>
<span class="changed"> 195   void print_worker_finished_task(AbstractGangTask* task, uint worker_id);
</span></pre>
        </blockquote>
        <br>
        No. I'll remove them.<br>
        <br>
        <blockquote cite="mid:5595784A.9050903@oracle.com" type="cite">
          <pre><span class="changed">

Rest looks good.  Just one question (just for my education).
 </span></pre>
          <a moz-do-not-send="true" class="moz-txt-link-freetext"
href="http://cr.openjdk.java.net/%7Estefank/8087324/webrev.01/src/share/vm/gc/shared/workgroup.cpp.frames.html">http://cr.openjdk.java.net/~stefank/8087324/webrev.01/src/share/vm/gc/shared/workgroup.cpp.frames.html</a><br>
          <br>
          <pre><span class="changed"> 336 void GangWorker::loop() {</span>
<span class="changed"> 337   while (true) {</span>
<span class="changed"> 338     WorkData data = wait_for_task();</span>
<span class="changed"> 339 </span>
<span class="changed"> 340     run_task(data);</span>
<span class="changed"> 341 </span>
<span class="changed"> 342     signal_task_done();</span>
 343   }
 344 }

Does this allow the same thread to execute more than 1
task ("data" here)?  Meaning if 2 threads are  requested but
1 thread is not scheduled to a cpu, will the other thread
do both chunks of work?</pre>
        </blockquote>
        <br>
        Yes, that's correct.<br>
        <br>
        Thanks for reviewing!<br>
        StefanK<br>
        <br>
        <blockquote cite="mid:5595784A.9050903@oracle.com" type="cite">
          <pre>Jon


</pre>
          <blockquote cite="mid:5591122F.2000704@oracle.com" type="cite">
            <br>
            - The IMPLEMENTS_SEMAPHORE_CLASS define was removed, since
            all platforms need to provide a Semaphore implementation.<br>
            <br>
            - Removed the need to pass down "max number of workers" to
            the Semaphore constructor.<br>
            <br>
            - Updated semaphore.hpp include path<br>
            <br>
            Thanks,<br>
            StefanK<br>
            <br>
            <br>
            <div class="moz-cite-prefix">On 2015-06-12 16:52, Stefan
              Karlsson wrote:<br>
            </div>
            <blockquote cite="mid:557AF21B.2090102@oracle.com"
              type="cite">
              <meta http-equiv="content-type" content="text/html;
                charset=UTF-8">
              Hi all,<br>
              <br>
              The current implementation to distribute tasks to GC
              worker threads often cause long latencies (multiple
              milliseconds) when the threads are started and stopped. <br>
              <br>
              The main reason is that the worker threads have to fight
              over the Monitor lock when they are woken up from the call
              to Monitor::wait. Another reason is that all worker
              threads call notify_all when they finish a task and there
              wakes all all sleeping worker threads, which will yet
              again force the worker threads to fight over the lock. <br>
              <br>
              I propose that we use semaphores instead, so that the
              worker threads don't have to fight over a lock when they
              are woken up.<br>
              <br>
              <br>
              The patches build upon the following patch which
              introduces a Semaphore utility class. This patch will sent
              out for review on the hotspot-dev, since it affects non-GC
              parts of the code:<br>
               <a moz-do-not-send="true" class="moz-txt-link-freetext"
                href="http://cr.openjdk.java.net/%7Estefank/8087322/webrev.00/">http://cr.openjdk.java.net/~stefank/8087322/webrev.00/</a><br>
               <a moz-do-not-send="true" class="moz-txt-link-freetext"
                href="https://bugs.openjdk.java.net/browse/JDK-8087322">https://bugs.openjdk.java.net/browse/JDK-8087322</a><br>
              <br>
              <br>
              The first patch that I would like to get reviewed is:<br>
               <a moz-do-not-send="true" class="moz-txt-link-freetext"
                href="http://cr.openjdk.java.net/%7Estefank/8087323/webrev.00/">http://cr.openjdk.java.net/~stefank/8087323/webrev.00/</a><br>
               <a moz-do-not-send="true" class="moz-txt-link-freetext"
                href="https://bugs.openjdk.java.net/browse/JDK-8087323">https://bugs.openjdk.java.net/browse/JDK-8087323</a>
              - Unify and split the work gang classes <br>
              <br>
              It prepares for JDK-8087324, by separating the generic
              WorkGang implementation from the more elaborate
              YieldingFlexibleWorkGang (CMS) implementation. By having
              this part as a separate patch, I hope it will be easier to
              review JDK-8087324. The patch changes the work gang
              inheritance from:<br>
              <br>
              AbstractWorkGang<br>
               WorkGang<br>
                FlexibleWorkGang<br>
                 YieldingFlexibleWorkGang<br>
              <br>
              to:<br>
              <br>
              AbstractWorkGang<br>
               WorkGang<br>
               YieldingFlexibleWorkGang<br>
              <br>
              Parts of the FlexibleWorkGang and WorkGang code that is
              going to be used by both concrete work gang classes, has
              been moved into AbstractWorkGang. I've duplicated some
              code in WorkGang and YieldingFlexibleWorkGang, but that
              code will be removed from WorkGang in the following patch.<br>
              <br>
              <br>
              The second patch I'd like to get reviewed is:<br>
               <a moz-do-not-send="true" class="moz-txt-link-freetext"
                href="http://cr.openjdk.java.net/%7Estefank/8087324/webrev.00/">http://cr.openjdk.java.net/~stefank/8087324/webrev.00/</a><br>
               <a moz-do-not-send="true" class="moz-txt-link-freetext"
                href="https://bugs.openjdk.java.net/browse/JDK-8087324">https://bugs.openjdk.java.net/browse/JDK-8087324</a>
              - Use semaphores when starting and stopping GC task
              threads <br>
              <br>
              It first simplifies the way we distribute the tasks to the
              GC worker threads. For example, the coordinator thread
              dispatches a task to a specific number of workers, and
              then waits for all work to be completed. There's no risk
              that multiple tasks will be scheduled simultaneously, so
              there's no need for the sequences number that is used in
              the current implementation.<br>
              <br>
              The patch contains two task dispatch / thread
              synchronization implementations:<br>
              <br>
              The first implementation uses Monitors, similar to what we
              did before the patch, but with a slightly lower overhead
              since the code calls notify_all less often. It still
              suffers from the "thundering heard" problem. When the
              coordinator thread signals that the worker threads should
              start, they all wake up from Monitor::wait and they all
              try to lock the Monitor.<br>
              <br>
              The second, and the more interesting, implementation uses
              semaphores. When the worker threads wake up from the
              semaphore wait, they don't have to serialize the execution
              by taking a lock. This greatly decreases the time it takes
              to start and stop the worker threads.<br>
              <br>
              The semaphore implementation is used on all platforms
              where the Semaphore class has been implemented in
              JDK-8087322. So, on some OS:es the code will revert to the
              Monitor-based solution until a Semaphore class has been
              implemented for that OS. So, porters might want to
              consider implementing the Sempahore class.<br>
              <br>
              There's also a diagnostic vm option
              (-XX:+/-UseSemaphoreGCThreadsSynchronization) to turn off
              the Semaphore-based implementation, which can be used to
              debug this new code. It's mainly targeted towards support
              and sustaining engineering.<br>
              <br>
              <br>
              The patches have been performance tested on Linux,
              Solaris, OSX, and Windows.<br>
              <br>
              The effects of the patch can be seen by running benchmarks
              with small young gen sizes, which triggers frequent and
              short GCs.<br>
              <br>
              For example, here are runs from the SPECjvm2008
              xml.transform benchmark with:<br>
              -Xmx1g -Xms1g -Xmn64m -XX:+PrintGC -XX:+UseG1GC -jar
              SPECjvm2008.jar -ikv xml.transform -it 30 -wt 30<br>
              <br>
              I got the following GC times:<br>
              <br>
              <tt>            Average    Median    99.9 percentile   Max</tt><tt><br>
              </tt><tt>Baseline:   </tt>
              <meta http-equiv="content-type" content="text/html;
                charset=UTF-8">
              <title></title>
              <meta name="generator" content="LibreOffice 4.4.0.3
                (Linux)">
              <style type="text/css">
                body,div,table,thead,tbody,tfoot,tr,th,td,p { font-family:"Liberation Sans"; font-size:x-small }
        </style><tt> </tt>
              <meta http-equiv="content-type" content="text/html;
                charset=UTF-8">
              <title></title>
              <meta name="generator" content="LibreOffice 4.4.0.3
                (Linux)">
              <style type="text/css">
                body,div,table,thead,tbody,tfoot,tr,th,td,p { font-family:"Liberation Sans"; font-size:x-small } </style><tt>8.76</tt><tt>
                ms    8.44 ms   25.9 ms           </tt><tt>3</tt><tt>4.7

                ms</tt><tt><br>
              </tt><tt>Monitor:</tt><tt>    6.1</tt><tt>7 </tt><tt>ms   

              </tt><tt>5.88 ms   26.0 ms           </tt><tt>49.1 ms</tt><tt><br>
              </tt><tt> </tt>
              <meta http-equiv="content-type" content="text/html;
                charset=UTF-8">
              <title></title>
              <meta name="generator" content="LibreOffice 4.4.0.3
                (Linux)">
              <style type="text/css">
                body,div,table,thead,tbody,tfoot,tr,th,td,p { font-family:"Liberation Sans"; font-size:x-small }
        </style><tt>Semaphore:  </tt><tt>3.</tt><tt>43 ms    </tt><tt>3.26 </tt><tt>ms






                  13.4 ms           33.4 ms</tt><tt><br>
              </tt><br>
              If I run an empty GC task 10 times per GC, by running the
              following code:<br>
              <a moz-do-not-send="true" class="moz-txt-link-freetext"
                href="http://cr.openjdk.java.net/%7Estefank/8087324/timedTask/">http://cr.openjdk.java.net/~stefank/8087324/timedTask/</a><br>
              <br>
              I get the following numbers to complete the empty GC
              tasks:<br>
              <br>
              <tt>            Average    Median    99.9 percentile   Max</tt><tt><br>
              </tt><tt>Baseline:   </tt><tt>1.43 ms    0.92 ms   3.43
                ms           9.30</tt><tt> ms</tt><tt><br>
                Monitor</tt><tt>:</tt><tt>    0.75</tt><tt> </tt><tt>ms   


              </tt><tt>0.72 ms   1.74 ms           2.78</tt><tt> ms</tt><tt><br>
              </tt><tt> </tt><tt>Semaphore:  </tt><tt>0.</tt><tt>07
                ms    </tt><tt>0.07 </tt><tt>ms   0.17 ms          
                0.26 ms</tt><tt><br>
              </tt><br>
              <br>
              <br>
              The code has been tested with JPRT and our nightly testing
              suites. <br>
              <br>
              I've created a unit test to run a small test with both the
              semaphore implementation and the monitor implementation:<br>
              <a moz-do-not-send="true" class="moz-txt-link-freetext"
                href="http://cr.openjdk.java.net/%7Estefank/8087324/workgangTest/">http://cr.openjdk.java.net/~stefank/8087324/workgangTest/</a><br>
              <br>
              But since we currently don't have code to shutdown worker
              threads after they have been started, I don't want to push
              this test (or clean it up) until we have that in place. I
              created this bug for that:<br>
              <a moz-do-not-send="true" class="moz-txt-link-freetext"
                href="https://bugs.openjdk.java.net/browse/JDK-8087340">https://bugs.openjdk.java.net/browse/JDK-8087340</a><br>
              <br>
              Thanks,<br>
              StefanK<br>
              <br>
              <br>
              <br>
            </blockquote>
            <br>
          </blockquote>
          <br>
        </blockquote>
        <br>
      </blockquote>
      <br>
    </blockquote>
    <br>
  </body>
</html>