<div dir="ltr"><div class="gmail_default" style="font-family:arial,helvetica,sans-serif">​&gt;&gt; I think I get the pro of using threaded interrupts - to decrease the maximum<br>
&gt;&gt; interrupt latency on RT workloads and/or RT machines (servers, embedded,<br>
&gt;&gt; etc.).<br>
​&gt;&gt; So, I&#39;d like to ask:<br>
&gt;&gt;    - Why not **all** of the drivers use the threaded interrupts?<br>
&gt;&gt;    - What are the cons of the threaded interrupts?<br>
&gt;&gt;<br>
&gt; Just a wild assumption: maybe the cost of incurring context switches ?<br>&gt; (comparing to tasklets)<br><br>I get that threaded IRQ is better than softIrqs because threaded IRQ supports<br></div><div class="gmail_default" style="font-family:arial,helvetica,sans-serif">priorities. On the other hand, added context switches surely get system slower.<br></div><div class="gmail_default" style="font-family:arial,helvetica,sans-serif">But from the practical point of view - why on relatively new Intel PC there are<br></div><div class="gmail_default" style="font-family:arial,helvetica,sans-serif">no threaded irqs at all?<br><br>```<br>$ uname -a<br>Linux kostaz-OptiPlex-7010 3.19.0-26-generic #28-Ubuntu SMP Tue Aug 11 14:16:32 UTC 2015 x86_64 x86_64 x86_64 GNU/Linux<br>```<br><br></div><div class="gmail_default" style="font-family:arial,helvetica,sans-serif">--- KostaZ<br></div><div class="gmail_extra"><br><div class="gmail_quote">On Thu, Sep 10, 2015 at 11:09 PM, Rami Rosen <span dir="ltr">&lt;<a href="mailto:roszenrami@gmail.com" target="_blank">roszenrami@gmail.com</a>&gt;</span> wrote:<br><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">Kosta,<br>
<br><div class="gmail_default" style="font-family:arial,helvetica,sans-serif;display:inline">​​</div>
Just a wild assumption: maybe the cost of incurring context switches ?<br>
(comparing to tasklets)<br>
<br>
Best Regards,<br>
Rami Rosen<br>
<a href="http://ramirose.wix.com/ramirosen" rel="noreferrer" target="_blank">http://ramirose.wix.com/ramirosen</a><br>
<div><div class="h5"><br>
<br>
<br>
On 10 September 2015 at 20:49, Kosta Zertsekel &lt;<a href="mailto:zertsekel@gmail.com">zertsekel@gmail.com</a>&gt; wrote:<br>
&gt; Hi guys,<br>
&gt;<br>
&gt; I hope I&#39;m on right mailing list. :-)<br>
&gt; I think I get the pro of using threaded interrupts - to decrease the maximum<br>
&gt; interrupt latency on RT workloads and/or RT machines (servers, embedded,<br>
&gt; etc.).<br>
&gt;<br>
&gt; Also, I see that in 4.2 there are only ~76 drivers that use threaded<br>
&gt; interrupt:<br>
&gt; ```<br>
&gt; $ git grep -l IRQ_WAKE_THREAD | sort | grep -v &quot;\.h&quot; | wc -l<br>
&gt; 76<br>
&gt; ```<br>
&gt;<br>
<div class="gmail_default" style="font-family:arial,helvetica,sans-serif;display:inline">​​</div>&gt; So, I&#39;d like to ask:<br>
&gt;    - Why not **all** of the drivers use the threaded interrupts?<br>
&gt;    - What are the cons of the threaded interrupts?<br>
&gt;<br>
&gt; Thanks,<br>
&gt; --- KostaZ<br>
&gt;<br>
</div></div>&gt; _______________________________________________<br>
&gt; Kernelnewbies mailing list<br>
&gt; <a href="mailto:Kernelnewbies@kernelnewbies.org">Kernelnewbies@kernelnewbies.org</a><br>
&gt; <a href="http://lists.kernelnewbies.org/mailman/listinfo/kernelnewbies" rel="noreferrer" target="_blank">http://lists.kernelnewbies.org/mailman/listinfo/kernelnewbies</a><br>
&gt;<br>
</blockquote></div><br></div></div>