On Saturday, June 11, 2016, Stefan Marr <<a href="mailto:smalltalk@stefan-marr.de">smalltalk@stefan-marr.de</a>> wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">Hi Tim:<br>
<br>
> On 11 Jun 2016, at 07:39, Tim Felgentreff <<a href="javascript:;" onclick="_e(event, 'cvml', 'timfelgentreff@gmail.com')">timfelgentreff@gmail.com</a>> wrote:<br>
><br>
> About also running TruffleSOM, while it might be interesting for some benchmarks, I am really interested in testing VMs that run the full image (including heartbeat, event sensor) while running those benchmarks.<br>
<br>
The JVM generates safe points, yield points, and all the things that are necessary for Java semantics… (i.e., all the equivalents for heartbeat/event sensor overhead).<br>
Also, such minor details don’t account for more than a few percent overhead on average.<br>
<br>
No excuses :-P<br></blockquote><div> </div><div>The point is that we want to benchmark a full Smalltalk system, not just a language runtime. There *is* a difference ;)</div><div><br></div><div>- Bert -</div><br><br>-- <br><div dir="ltr">- Bert -</div><br>