Pure Interpreter Speed

Thomas Wuerthinger thomas.wuerthinger at oracle.com
Fri Oct 24 00:37:13 UTC 2014


Can you post the “-version” results for the two VMs you are comparing and the two command lines?

Thanks, thomas


On 23 Oct 2014, at 15:53, Stefan Marr <java at stefan-marr.de> wrote:

> Hi:
> 
> I was measuring the pure interpreter speed of TruffleSOM and would just like to hear whether the following is plausible.
> 
> I ran my benchmarks on top of a plain Hotspot as well as one with Graal.
> For the one with Graal, I also passed in the following flag to ensure I see the interpreter speed only:
> -G:TruffleCompileOnly=__something_that_really_doesnt_exist__so_that_nothing_is_compiled__
> 
> As a result, I see the pure interpreter on top of Graal being somewhere between 20% and 80% slower than on plain Java.
> Does that sound about right? I assume this overhead is the additional book keeping done to facilitate compilation?
> 
> Thanks
> Stefan
> 
> 
> -- 
> Stefan Marr
> INRIA Lille - Nord Europe
> http://stefan-marr.de/research/
> 
> 
> 



More information about the graal-dev mailing list