|Posted by Russel Winder||PermalinkReply|
- signature.asc (This is a digitally signed message part)
On Tue, 2011-12-20 at 23:13 -0800, Isaac Gouy wrote:
> > From: Russel Winder <email@example.com>
> > Actually I am surprised that Java does so well in this comparison due to its start-up time issues.
> Perhaps the start-up time issues are less than you suppose.
Very possibly the case, I have only switched to Java 7 recently and haven't had time to assess start up or JIT kick in times. Great strides in startup time have been made with each release of Java, at least on Linux, using mmap and preloaded runtime infrastructure.
> The Help page shows 4 different measurement approaches for the Java program, and for these tiny tiny programs, with these workloads, the "excluding start-up" "Warmed" times really aren't much different from the usual times that include all the start-up costs -
My experience is similar, that JIT warmup is only a small effect when using int and yet quite dramatic when using long. This is to be expected though because a long is not an atomic type in the JVM but is two ints.
> Your "Groovy rewrite effort" didn't contribute a single program in 6 months !
In the interim Groovy had been ejected so there was no point. No real need, and probably highly inappropriate to rehearse all the arguments here. It's water under the bridge. There is no no enthusiasm for getting back into the shootout since Groovy is not a language for computationally intensive code, that is the realm of D, C, C++, Fortran, and sometimes Haskell. If we want to progress this point we should do so on the Groovy lists, or privately, rather than here.
> > There is no point in a language development team running a benchmark.
> Tell that to the PyPy developers http://speed.pypy.org/
That is only a comparison of PyPy against CPython using the CPython benchmarks. Those are internal Python benchmarks. So yes very reasonable.
> Tell that to Mike Pall http://luajit.org/performance_x86.html
An internal comparison of Lua engines. Entirely reasonable.
> Tell that to the Go developers
Do you have a particular URL in mind?
My point though was that (and this is where Walter received so much
flak) where a language vendor tries to do comparisons with other
languages there are always claims of bias whether true or not. In this
sense the Alioth Shootout has the benefit of being clearly independent
of any language vendor. The examples above are not at all inconsistent
with my point.
Russel. ============================================================================= Dr Russel Winder t: +44 20 7585 2200 voip: sip:firstname.lastname@example.org 41 Buckmaster Road m: +44 7770 465 077 xmpp: email@example.com London SW11 1EN, UK w: www.russel.org.uk skype: russel_winder