On Sun, Nov 15, 2009 at 02:54:23PM +0100, Tobias Ulmer wrote:
On Sun, Nov 15, 2009 at 02:24:41PM +0100, frederic wrote:
All this is more than just nit-picking. Pike claims a 10-20% loss
compared to C, which would still be quite good. However, the first
benchmarks tell another story at the moment:
http://shootout.alioth.debian.org/u64/benchmark.php?test=all&lang=go
Be careful with this link, the results may be ok for higher level
languages with built-in datatypes, but they are not for low level
languages. One example i've looked at just yesterday demonstrates
this nicely:
http://shootout.alioth.debian.org/u64/benchmark.php?test=knucleotide&lang=all&box=1
Indeed, I found 3 more of this "anomalies". I think it has more to do
with social aspects - that is the fact that there are people interested in
improving the benchmark results in language X - than with high/low level;
the general shape of the results shows quite the opposite of your
hypothesis.
I didn't dig into your example, but one possible reason C++ does better
than
C is that the programmer uses templates in order to perform compile-time
evaluation.
Take a look at the implementation before you draw any conclusion,
that's all.
I pointed out this benchmark precisely because I was puzzled by the low
performances of Go - I don't think Pike is lies. I was wondering if the
10-20% loss applied to gccgo or 6g.
When I compiled Go, I saw that these benchmarks were in the package, so
I thought that they were ok for the shootout (including appropriate
makefiles).
But apparently the submitters have screwed it up.
I'd also mention that that compares 6g, rather than GCC Go, to GCC C. 6g
uses very meager optimizations, and they claim that the GCC version is
faster. The standard caveat of the shootout stands: it compares
implementations, not languages.
Yes, I was also puzzled that they use 6g instead of gccgo, or even better
that
they didn't submit both. It looks like sabotage.
That said, I don't see what's wrong with using this benchmark to compare
the
efficiency of two languages - provided that one interprets the results
carefully.
