|
BenchResults
Results of the current completed benchs.
The generated results can be found on the on the SVN : Time bench Size bench Versions
Time resultshttp://vala-benchmarks.googlecode.com/svn/trunk/bench-time.svg
Size resultshttp://vala-benchmarks.googlecode.com/svn/trunk/bench-size.svg
Memory consumed resultsThe method of memory measurement is pretty simple : time --format=%M
%M : Maximum resident set size of the process
during its lifetime, in Kilobytes.The project's goal is to become part of "The Computer Language Benchmarks Game", not to implement complex memory measurement systems :). due to the time command bug, these results are not available |
Sign in to add a comment
Are you still thinking of adding support for new environments? :)
As I have said before, it's very hard to make. It will break the automatic way my benchmark work by now.
Using Mingw based binaries may also introduce false results (using cygwin will probably moreover change the results). Furthermore all the tools I use to perform automatic rendering are Unix based (look at bench.sh). If you want you can take the C# file and make a VS project. I will pleasely add it to the bench (out of the automatic way).
I am also interested in comparission to D and MS .NET.
Read the issue 6 and FAQ for more information : http://code.google.com/p/vala-benchmarks/issues/detail?id=6
I think all overhead of C++ is in <iostream>. That is reason why C++ lose plain-C in some tests.
I don't know if the difference come from <iostream> but as seen in the code size a huge piece of code is used by C++ and not by plain-C.
grep VmHWM /proc/<pid>/status is a nice way to measure real memory comsumption.