On Wednesday 19 November 2008 16:59, Hansi wrote:
I have made a few tests with the library. I think the library is good, but I don't know if I make something wrong, but it seems for me that the library is slow. I have made a few tests with VC++ 8.0 and _SCL_SECURE=0 and it seems that it is about 50 times slower than a normal boost::function with a bind.... It's a lot...
I distilled your benchmark down to something minimal I could compile under Linux (attached) and I get: $ ./a.out 0.25 s 10000000 2.24 s when compiling with gcc with -O3. So, I'm getting a factor of 9. The original Boost.Signals doesn't do any better in this particular benchmark. One odd thing I noticed is the signals2 benchmark actually runs slower (2.8 sec) if I define NDEBUG!