Thanks,
I'll try to figure out how to benchmark and profile the code then.
IMP is a very large and complex piece of software... so you'll have to be > more specific - *what* is slower? Benchmarking, as Riccardo suggests, is > one option - another would be to profile your code. > > IMP 2.0.1 seems to be much faster (up to 6 times faster) as compared to >> versions 2.1, 2.2 and 2.3.1 (compiled with the "Release" flag). >> > > What I meant is that IMP 2.0.1 takes 6 times less then the others to perform the exact same task (e.g. one times unit vs six time units).
> What do you mean by "Release" flag? If you mean the cmake build type > ("-DCMAKE_BUILD_TYPE=Release") then that doesn't do a whole lot (IIRC, > maybe some slightly different compile options on Linux; on Windows it'll > change which C runtime you link against). What will make the most > difference is what IMP check level you build with; if you use > -DIMP_MAX_CHECKS=INTERNAL then the internal checks will definitely slow > things down.
I forgot to say that I am using the same cmake build types when complaining different IMP versions (-DCMAKE_BUILD_TYPE=Release -DIMP_MAX_LOG=SILENT -DIMP_MAX_CHECKS=NONE). Anyway, I will do the benchmark.
Davide
> > > Ben > -- > ben@salilab.org http://salilab.org/~ben/ > "It is a capital mistake to theorize before one has data." > - Sir Arthur Conan Doyle > > _______________________________________________ > IMP-dev mailing list > IMP-dev@salilab.org > https://salilab.org/mailman/listinfo/imp-dev >