Subject: Re: [boost] Final benchmark graphs for Colony vs std:: containers now available
From: Soul Studios (matt_at_[hidden])
Date: 2016-05-09 19:01:13
> What would be interesting to see is a bench-mark against tbb's (Thread
> Building Blocks) concurrent_vector (and maybe some of the other containers
> in that library) as this has a similar growth strategy (and other
> characteristics), but alllowing for (some) concurrent access...
Actually tbb concurrent_vector doesn't allow erasures other than
clear(), so it wouldn't be suitable for comparisons in the situations
where you'd use a colony. Interesting structure though.
A colony should be able to have more concurrent accesses than an
equivalently-multithreaded vector, as the block-based approach means you
can have mutexes on individual blocks rather than the whole thing. Also
some reads and writes can occur at the same time.
> Bench-marking anything against std::deque on Windows is a rather futile
> exercise as the implementation is broken, the result of a maximum
> chunk-size of 16 bytes (no, no typo) or the size of the (one) object if
> larger. Changing this is on the M$-to-do-list, but will not feature untill
> a major version change (source STL)...
Yes, I noticed the MSVC deque performance results were weak, that's a
pretty extraordinarily bad implementation! Wow.
Hence why the main benchmarks are in GCC.
I've noticed deque is actually better than vector under GCC, for most
circumstances. Though I don't know how that holds up under vectorisation.
Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk