J
Juha Nieminen
I wrote this allocator which you can use with std::list, std::set and
std::map (cannot be used with std::vector nor std::deque, but that
doesn't matter because there would be no benefits with those).
I have measured amazing speedups when using this allocator. For
example I have made several speed tests using an std::list<int> using
the default allocator and my allocator, adding and removing large
amounts of elements from the list repeatedly and I have measured huge
speedups using my allocator. For example with the test I describe in the
webpage below, 1min 12secs vs. 16secs. In a test program which I created
specifically to thoroughly test the validness of the library I got
9mins 33secs vs. 1min 2secs.
Peak memory usage is harder to measure because there doesn't seem to
be any reliable way of doing that in linux. However, from the little I
have been able to measure, the peak memory usage dropped considerably
when using my allocator (almost to half with std::list<int>).
Of course there are some disadvantages to the allocator, and it isn't
suitable for all possible situations. Read more at:
http://warp.povusers.org/FSBAllocator/
Tests and feedback would be appreciated.
std::map (cannot be used with std::vector nor std::deque, but that
doesn't matter because there would be no benefits with those).
I have measured amazing speedups when using this allocator. For
example I have made several speed tests using an std::list<int> using
the default allocator and my allocator, adding and removing large
amounts of elements from the list repeatedly and I have measured huge
speedups using my allocator. For example with the test I describe in the
webpage below, 1min 12secs vs. 16secs. In a test program which I created
specifically to thoroughly test the validness of the library I got
9mins 33secs vs. 1min 2secs.
Peak memory usage is harder to measure because there doesn't seem to
be any reliable way of doing that in linux. However, from the little I
have been able to measure, the peak memory usage dropped considerably
when using my allocator (almost to half with std::list<int>).
Of course there are some disadvantages to the allocator, and it isn't
suitable for all possible situations. Read more at:
http://warp.povusers.org/FSBAllocator/
Tests and feedback would be appreciated.