tbb::cache_aligned_allocator or scalable_allocator?

[buraktamturk]

Hello, i am tring to use tbb's new allocator on every container and string. Which allocator is faster for strings and stringstreams?

std::basic_stringstream, tbb::tbb_allocator> content1;
or like that std::map<:BASIC_STRING>, tbb::tbb_allocator>, std::basic_string, tbb::tbb_allocator>, less<:BASIC_STRING>, tbb::tbb_allocator>>, tbb::tbb_allocator<:PAIR><:BASIC_STRING>, tbb::tbb_allocator>, std::basic_string, tbb::tbb_allocator>>>> qlUser;

and how i can use tbb::interface4::concurrent_hash_map, there is a several example but i cannot get work with small isapi filter. And does it will incrase speed just for 6-7 data?

and there is a tbbmalloc_proxy.h header file,
when i use it in isapi ext and filter will my iis use tbb allocator, or just my codes, how it works? Does it replace normal malloc function pointers to scalable ones?

Thank You,
Best Regards.

[TimP · Accepted solution]

There's a separate forum section for tbb. If I had to guess, I wouldn't expect signficant differences among allocators, provided that you minimize the number of separate allocator invocations.