Intel® Threading Building Blocks

Why does a range need a copy (instead of a move) constructor?

TBB's range concept requires a copy constructor. Unfortunately, the reference manual does not specify why this is required, and algorithmically duplicating a range makes no/little sense. I can see that moving a range around between different parts of the algorithms may be necessary, but this can be achieved either by copying pointers or by using the move constructor. Are there plans to relax the concept to allow move-only constructible ranges (but still work with copy-constructible)? Or are there reasons (which) why this cannot be done?

can't build TBB with clang

Using tbb41_20130116oss and clang on Mac OS X with the following invocations

$ make compiler=clang arch=ia32
$ make compiler=clang arch=intel64

The build fails and the output contains ...

Undefined symbols for architecture <arch>:
"vtable for tbb::tbb_exception", referenced from:
 -exported_symbol[s_list] command line option

If you also have set the following environment variables

tbb_config file

Hi,

I am using the new tbb release within vs2010 and using icl 12.1. All builds are 64bit  (win7).

I am trying to use the scalable_allocator construct method. I assume that because variadic templates are supported in icl 12.1,

I will be able to use the overloaded version of the function that forwrds arguments (variadic argument list).

However, the compiler does not "pick up". I was hoping that this would be activated:

tbb::concurrent_queue push and pop with move semantics

Hello Intel,

I'm writing a c++11-native set of libraries and am using std::unique_ptr<> - style pointers in all relevant circumstances. I see that the tbb::concurrent_queue push signature requires a copy. 

Would you please consider adding a std::move-enabled move to your push and pop signatures, as in:

void push( const T&& source );

bool try_pop( T&& destination );

Perhaps under the covers this is a bad idea, but I don't see why it would be. 

Thank you,

-Jay

TBB spawns more threads than requested

Hi all,

I am implementing and tuning an algorithm and need to specify for the exact number of threads to use. I can observe that no more threads than specified are active at a time, however the scheduler seems to sometimes spawn new threads. Old theads then become idle.

Can anyone please elaborate why the scheduler thinks this is worthwhile?

Thanks and Best Regards!

how to use the specific core to run specific application program

Dear all

We want to running the nake running program on specific interl cpu core on which there isn't linux operation system. Alternatively, we should running the specific Linux application program on specific core.

Please tell me how to realize this requirement.

Thank you very much.

页面

订阅 Intel® Threading Building Blocks