CUDA core median of choice

Question

I need to calculate the median of an array of size p inside the CUDA kernel (in my case p is small, for example p = 10). I use the O (p ^ 2) algorithm for its simplicity, but at the expense of time performance.

Is there a “function” to find the median efficiency that I can call inside the CUDA core?

I know that I can implement a selection algorithm, but I'm looking for a function and / or verified code.

Thank!

Answer 1

O (p * log (p)), O (p ^ 2) . p-, , O (log (p)), p. . :

:

Answer 2

:

2. : QuickSelect - QuickSort k- . , , TLP O (log ^ 2 n). 8- , , . , , , . , ( , , SM ), .
3. . , . .
4. , , shared memory, , .
5. , (, 3x3 5x5), , . , , , , .
6. , .

, . , CUDA, . , CUDA, , Visual Profiler.