A temporary workarround for parallel find_if

[maribu]

• The new function diffract_gather2 allows an extended collapse function, which also retrieves the subrange belonging to the current value as additional argument (a pair of iterators)
• examples/find_if.cpp - a stupidly simple example to test find_if - was added

[t-lutz]

Thanks for the PR. Could you also add a test case in alg.nonmodifying/alg.gen.find and modify std::find and std::find_if_not for consistency?

[t-lutz]

I pushed a slightly different version using a comparison function instead of a reduction function.

The advantages of my approach compared to yours is that the two diffract variants have the same prototype - except the final operator -, and the comparison is easier to parallelize than the accumulation.

If you want to submit another PR for the example, i will gladly merge it.

Thanks.

[maribu]

Thanks for all your work! Sorry it took me so long to answer.

By the way: You might be interested I'm working with your ParallelSTL. I'm a student at the university of Münster and we're implementing two new policys in project, one running on GPUs using CUDA and one using an experimental compiler that allows programming any CUDA and OpenCL device in C++, which is developed here at the university.

Regards, Marian

[t-lutz]

Is that PACXX? I am aware of it. I wrote a compiler for ParallelSTL with a PTX and a SPIR backend as well, but it is not open source.

[maribu]

Yes it is ;-)

I wrote a compiler for ParallelSTL with a PTX and a SPIR backend as well, but it is not open source.

Very cool!