]>
Commit | Line | Data |
---|---|---|
1 | [/=========================================================================== | |
2 | Copyright (c) 2013-2015 Kyle Lutz <kyle.r.lutz@gmail.com> | |
3 | ||
4 | Distributed under the Boost Software License, Version 1.0 | |
5 | See accompanying file LICENSE_1_0.txt or copy at | |
6 | http://www.boost.org/LICENSE_1_0.txt | |
7 | =============================================================================/] | |
8 | ||
9 | [section:design Design] | |
10 | ||
11 | [section Library Architecture] | |
12 | ||
13 | The Boost Compute library consists of several different components. The core | |
14 | layer provides a "thin" C++ wrapper over the OpenCL API. This includes classes | |
15 | to manage OpenCL objects such as | |
16 | [classref boost::compute::device device]'s, | |
17 | [classref boost::compute::device kernel]'s and | |
18 | [classref boost::compute::device command_queue]'s. | |
19 | ||
20 | On top of the core layer is a partial implementation of the C++ standard | |
21 | library providing common containers (e.g. | |
22 | [classref boost::compute::vector vector<T>], | |
23 | [classref boost::compute::array array<T, N>]) along with common algorithms | |
24 | (e.g. [funcref boost::compute::transform transform()] and | |
25 | [funcref boost::compute::sort sort()]). | |
26 | ||
27 | The library also provides a number of "fancy" iterators (e.g. | |
28 | [classref boost::compute::transform_iterator transform_iterator] and | |
29 | [classref boost::compute::permutation_iterator permutation_iterator]) which | |
30 | enhance the functionality of the standard algorithms. | |
31 | ||
32 | Boost.Compute also supplies a number of facilities for interoperation with | |
33 | other C and C++ libraries. See the section on [link boost_compute.interop | |
34 | interoperability] for more information. | |
35 | ||
36 | See the [link boost_compute.reference.api_overview API Overview] section for | |
37 | a full list of functions, classes, and macros provided by Boost.Compute. | |
38 | ||
39 | [endsect] [/ library architecture] | |
40 | ||
41 | [section Why OpenCL] | |
42 | ||
43 | Boost.Compute uses [@http://en.wikipedia.org/wiki/OpenCL OpenCL] as its | |
44 | interface for executing code on parallel devices such as GPUs and multi-core | |
45 | CPUs. | |
46 | ||
47 | OpenCL was chosen for a number of reasons: | |
48 | ||
49 | * Vendor-neutral, standard C/C++, and doesn't require a special compiler, | |
50 | non-standard pragmas, or compiler extensions. | |
51 | * It is not just another parallel-library abstraction layer, it provides direct | |
52 | access to the underlying hardware. | |
53 | * Its runtime compilation model allows for kernels to be optimized and tuned | |
54 | dynamically for the device present when the application is run rather that the | |
55 | device that was present when the code was compiled (which is often a separate | |
56 | machine). | |
57 | * Using OpenCL allows Boost.Compute to directly interoperate with other OpenCL | |
58 | libraries (such as VexCL and OpenCV), as well as existing code written with | |
59 | OpenCL. | |
60 | * The "thin" C++ wrapper provided by Boost.Compute allows the user to break-out | |
61 | and write their own custom kernels when the provided APIs are not suitable. | |
62 | ||
63 | [endsect] [/ why opencl] | |
64 | ||
65 | [endsect] |