Subject: Re: [boost] [Boost-users] [compute] Review period starts today December 15, 2014, ends on December 24, 2014
From: Denis Demidov (dennis.demidov_at_[hidden])
Date: 2014-12-16 02:25:05
I am the one who submitted the review to the boost incubator, but I think
it should be updated.
Also I think its easier to read the review as the whole piece here than on
the incubator site.
### Design ###
Boost.Compute provides a thin C++ wrapper around OpenCL host API at its core
and builds a set of STL-like algorithms on top of that core. The user
strongly resembles the STL and hence should be familiar to any C++
The only minor problem I have with the design is the decision to provide
another C++ wrapper for OpenCL host API instead of using the standard C++
bindings header  provided by the Khronos group (the body behind the
standard). This decision makes interaction with the existing OpenCL
(that use Khronos C++ bindings) somewhat complicated at times. I don't
its possible to change the design at this point though, so I am prepared to
live with it.
### Implementation ###
Having proposed several patches to the library, I can say that I am familiar
with its implementation. The library is well structured, the code is well
designed, well formatted and is easy to read and understand. The library
provides a large number of examples and an extensive set of unit tests.
When the first public announcement of the library was made here on Boost
mailing list, there were several performance problems. In particular the
compute kernels were not cached at the first invocation, and some algorithms
only provided serial implementation (some still do ). I know that a lot
effort has been put into the implementation since then, and the situation
much improved. The performance page  of the documentation shows that
Boost.compute is able to outperform Nvidia's Thrust for some algorithms, but
there is still some work to be done.
### Documentation ###
The documentation does a good job at providing both an overview of the
and an extensive API reference. Boost.compute uses Boostbook as the
documentation generator, and has a look and feel compatible with the
of Boost libraries.
### Potential usefulness of the library ###
I'd say that a library that allows to easily harvest the performance
by the modern graphic processors and accelerators is extremely useful. For
as an end user, the convenience could even outweigh a loss of a fraction of
Since the library interface is so close to STL, it is extremely easy to try
use. I have successfully compiled the library with the recent versions of
and Clang. The unit tests run fine on NVIDIA, AMD, and Intel OpenCL
Among the current alternatives to the library that provide an STL-like set
containers and algorithms the Boost.Compute is the most portable, being
on top of standard OpenCL (see  for my take on differences between
Boost.Compute and alternatives at stackoverflow.com).
One thing that could potentially make Boost.compute obsolete is the
of n3960  into standard. Kyle, what do you think about this?
### Familiarity with problem domain ###
I am the author of VexCL  library that has similar functionality to
Boost.Compute, but provides higher level interface. I would say that I am
familiar with GPGPU programming, both CUDA and OpenCL. I have provided an
implementation of Boost.Compute backend (algebra and operations) for
Boost.Odeint library , and made a couple of Boost.Compute algorithms
available through VexCL interface.
### Conclusion ###
I think an inclusion of a GPGPU library into Boost is long overdue. In my
opinion, Boost.Compute deserves to be accepted. The interface of the
well designed, but it needs some work on the performance of the provided
algorithms. Due to high reputation of Boost collection of libraries, a newly
included library almost automatically becomes a de-facto standard in its
This is why I want a GPGPU library accepted into Boost to show a state of
art performance. I still believe the work on performance may be continued
the library is accepted into Boost.
### References ###
1. [cl.hpp](http://www.khronos.org/registry/cl/api/1.2/cl.hpp) -- OpenCL
Bindings Header File, implementing the [C++ Bindings
3. Boost.compute [performance](
4. Answer to [Differences between VexCL, Thrust, and
5. [TS for C++ Extensions for Parallelism](
6. [VexCL](https://github.com/ddemidov/vexcl) -- a C++ vector expression
template library for OpenCL/CUDA.
7. [Boost.compute backend for Boost.odeint](http://goo.gl/xZSd10).
Senior researcher at Supercomputer center of the Russian Academy of