I am using Cuda in C++ for some time now and I would like to try cuDNN. I would like to go straight in C++ but I can mostly only find examples and tutorials that are based on different platforms like Caffè or TensorFlow. Does it mean that I cannot use cuDNN in plain C++?
I need to have the highest possible performance.
Thanks for your help
You can use cuDNN in plain C++.
As a library, it is targeted at Deep Neural Networks/Deep Learning, so many DL Frameworks use it for acceleration of their workloads.
But there is nothing that prevents you from using the library directly, and there should be sufficient API level documentation for you to do so.
If you're unfamiliar with some of the concepts (e.g. tensors), you may find the library more difficult to comprehend, so certainly general familiarity with current common DL concepts will be beneficial in understanding how to use it.
Related
I'm looking to getting into image/video processing and was searching for a good library to start with. I've heard of two, OpenCV and IVT. I'd like to hear your opinion about which one is better to start with, what are the advantages/disadvantages of both and which one is better for possible commercial use later on.
Both offer very similar functionality - about 95% of it overlaps.
Both are BSD(ish) licensed and are widely used in commercial packages.
IVT is a cleaner more modern C++ design, but the new c++ bindings to opencv work well. Opencv has a few more 'C' type macros but it also means it's usable from C. Opencv is also very well supported by python and other languages, don't know about IVT.
There is also CImg. It requires only a single header file and uses C++ type templates so you write code like result = image.blur().sharpen().edge() almost like Mathematica!
It doesn't have the same depth of functionality, especially in things like recognition and machine learning, but is definitely easier to use. It's GPL/LGPL so usable commercially with care.
OpenCV is much more widely used, so has a bigger set of users who might answer questions - but it also has a MUCH bigger set of beginners asking questions !
The decider for me is that openCV is moving to support (almost) all the functionality in CUDA (ie on a parallel GPU) which is fantastic for anything needing realtime video processing.
Other than this I couldn't comment on the performance I didn't really benchmark IVT enough. OpenCV does use custom SSE2 assembler for a lot of the operations and uses TBB to parallel the rest if you have a multicore/hyperthreaded CPU.
I am a beginner like you and I personally say OpenCV. My first learning experience with OpenCV was more effortless than IVT because documentation was so neat and clear, also there are common beginner books, many tutorials and example projects for OpenCV.
Greetings,
I am trying to decide between C and C++ for my robot. I am a 5+ year veteran of Visual Basic.NET, however I'm going with Linux (Ubuntu) on this robot, and there is a compatibility problem between Linux and the .NET Framework. I want to stick with whichever language I choose for all of my projects, so I want to make sure that I choose the most appropriate one for the task.
For reference, I will describe my current robot in progress and what I am going to do with it. I am in the process of building a full-sized R4 Astromech (yep, I'm one of those guys). I have incorporated a PC motherboard with an Intel Core 2 2.1 GHz processor, 1 GB RAM. I will be using a scratch-built parallel interface card to control the drive motors, head motor, as well as a secondary parallel interface card (going to a second parallel port) which all of the sensors (IR, Ultrasonic Ranging, Visual Recognition via webcam, etc.) will be going to. Commands will be given using speech recognition (currently have a VB.NET scratch-built recognition program that I will be adapting to the new language).
Given the specifications and desired goals listed above, would I be better off with C or C++? I greatly appreciate any suggestions that you might have.
Thanks!
Thaskalas
What do you mean by a compatibility problem? Have you looked at Mono? It's an open-source implementation of the .NET libraries. It's geared toward C# not VB.NET but if you're more comfortable in a .NET environment use that. Speed isn't really an issue here as a Core2Duo is plenty fast for what you need to do.
If Mono won't work for you, I'd recommend C++. There are a lot more libraries out there for C++ (or at least, I am familiar with more, e.g. Boost), which can use most C libraries too. There's no real speed penalty for using C++. While using C wouldn't be bad per-se, C++ has some benefits and no drawbacks, so it's probably the better choice.
I would recommend using ROS. It will let you get started with a sophisticated Inter-Process Communications manager, as well as a large library of sophisticated robotics code, including multiple implementations of SLAM and other critical robotics algorithms. ROS also lets you program in multiple languages, including C, C++, and Python, so you aren't stuck with one language or another down the road.
I would also recommend C++ and ROS. In our company we're migrating to it, because there's so many people working on it, expanding it, and adding lots of cool features.
With this, you can forget about implementing most of the basic low-level stuff and start working on what you intend to research.
It's really easy to set up and start developing.
Since you're running Linux on it, I'd recommend a split approach, where you do the lower-level (device interface, where you may need fast performance) stuff in C (or C++), and the higher level stuff in a modern language like C# (using Mono) or Java, or even Python.
Python especially is hugely expressive, has a large set of libraries, and has a pretty straightforward C interface.
Writing your high-level control stuff in a low-level language like C/C++ will get old fast (IMHO). Robots should be fun!
Have you considered D? It's a fairly new language, is compiled to native code and can link directly to C. (The entire C standard library is even available from D, and bindings to the POSIX API are included in the standard library.) Basically all you need to do to use any C library from D is compile it with a C compiler and translate the function prototypes, constant declarations, etc. in the header file.
D is low-level enough that an experimental kernel is written in it, but has modern features like garbage collection (though manual memory management is still permitted), builtin strings and arrays, and more advanced/easier to use template metaprogramming facilities than C++. The biggest disadvantage is lack of a mature toolchain and libraries for enterprise-y things, but for your purposes that probably doesn't matter. BTW, if you need to do a bunch of matrix manipulation and stuff, there's the SciD project, which provides nice templated wrappers over LAPACK and BLAS.
Use C++. You have the space. You can use it "as a better C" to start with.
C++ is a bigger tool bag; why would you not want that!? You need not use all the tools, but with C you'd have no choice. Most importantly with C++ you have the choice of using both C and C++ third-party libraries.
I have a fair hang of programming in various languages. I have been implementing my codes for research using MATLAB (during the past few months) and for the first time really noticed the difference in execution speed of MATLAB v$ C. (As much as I love the blazingly fast prototyping capabilities).
I am looking to pickup C++ and start using it in my research. I am aware of OOP and have programmed fair bit of Java (relatively long back) and C++ (even longer back). I would like to really get deep into C++ now and hence need suggestions for resources on the same:
What C++ things I need to pick up (STLs and. ) to really make good use of C++?
What is a good tutorial/manual to get started with?
What are the numerical/scientific libraries for C++? GSL? Is there a equivalent (features) of Scipy/Numpy for C++?
I shall be programming on Linux, so I shall be using g++ .
Any pointers to previous SO questions also appreciated.
You'll want to get to grips with parallel programming as quickly as possible. For message-passing I like this book by Karniadakis and Kirby. Of the books on OpenMP, for distributed-memory programming, this one is the best.
If you can get access to them, then Intel's Threading Building Blocks, Maths Kernel Library, and Integrated Performance Primitives are good to have. If not, there are plenty of open source alternatives, start looking at Netlib.
Oh, I almost forgot BOOST, which is a must.
In regards to numerical stuff like Numpy, you should have a look at both:
Blitz++ http://www.oonumerics.org/blitz/
and
Jama/TNT http://math.nist.gov/tnt/download.html
On the library side, check out Armadillo. It almost gives you the full extent of MATLAB's array manipulation syntax and uses LAPACK and BLAS (ATLAS) under the hood.
This tutorial absolutely rocks, but you may not want to tackle it initially.
http://www.parashift.com/c++-faq/
Make sure to read up on the STL (standard template library) and other stuff, using sites like:
http://cplusplus.com/
And, check out the Boost library:
http://www.boost.org/
To make really good use of C++, you need to learn at least the STL, that alone will save you lots of time, but as parashift mentions, C++ OOP is only programming with objects, if you don't use dynamic bindings.
TRNG is a parallel random number generation library. It allows you to create multiple independent streams and was designed for use on clusters.
I'm looking to add full text indexing to a Linux desktop application written in C++. I am thinking that the easiest way to do this would be to call an existing library or utility. This article reviews various open source utilities available for the Gnome and KDE desktops; metatracker, recoll and stigi are all written in C++ so they each seem reasonable. But I cannot find any notable documentation on how to use them as libraries or through an API. I could, instead, use something like Clucene or Xapian, which are generic full text indexing libraries. They seem more straightforward but if I used them, I'd have to implement my own indexing daemon, an unappealing prospect.
Also, Xesam seems to be the latest thing, does anyone have any evidence that it works?
So, does anyone have experience using any of the applications or libraries? How did you use it and what documentation was useful?
I used CLucene, which you mentioned (and also Lucene.NET), and found it to be pretty good.
There's also Strigi which AFAIK works with Xesam and is the default used in KDE.
After further looking around, I found and worked with Recol. It believe that it has the best C++ interface to a full text search engine, in this case Xapian.
It is important to realize that clucene and Xapian are both highly complex libraries designed primarily for multi-user server applications. Cutting them down to a level appropriate for a client-system is not easy. If I remember correctly, Strigi has a complex, pure C interface which isn't adapted.
Clucene also doesn't seem to be that actively maintained currently and Xapian seems to be maintained. But the thing is the existence of recol, which allows you to index particular files without the massive, massive setup that raw Xapian or clucene requires - creating your own "stemming" set is not normally desirable, etc.
I am writing a small C++ program for fun and for extending my C++ skill. Since its scope is relatively small, I also planning to try out cross-platform development by making this program support both Windows and Linux.
I reckon my C++ proficiency is sitting somewhere between casual and intermediate level: OO, a bit of templates and design patterns, used STL before and trying to look into it more in details, ... However, while coding this little program, I find that the deeper I dig into C++, the more pain I feel, especially when I come to understanding and dealing with differences between different platform's/vendor's implementation.
The use of cross-platform frameworks like Qt, ACE, Boost seems help to speed up development a lot thus make life easier, but I worry if this will beat my purpose. Can somebody give some advice if there is any "best practice" for doing C++ cross-platform development? Thanks.
Can somebody give some advice if there is any "best practice" for doing C++ cross-platform development?
There are three things:
Write your own code so that it's portable
Wrap platform-specific APIs behind an abstraction/insulation/utility layer
Choose cross-platform libraries
You can choose option #2 and/or #3.
Advantages of #3 over #2 tend to be things like, "It's already written, debugged, and supported"; and the disadvantages are like, "I have to learn it, I might have to pay for it, I can't necessarily support it myself, and it may not do exactly what I want."
Developers will often prefer option #3 instead of #2, especially if it's free open source (which all three of the libraries that you cited are).
http://blog.backblaze.com/2008/12/15/10-rules-for-how-to-write-cross-platform-code/
Should provide more detail to the answers already given.
Also I suggest using existing libraries that abstract endianness, data type sizes and differences. The following should be considered before starting your cross-platform project.
GUI
Qt
XVT
wxWidgets
General Libraries/frameworks
STL (Incorporated in most platform libraries already)
Boost
Game Development
SDL
Cocos2d-x
Use gcc. It's available on both Windows and Linux and the libraries and language syntax is identical on both platforms.
For cross platform GUI applications, Qt is a good idea. There is no getting away from having a dependency on a GUI framework if you are trying to achieve platform independence.
Unless you are doing GUI stuff cross platform isn't a big problem.
There are some small issues to do with filesystems ( different / \ separators, allowed characters in filenames etc) but these are at the application level rather than the c++.
Doing major applications gets more complex, you need to handle help, file locations an possibly security and user info in a cross platform way. For simple algorithm type programming there isn't a problem.
Qt is mainly a GUI library, although it has extra cross platform filesystems stuff. STL, Boost, ACE are cross platform but that isn't there main point.
Use them! Seriously. The only reason you may not want to use them is if you plan on working in an environment where they're not available. But, given their cross-platform nature, that's not likely.
You will find that the benefit you get from using them is immense, even if they weren't cross-platform. The "best practice" you speak of is to be able to deliver your "product" as quicly and easily as possible.
I once answered a question from someone who stated he didn't want to use GUI libraries at a level above Xlib. If he'd actually ever used Xlib, he'd know the pain we'd all felt when forced to code at such a low level of abstraction. This makes about as much sense as wanting to code in assembler because C/C++?Python/Perl/everthting-else is simply a higher-level abstarction.
this answers are really good and you can make a list for find where is the beginning. but i think you should read some articles about "porting application".not relevant with cross-platform development but this can give you very large perspective about cross platform development. In cross-platform developing, one of more importing thing is memory issues like "endian" (byte order- byte order can show differences for tehnologies or platforms)
use boost. they take care of cross-platform stuff for you.
boost::filesystem is a great example
i think you can learn a lot from using ACE or equivalent libraries. they will boost your understanding of c++ and design patterns. i think this is the best thing you can do to improve your coding skills.
If your are really interested in making your code as cross-platform as possible, use as many compilers as possible. If you are using Windows and Linux, use VC and gcc, at the minimum. This will ensure that you don't use complier specific features, and that you don't rely on system specific behavior. Use more compilers (Intel, IBM, etc) and OSs (OS X, Solaris) if you have access to them.
You can try using U++ > http://www.ultimatepp.org/index.html