Projects tagged ‘gpgpu’

Perl module that provides tight integration between OpenGL and various imaging libraries. In the case of ImageMagick, C pointers may be shared between GPU and IM's image cache for optimized performance.

Perl module that abstracts OpenGL APIs for use with ARB (assembly), Cg and GLSL shading languages.

Perl OpenGL Module * Portable * Comparable to C performance * Supports over 50 OpenGL extensions

Brahma is a library written for the .NET platform, to provide high-level access to graphical and general purpose processing on the GPU. It eliminates the need for a shading language, generating shader code directly from IL at runtime.

Thrust is a CUDA library of parallel algorithms with an interface resembling the C++ Standard Template Library (STL). Thrust provides a flexible high-level interface for GPU programming that greatly ... [More] enhances developer productivity. Develop high-performance applications rapidly with Thrust! [Less]

C# wrapper for Ageia PhysX. It allows using the PhysX api in the .NET framework or XNA applications.

It's a research project which aims to implements tomography algorithm in Stream technology.

This project is a quick port of the "Standard Model of Visual Cortex" (CBCL, MIT, by Riesenhuber M., Poggio T., Serre T., Wolf L., ; I've got some missing names please let me know) I first started ... [More] working on this model in 2007, but I only worked a month or two on the project (after coding a working gpu port). After some years of research in computational neuroscience (and mastering GPU programming) I've decided to start the code from scratch, as I needed to experiment a bit further with it, comparing it to other models and simulations. The resulting code is fast and robust, it's a mere ~500 lines of code and I managed to code it in less than a week (probably it would have taken just an afternoon if it wasn't GPU code, which requires days to fix small bugs :P). Code is released under GPL2. Feel free to use it without problems. Benchmark: The code has been benchmarked on the BASS supercomputer, by University of North Carolina. The test node features an nVidia Quadro FX 5600 GPU (equivalent to nVidia GTX 8800) GPU cores, 4GB RAM, 1.35Ghz shader clock, 76GB/s memory speed. Time to process a 256x256 image is ~345ms (of which ~330ms to compute S2 features); to process a 320x240 image requires ~405ms. I'll try to compare with the official (released) code to determine the speedup. Some quick notes: ** Current implementation only allows 4x4 and 8x8 C1 patches to be used at S2 level (test programs use 1024 4x4 features and 256 8x8 ones). ** I'm exploring a variant of the base model (code is not in the repository, though) which adds a DoG filter before passing the input image on to the StandardModel. Performance on Car Vs Non-Car (UIUC database) with C1 patches extracted at random from a different database is about ~70-75% with normal model, and ~85% using DoG images as input. To try it out just modify this piece of code under cuda/standard.cu, SetPyramid_0() function, at the top: cvCvtColor(im, pyr0, CV_BGR2GRAY); IplImage *g1=cvCreateImage(cvSize(im->width, im->height), 8, 1); IplImage *g2=cvCreateImage(cvSize(im->width, im->height), 8, 1); cvSmooth(pyr0, g1, CV_GAUSSIAN, 5, 0, 1.5); cvSmooth(pyr0, g2, CV_GAUSSIAN, 5, 0, 1.0); cvAbsDiff(g1, g2, pyr0); cvThreshold(pyr0, pyr0, 5.0, 255.0, CV_THRESH_TOZERO); cvReleaseImage(&g1); cvReleaseImage(&g2); References: - 1) "Object Recognition with Features Inspired by Visual Cortex", Serre T., Wolf L., Poggio T. link - 2) "Robust Object Recognition with Cortex-Like Mechanisms", Serre T., Wolf L., Bileschi S., Riesenhuber M., Poggio T. link - 3) "Multiclass Object Recognition with Sparse, Localized Features", Mutch J., Lowe D. G. link - 4) "A Theory of Object Recognition: Computations and Circuits in the Feedforward Path of the Ventral Stream in Primate Visual Cortex", Serre T., Kouh M., Cadieu C., Knoblich U., Kreiman G., Poggio T. link [Less]

"In Roman mythology, Laverna was a goddess of Thieves, Cheats and the underworld. She was propitiated by libations poured with the left hand. The poet Horace and the playwright Plautus call her a ... [More] goddess of thieves." Laverna's Brute: A hash cracker auditor utilizing a custom high-speed processing engine, support for a nearly limitless number of threads, cracking auditing NTLM hashes, (Support for others can be easily added) and future plans for GPGPU acceleration via OpenCL. (ATI and Nvidia support) For users without a supported GPU, a high-performance CPU-only processing path is available. Anything that can be offline cracked audited is a potential candidate. For a complete summary, visit http://code.google.com/p/lavernasbrute/wiki/Description Follow the progress of Laverna's Brute at the Laverna's Brute Project Blog. [Less]

White space sensing algorithms and their cuda implementations.