CUDA
CUDA
cuda_introduction
this summer, the face of supercomputing will change
Presenting The CUDA Hands-On Workshop, part of Shaastra '09. CUDA (Compute Unified Device Architecture) is nVIDIA's API to run general purpose C programs on their Graphics Cards, which can perform many operations in parallel.
In this workshop, you'll learn about the supercomputer that is lurking
inside your desktop, and "get your hands dirty" trying to execute
programs on the Graphics Card and seeing insane speedups when compared
to the embarrassingly serial programs.
Supercomputing is no longer expensive and the prerogative of the elite. It is pervasive, affordable and for the first time, open to all. Be part of the revolution!
Before you let stuff like API and parallel program scare you, check this out. CUDA applications have been developed with great success in diverse areas like Digital content creation, Image processing, Signal Processing, Numerics, Life Sciences, Computational Fluid Dynamics, Medical Imaging, etc. The raw power of CUDA lends itself nicely to simulations and other programs that are highly compute-intensive.
Applications already developed (with speedup achieved) include
- Particle Swarm Optimization - 270x
- Cubic Interpolation - 327x
- Accelerating Leukocyte Tracking - 29x
- AES Cryptography Acceleration - 12x
- SETI@HOME - 10x
- Molecular Dynamics of DNA and Liquids - 18x [source:www.nvidia.com]
Supercomputing is no longer expensive and the prerogative of the elite. It is pervasive, affordable and for the first time, open to all. Be part of the revolution!
Registrations for the workshop is closed!
In case you're shortlisted, we'll email you by September 22nd.
There might be a walk-in session for those who've not registered, on a first-come first-serve basis. Please contact the Help Desk for details during Shaastra.
There might be a walk-in session for those who've not registered, on a first-come first-serve basis. Please contact the Help Desk for details during Shaastra.
cuda_faq
1. Sounds fascinating, but I'm not a CS guy. Will I gain by attending? Is there any point?
Of course! In fact, we encourage people from diverse disciplines to register, as this is highly interdisciplinary in nature.
2. Parallel programming? API? Sounds complicated.. You sure I can manage?
We think you can. All it needs is basic familiarity with C.
3. I don't have a graphics card.
Np. You can use the API and code, compile and run on a system without a graphics card too. The compiler (which is free, btw) has an option for this, called the device emulation mode.
4. Help! I've never programmed on CUDA before!
Np again. The workshop will start from the basics.
5. Will it be another boring class-room lecture?
No!! It'll be completely hands-on, in a lab. We don't talk much. We tell you the basics. You read. You code.
6. Cool! How can I know more about this?
You could read up online on GPGPU, CUDA, SIMD, Data Parallelism, and also visit the nVIDIA's CUDA page for a lot of information and material, including the manual and the compiler.
Of course! In fact, we encourage people from diverse disciplines to register, as this is highly interdisciplinary in nature.
2. Parallel programming? API? Sounds complicated.. You sure I can manage?
We think you can. All it needs is basic familiarity with C.
3. I don't have a graphics card.
Np. You can use the API and code, compile and run on a system without a graphics card too. The compiler (which is free, btw) has an option for this, called the device emulation mode.
4. Help! I've never programmed on CUDA before!
Np again. The workshop will start from the basics.
5. Will it be another boring class-room lecture?
No!! It'll be completely hands-on, in a lab. We don't talk much. We tell you the basics. You read. You code.
6. Cool! How can I know more about this?
You could read up online on GPGPU, CUDA, SIMD, Data Parallelism, and also visit the nVIDIA's CUDA page for a lot of information and material, including the manual and the compiler.
cuda_contact
Karthik: cuda@shaastra.org