CUDA Supercomputing
|
CUDA stands for Compute Unified Device Architecture
and describes a certain GPU architecture as well as the
corresponding development environment for software
engineers to generate the program code for NVIDIA
GPUs.
This includes three important components:
The CUDA toolkit is in principle a C development environment and includes the actual compiler (nvcc), an update of the PathScale C compiler, optimised FFT and BLAS libraries as well as a profiler, debugger (gdb), shared libraries for the runtime environment for CUDA programs (the “Runtime API”) and last but not least, comprehensive documentation including a developer’s manual. The CUDA SDK includes examples with source codes for matrix calculation, pseudo random number generators, image convolution, wavelet calculations and a lot more besides. CUDA runs on all NVIDIA GPUs from series G8X and above including the GeForce, Quadro and Tesla series and enables programmers access to the highperformance capacity of massive parallel calculation components in GPUs. Unlike CPUs, GPUs have a “many-core architecture” which means that they deliver a parallelisation degree two times larger than standard CPUs. Furthermore, a GPU has its own memory which can be communicated with by the execution units of the individual cores with its extremely high bandwidth. If the entire GPU architecture is outsourced to an external system, a gateway system is required as the main computer to which the GPU systems are connected via external PCI Express connectors. |
|
|
|
||
Downloads |