We present the development of one of the first libraries based on the so-called expression templates technique to simplify the implementation of CPU and parallel GPU codes. Expression templates allow to express matrix algebra operations to be executed either on the CPU or on the GPU with a syntax very close to the natural mathematical one. The developed library has been deeply optimized so that the same performance of standard C/C++ or CUDA codings is kept.
Achieving natural mathematical expression programming on GPUs via expression templates / A., Breglia; Capozzoli, Amedeo; Curcio, Claudio; Liseno, Angelo. - (2013), pp. 469-474. (Intervento presentato al convegno Europea Modelling Symposium 2013 tenutosi a Manchester, UK nel 20-22 Novembre 2013) [10.1109/EMS.2013.84].
Achieving natural mathematical expression programming on GPUs via expression templates
CAPOZZOLI, AMEDEO;CURCIO, CLAUDIO;LISENO, ANGELO
2013
Abstract
We present the development of one of the first libraries based on the so-called expression templates technique to simplify the implementation of CPU and parallel GPU codes. Expression templates allow to express matrix algebra operations to be executed either on the CPU or on the GPU with a syntax very close to the natural mathematical one. The developed library has been deeply optimized so that the same performance of standard C/C++ or CUDA codings is kept.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
