This course gives an overview of the most relevant GPGPU computing techniques to accelerate computationally demanding tasks on HPC heterogeneous architectures based on GPUs.
The course will start with an architectural overview of modern GPU based heterogeneous architectures, focusing on its computing power versus data movement needs. The course will cover both a high level (pragma-based) programming approach with OpenACC for a fast porting startup, and lower level approaches based on nVIDIA CUDA and OpenCL programming languages for finer grained computational intensive tasks. A particular attention will be given on performance tuning and techniques to overcome common data movement bottlenecks and patterns.
Overview of architectural trends of GPUs in HPC. GPGPU parallel programming in heterogeneous architectures. Basis of OpenACC, CUDA and OpenCL programming.
By the end of the course, students will be able to:
Researchers and programmers interested in porting scientific applications or use efficient post-process and data-analysis techniques in modern heterogeneous HPC architectures.
A basic knowledge of C or Fortran is mandatory. Programming and Linux or Unix. A basic knowledge of any parallel programming technique/paradigm is recommended.
The lunch for the three days will be offered to all the participants and some grants are available. The only requirement to be eligible is to be not funded by your institution to attend the course and to work or live in an institute outside the Roma area. The grant will be 300 euros for students working and living outside Italy and 150 euros for students working and living in Italy (outside Roma). Some documentation will be required and the grant will be paid only after a certified presence of minimum 80% of the lectures.
Further information about how to request the grant, will be provided at the confirmation of the course: about 3 weeks before the starting date.
Coordinating Teacher: Dr. L.Ferraro