PAROLE CHIAVE API MANAGEMENT, GRAPHIC PROCESSING UNITS, PARALLEL COMPUTING, ARCHITECTURE SOFTWARE DISTRIBU, PROCESSORI, SOFTWARE DEVELOPMENT

RIFERIMENTI JOSIE ESTEBAN RODRIGUEZ CONDIA, MATTEO SONZA REORDA

RIFERIMENTI ESTERNI Juan David Guerrero Balaguera

TIPO TESI EXPERIMENTAL, EXPERIMENTAL/MODELLING

DESCRIZIONE Currently, modern generations of GPU-accelerated computing (e.g., from embedded systems to data centers and high-performance computing) are crucial for several application domains. These systems commonly resort to strategies to monitor and observe their operational status, involving software or hardware mechanisms. This thesis seeks to explore the potential of software codes to monitor and verify the status of GPUs. The thesis also aims to develop small stress programs to push GPUs to their limits.

Objectives:

- Investigate and understand the GPU structures used to monitor and verify their operational status.
- Develop innovative programs to leverage them effectively.
- Develop focused programs for GPUs to stress and push them on execution limits.

What could you learn?

- Develop skills in using GPUs to develop and deploy parallel programs.
- Develop skills in using debugging and profiling tools for GPUs.
- Understand of GPU structures and their relevance in optimizing performance and monitoring their status.
- Develop skills for deploying GPU programs in servers and data center systems.

The thesis provides an opportunity for collaboration with a local company in the sector (Intelsa SanPaolo) to explore and validate the developed solutions.

CONOSCENZE RICHIESTE General knowledge in Computer's/Processor's architecture
General Knowlegde in C and C++
General knowledge in scripting languages (bash or python)
Knowledge in Parallel programming for GPUs (e.g., with CUDA) is not required

SCADENZA VALIDITA PROPOSTA 26/02/2025