Download presentation
Presentation is loading. Please wait.
1
The Case for Operating System Services on GPUs
GPUfs
2
Why I chose this article
GPUs are increasing in performance compared to stagnating CPUs I’ve used dedicated programs for GPUs before It brings multi-threaded programs to a whole new level
3
What is GPUfs? A software layer providing native support for accessing host files on discrete GPUs
4
Background GPUs are built up of multiprocessors
GPUs are structured for parallelism No OS integration support for GPUs Used similarly to a function call, for operations All functions for GPUs had to be hand written and fine tuned.
5
GPUfs Thesis We have this incredible hardware
How can we tap into it What is preventing us from using it Why is GPUfs the answer “The time has come to provide GPU programs the useful system services CPU code already enjoys.“
6
Where is GPUfs currently?
GPUs are just peripherals Discrete vs Hybrid “GPUfs is a system software co-resident with GPU programs but is less intrusive than a complete operating system in that it has no active continuously running components on the GPU”.
7
What needs to be addressed?
Functions that accomplish more Virus scanning, searches GPUs are mathematically faster But still require CPU interference Big Three Separate Memory Massive Parallelism Non-preemptive hardware scheduling Main focus has only been a file-system layer
8
My thoughts When will GPUfs become important
Discrete GPUs are still relevant GPUs are used like co-processors When would I need a GPU-CPU partnership
9
Your thoughts Any questions?
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.