23 ms·

Bounds Checking on GPU (2021) [pdf]

• scoopr 4y ago
  The link doesn't seem right, I think this might work better: https://futhark-lang.org/publications/hlpp20.pdf https://futhark-lang.org/publications/hlpp20.pdf
• anonymousDan 4y ago
  What's the overhead?
  • jsmith45 4y ago
    On average a 4-6 percent overhead for the most optimized version on the NVidia and AMD gpus tested, but which also has its own limitations. Which is not wonderful, but is not that terrible either.
    • anonymousDan 4y ago
      That's pretty damn good in my book! What are the main limitations out of interest? I have a student working on a project in this area. Also, regarding the threat model, presumably this focuses on spatial memory safety. Have you come across a need for temporal safety?