[Lldb-commits] [lldb] [lldb-server] Add accelerator plugin infrastructure for debugging hardware accelerators like gpus (PR #198907)

[David Spickett via lldb-commits]

DavidSpickett wrote:

Thanks, I get the picture now and I don't have any objections to it.

One thing I had been thinking about is how you knew which plugin to "select". For example if I see AMD and Nvidia plugins. But those breakpoints do that job I think. You have to hit one of those initial breakpoints to start using the plugin so if the program never hits it, or doesn't contain the symbol to break on in the first place, there's no overhead.

And if you have a program that uses multiple different types of accelerator, whether it picks one at runtime, or uses everything it can find, it sounds like this scheme would handle that too. Each accelerator would be an extra target.

https://github.com/llvm/llvm-project/pull/198907