Re: [gem5-dev] remote GDB debugging

2018-01-16 Thread Gabe Black
So, as I've dug into this a bit, I realized that System objects aren't specialized per-ISA in SE mode, I think because the "System" is supposed to be a more abstract entity which doesn't need to set things up for a kernel to start. There is some merit to that, but it makes it harder to hook in the

Re: [gem5-dev] remote GDB debugging

2018-01-09 Thread Jason Lowe-Power
Hi Gabe, All of this seems reasonable to me. I've never really used the remote GDB feature, so I don't have any insight to give you. I strongly agree that multiple system objects can make sense (I have some patches which use this ability), so we shouldn't break that. For multithreaded GDB support,

[gem5-dev] remote GDB debugging

2018-01-09 Thread Gabe Black
Hi folks. As part of looking at refactoring how ISAs plug into things, I started looking at how remote GDB is set up. Currently the system object does that through its registerThreadContext function, where it queries the remote GDB port (set through a gem5 command line option), and then instantiate