High-performance QEMU Instruction and Memory Access Tracing
Вставка
- Опубліковано 14 тра 2022
- Here's a short teaser for a high-performance QEMU introspection library I'm working on. It is about a 20-40% slowdown over stock qemu-user for recording full PC traces as well as memory (addr + val) traces. Super exciting! In this video I'm emulating a MIPS build of `mpv` playing Bad Apple!
This can achieve about 2 x86 cycles per MIPS instruction with full memory and instruction tracing, which was honestly really hard to get. Stay tuned for more info! - Наука та технологія
Awesome, dude! Love seeing your work, as it is always inspiring.
Hell yeah, looking forward to the vods on this project
let's hope together, even if thats never going to happen
Out of curiosity, why is this so significant?
@@boball1228 Emulating something on qemu with full memory and instruction tracing is really slow ussually, the fact that his implementation can even play a video is amazing
This is nutty. I love it
I really dig visualization software and this is blazing fast
incredible results
absolutely insane
time to duct tape machine vision to afl-fuzz to make memory access go brrrr
this thing fucks
okey cool, where are the logs ?
I’ve got so many questions…
Does this mean you can record a frame of my software and visualize physical (ram and cpu?) memory reads/writes instruction-by-instruction? I can imagine one way to visualize it is a zoomable memory canvas with known data structures mapped in. You could do clipping to avoid doing the whole scan too. Again, not sure what I’m looking at nor the limits, but would be cool for debugging.
Not to be rude, but I don’t get it. What’s the point of all of that?
ᑭᖇOᗰOᔕᗰ ✌️