Better test case timings: I changed from doing entire linear write testing to just touching the pages, to take memory bandwidth out of the equation. VirtualAlloc takes 0ms on 4k pages, 6ms on 2mb pages. Fault handling takes 60ms on 4k pages, 0ms on 2mb pages. So that's that.
I fail to see how this is any different whatsoever to the driver having locked the entire 2mb page, which it could have done?
-
-
The system caps the number of locked pages so a driver can't just lock all the memory in the system. The problem here is that just 9 locked pages could prevent a 16GB system from paging in a 2MB page. I'll ask my owner to discuss this further on the blog.
-
If your owner is so inclined, I have posted an example repository for commentary. When run, it demonstrates the dramatic difference in cold-start performance between 4k and 2mb pages, where 4k pages are often 10x slower, etc.:https://github.com/cmuratori/largepages …
- Show replies
New conversation -
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.