Hyeonggon Yoo
Posts
474Following
89Followers
94- Ex-Intern @ NVIDIA Korea (Security System Software) (2024.06-2024.11)
- Ex-Intern @ Panmneisa (CXL emulation stuff) (~2023.12)
- Undergraduate majoring CSE (estimated graduation: Feb. 2025)
- Working as reviewer at Linux Slab subsystem
- Born in August 6, 2000
Opinions are my own.
My interests are:
Memory Management,
Computer Architecture,
Circuit Design,
Virtualization
@ljs lol I recall reading it in "Linux Kernel in a Nutshell" book and I was shocked in the same way
(picture taken from the book http://www.kroah.com/lkn/ under CC BY-SA 2.5)
(picture taken from the book http://www.kroah.com/lkn/ under CC BY-SA 2.5)
2
0
2
So I measured one change (the SLUB numa locality change) at a time and the data seems to be nice I think? (at least on my old Broadwell machine with 2 sockets)
The "Neither a win nor a loss" situation was because one change was a win and the other change was a loss.
But it's still synthetic hackbench and I wonder how it would perform on network benchmarks - but not sure if I could benchmark this enough on local env cuz I only have 2.5Gbps lan cards.
The "Neither a win nor a loss" situation was because one change was a win and the other change was a loss.
But it's still synthetic hackbench and I wonder how it would perform on network benchmarks - but not sure if I could benchmark this enough on local env cuz I only have 2.5Gbps lan cards.
0
0
0
Hyeonggon Yoo
hyeyoo
Edited 1 year ago
watching a tutorial on Oppenheimer before going to the cinema
0
0
1
Hyeonggon Yoo
hyeyoo
Edited 1 year ago
Making a small progress on one of my TODOs: run automatic builds & tests using CI tools for MM development kernels
This is a jenkins configuration matrix specially targeting SLUB to cover various SLUB configurations. Only boot test is done atm.
There are many milestones on this project, which demands more time :P
This is a jenkins configuration matrix specially targeting SLUB to cover various SLUB configurations. Only boot test is done atm.
There are many milestones on this project, which demands more time :P
0
0
2
Hyeonggon Yoo
hyeyoo
Edited 1 year ago
When there are no free pages in free list of the requested order, buddy allocator splits a larger page into smaller ones, and allocates a one of the split pages.
TIL: The way buddy splits large pages (and allocating first or last part of a large page) during the expand() function significantly impacts the mergability of the scatter-gather list for IO. The likeliness that pages are allocated in PFN order is important for mergability. This is so subtle and I didn't realize this :(
Consider a scenario where there is only an order 2 block, and four order 0 requests are served. In the diagram above, they are allocated in PFN order, while in the other diagram, they are allocated in the reverse order.
https://git.kernel.org/pub/scm/linux/kernel/git/tglx/history.git/commit/?id=c75b81a5da09be94099cf7b46bdb32fedc8ce71a
And in the history this thing has introduced a regression because the order was not preserved (by mistake) when pages are allocated from PCP lists with __GFP_COLD ;)
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=e084b2d95e48b31aa45f9c49ffc6cdae8bdb21d4
TIL: The way buddy splits large pages (and allocating first or last part of a large page) during the expand() function significantly impacts the mergability of the scatter-gather list for IO. The likeliness that pages are allocated in PFN order is important for mergability. This is so subtle and I didn't realize this :(
Consider a scenario where there is only an order 2 block, and four order 0 requests are served. In the diagram above, they are allocated in PFN order, while in the other diagram, they are allocated in the reverse order.
https://git.kernel.org/pub/scm/linux/kernel/git/tglx/history.git/commit/?id=c75b81a5da09be94099cf7b46bdb32fedc8ce71a
And in the history this thing has introduced a regression because the order was not preserved (by mistake) when pages are allocated from PCP lists with __GFP_COLD ;)
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=e084b2d95e48b31aa45f9c49ffc6cdae8bdb21d4
0
0
1
Hyeonggon Yoo
hyeyoo
Edited 1 year ago
I've just built my first 2 socket server (36C/72T, DDR4 256GB ECC/REG RAM) for experimenting with NUMA.
It's an old xeon server (Broadwell), but enough to play with!
It's an old xeon server (Broadwell), but enough to play with!
1
0
1
