This is insanely slow given its 200+GB/s memory bandwidth. As a comparison, I've tested GPT OSS 120B on Strix Halo and it obtains 420tps prefill and >40tps decode.
Not very unique to Japan. You can also see these techniques in other East Asian countries like China, Korea. Japanese guys are good at promoting their cultures anyway.
I am not 100% sure that all of the mitigation overhead
comes from syscalls, but it stands to reason that a lot
of it arises from security hardening in user-to-kernel
and kernel-to-user transitions.
Will io_uring be also affected by Spectre mitigations given it has eliminated most kernel/user switches?
And did anyone do a head-to-head comparison between io_uring and DPDK?
I'm using Okular and I just configured Okular to hide most panels and only leave the scroll bar on the window so I think it's minimalist as well. I will use Ctrl+M to toggle the menu bar when I need other functionalities.
Cuckoo hash must be a great example. I was interested in the problem solved by this paper solely because I learnt the analysis of time complexity of cuckoo hash just a few months ago.
A simple side note: Most people in China knows what happened at Tiananmen Square in 1989 although CCP has made much effort to conceal it. But almost none of them believes Uyghurs are under genociding. Instead, they believe Chinese government has been making effort in countering terrorism and keeping peace in Xinjiang for over 30 years.
That makes sense. ByteDance has many employees who have experiences in competitive programming. These guys tend to be more sensitive and aggressive in algorithmic changes.
