Operating Class(简称 OP Class)是 IEEE 802.11 标准中定义的一个 逻辑概念,它是一个监管域相关的频谱配置打包标签,它把 频段 + 信道集合 + 信道间距/带宽 + 其它行为限制 用一个数字编码起来,方便设备在不同国家和地区快速理解“哪些信道可以合法使用”。

Operating Class (abbreviated as OP Class) is a logical concept defined in the IEEE 802.11 standard. It is a regulatory-domain-related spectrum configuration packing label that encodes frequency band + channel set + channel spacing/bandwidth + other behavior restrictions into a single number, so that devices can quickly understand “which channels can be legally used” in different countries and regions.

WikiWLAN

本文记录如何在 Ubuntu 24.04 上编译 AOSP,并使用 Cuttlefish 模拟运行编译好的系统镜像。

在中国想要编译并模拟运行 AOSP,最好首先准备好一个质量不错的代理。因为无论是同步代码还是安装 Cuttlefish,都需要访问一些被屏蔽的服务器。

This article records how to compile AOSP on Ubuntu 24.04 and run the compiled system image in a Cuttlefish virtual device.

If you want to compile and emulate AOSP in China, it’s best to prepare a good-quality proxy first. Whether it’s syncing the source code or installing Cuttlefish, you’ll need to access some blocked servers.

WikiAndroid

用户空间 (Android init/ueventd)内核 (firmware_loader)驱动程序用户空间 (Android init/ueventd)内核 (firmware_loader)驱动程序alt[找到 firmware][未找到]调用 request_firmware()查找 /lib/firmware返回 firmware触发 sysfs fallback发送 ueventinit/ueventd 接收 uevent根据 ueventd.rc 依次查找/vendor/firmware/firmware/image...通过 sysfs fallback 提供 firmware返回 firmware
Userspace (Android init/ueventd)Kernel (firmware_loader)DriverUserspace (Android init/ueventd)Kernel (firmware_loader)Driveralt[Firmware found][Not found]Call request_firmware()Look up /lib/firmwareReturn firmwareTrigger sysfs fallbackSend ueventinit/ueventd receives the ueventSearch /vendor/firmware/firmware/image... in orderaccording to ueventd.rcProvide firmware via sysfs fallbackReturn firmware
WikiLinuxAndroid

上一次滨海路徒步还是 2021 年,在疫情封校的间隙,实验室组织了一次徒步活动。

时隔 4 年,再走滨海路。

但人数从 20 多人变成了 2 人。

The last Binhai Road hike was back in 2021, during a break from the COVID campus lockdown, when the lab organized a hiking activity.

Four years later, we walked the Binhai Road again.

But the group had shrunk from over 20 people to just 2.

结绳记事

背景

工作学习要用 Ubuntu,打游戏要用 Windows,双系统的方案开关机不太方便,而且也不是很稳定,这一点在读研期间就尝试过。

听说 WSL2 可以在 Hyper-V 上直接运行 Linux(微软定制版 kernel),性能比 WSL1 有了很大提升。

所以本文要测试一下在 Windows 11 上的 WSL2 和原生 Windows 11 系统上进行深度学习训练的性能差异。

B 站视频中也查到了相关的测试,基本上是 WSL2 比 Windows 11 快很多,百分之25?

但本着求真务实的态度,还是自己测试一下。

Background

For work and study I need Ubuntu; for gaming I need Windows. The dual-boot approach is inconvenient to switch between and not very stable either — I tried that during grad school.

I heard that WSL2 can run Linux directly on Hyper-V (with Microsoft’s customized kernel), with a big performance boost over WSL1.

So this article tests the performance difference in deep learning training between WSL2 on Windows 11 and native Windows 11.

I also found related tests in Bilibili videos — basically WSL2 is much faster than Windows 11, 25%?

But in the spirit of seeking truth from facts, I decided to test it myself.

WikiPython科研