paddle-mobile将开发一个新的版本，为了：

• 支持Paddle Fluid
• 更小的体积
• 更快的速度
• 更广泛地支持各种arm设备和终端平台

因此，paddle-mobile将会进行一次完全的重构，从框架设计到代码开发。在新版本开发完成后，将会删除原有mobile-deep-learning的代码。新版本开发期间，代码将放置在paddle-mobile repo的一个子目录下。对于子目录的命名，目前有3个候选，希望大家投票：

• 👍 zygote
• 👎 paddle
• ❤️ fluid

有更好的建议，请回复issue。相关issue：#121
另外，待新版本代码开发完成，移除老版本mobile-deep-learning代码后，新版本代码将移出子目录，直接放置在paddle-mobile repo下面。

MeshNormalSmooth() takes a vertex and iterates all the others, looking for vertices with a distance < epsilon.
Then it calculates their normals and sums it.

That's not a good idea, IMO:

• Two close vertices can be not physically linked, I don't think summing normals makes sense.
• It's quadratic.

A simpler and faster way to smooth a model could be:

• Iterating through each triangle

I was talking with Gabor about contributing with LVGL, and one of the points he was interested in was in writing unit and functional tests. I'm interested on contributing, as I use LVGL in some of my projects. If you'd like, we could discuss the following steps here.

Is your enhancement proposal related to a problem? Please describe.
Refer to: zephyrproject-rtos/zephyr#28932

Describe the solution you'd like
Refer to: zephyrproject-rtos/zephyr#28932 (comment)

Describe alternatives you've considered
NA

Additional context
NA

Implement Python 3 bindings, so wasm3 can be easily used from Python.

Resurrecting the second bullet from #1334

• Figure out how to fix favicon / logo more generally (here's a start https://github.com/tock/tock/tree/doc-updates)

Tested on MagTag, PyBadge, CPX and QT Py. The issue is not present on QT Py which requires adafruit_pypixelbuf. It is present on all other boards, which may narrow it down to an issue in _pixelbuf.

Used simple code that uses board.NEOPIXEL to turn the pixels red, and then tested brightness:

• Setting brightness = 0 makes the on-board NeoPixels max brightness. They are the appropriat

On Linux, packets that stay local to the machine (moving from the host to a container for example) don't always have their checksum computed, see for example https://patchwork.ozlabs.org/patch/261822/.
The proper fix would be to check for TP_STATUS_CSUMNOTREADY, but an easy workaround is to have a way to change the ChecksumCapabilities or DeviceCapabilities.

As a rust noob here my solution for