Any suggestions about possible optimizations for aarch64 (arm64) device? #61

New Issue

2021-07-22T17:04:53+02:00

andrewrush commented

2021-07-22 17:04:53 +02:00

(Migrated from github.com)

I switch backend to CPU or GPU (depends on CUDA-checking lines), copy SimSwap to Android device, install CPU version of dependencies via termux and it just works but too slow.

So I'm thinking about buying a NVIDIA arm device (such as Jetson Nano, Jetson Xavier etc.), but I don't know if CUDA will work on it.

In addition, the recent NVIDIA presentation opens up wide opportunities for the joint use of arm processors and RTX video cards.

Proof-of-work: https://www.youtube.com/watch?v=GZNlUCJNGok

Also powershell script which patches SimSwap as universal (CPU or GPU, no NVIDIA card required or any card, just 3Gb RAM (or less) and any processor, you should build or find compatible dependencies): https://pastebin.com/y1FxV8jA

In-house benchmark results:

I switch backend to CPU or GPU (depends on CUDA-checking lines), copy SimSwap to Android device, install CPU version of dependencies via termux and it just works but too slow. So I'm thinking about buying a NVIDIA arm device (such as Jetson Nano, Jetson Xavier etc.), but I don't know if CUDA will work on it. In addition, the recent NVIDIA presentation opens up wide opportunities for the joint use of arm processors and RTX video cards. Proof-of-work: https://www.youtube.com/watch?v=GZNlUCJNGok Also powershell script which patches SimSwap as universal (CPU or GPU, no NVIDIA card required or any card, just 3Gb RAM (or less) and any processor, you should build or find compatible dependencies): https://pastebin.com/y1FxV8jA In-house benchmark results: ![изображение](https://user-images.githubusercontent.com/8232067/126667092-ce6f39ee-1304-4026-bef3-2e4987f96141.png)

Sign in to join this conversation.

1 Participants

Notifications

Due Date

No due date set.

Dependencies

No dependencies set.

Reference: CalvinBackup/neuralchen-SimSwap#61