Skip to content

Support loongson, loongarch, 龙芯CPU' translates to English as: 'Support loongson, loongarch, Loongson CPU #2689

Closed
@winlinvip

Description

@winlinvip

LOONGARCH

The multimedia open-source project that Loongson has already adapted: Media

  • Audio and video application development framework, audio and video codec projects: FFmpeg, X264, OpenH264, libvpx, libyuv
  • Basic compression and decompression, image codec, and graphics rendering projects: zlib, libpng, libjpeg-turbo, pixman, skia
  • High-performance mathematical libraries for signal processing, linear algebra, etc.: OpenBLAS, FFT

If you want to use Loongson servers for video cloud, you still need a streaming media server.

Build Loongarch SRS in Qemu

Loongson has a QEMU image that can be run on the local machine. Below are the steps to follow.

First, download the QEMU environment. For macOS, execute the following command. For other systems, refer to this link:

brew install qemu

Then, download the Loongson ISO image. It's still on the way, please wait...

LOONGSON 3A5000/3B5000

Loongson 3A5000/3B5000

The Loongson 3A5000/3B5000 is a general-purpose processor designed for personal computers, servers, and other information technology fields. It is based on the Loongson proprietary instruction set architecture (LoongArch®) and the LA464 microarchitecture, which further improves frequency, reduces power consumption, and optimizes performance. While maintaining pin compatibility with the Loongson 3A4000 processor, the frequency has been increased to 2.5GHz, power consumption has been reduced by over 30%, and performance has been improved by over 50%. The Loongson 3B5000, based on the Loongson 3A5000, supports multi-way interconnection.

image

ST for LOONGARCH

First, it is necessary to adapt ST to the Loongson CPU. SRS 5.0 already supports this, please refer to ST #24.

For the adapted code, please refer to the commit: ossrs/state-threads@db42969

Build for LOONGARCH

Additional compilation options for some dependent libraries need to be adapted, mainly for SRTP and Opus. Please refer to the commit: 5b44cc6

TRANS_BY_GPT3

Metadata

Metadata

Assignees

Labels

FeatureIt's a new feature.TransByAITranslated by AI/GPT.

Type

No type

Projects

No projects

Milestone

Relationships

None yet

Development

No branches or pull requests

Issue actions