JPEG-XL libjxl 0.12 Brings More Performance Optimizations
([Free Software] 16 Minutes Ago
libjxl 0.12)
- Reference: 0001644217
- News link: https://www.phoronix.com/news/JPEG-XL-libjxl-0.12
- Source link:
A new release of libjxl is now available for this reference implementation of the JPEG-XL image format, including both image encode and decode support.
The libjxl 0.12 release debuted yesterday and added support for new tooling features, the new jxltran tool for extracting codestreams from JXLP boxes, and a "major overhaul" for faster decoding and better handling of progressive lossless images.
With JPEG-XL's libjxl 0.12 release, progressive lossless images are 30~40% smaller than before while with new multi-threaded handling there is a 2-5x performance improvement in encoding performance. Lossless images in general are decoding faster too and 30~80% smaller.
Libjxl 0.12 also improves encoding speeds by SIMDifying EstimateCost for a 5% performance improvement, up to 4x faster performance for accelerated modular encoding and decoding via SIMD optimizations, a "massive" overhaul to histogram encoding/decoding for more performance too, and a variety of other SIMD optimizations and performance improvements at large.
Downloads and more details on the many libjxl 0.12 changes can be found via [1]GitHub .
[1] https://github.com/libjxl/libjxl/releases/tag/v0.12.0
The libjxl 0.12 release debuted yesterday and added support for new tooling features, the new jxltran tool for extracting codestreams from JXLP boxes, and a "major overhaul" for faster decoding and better handling of progressive lossless images.
With JPEG-XL's libjxl 0.12 release, progressive lossless images are 30~40% smaller than before while with new multi-threaded handling there is a 2-5x performance improvement in encoding performance. Lossless images in general are decoding faster too and 30~80% smaller.
Libjxl 0.12 also improves encoding speeds by SIMDifying EstimateCost for a 5% performance improvement, up to 4x faster performance for accelerated modular encoding and decoding via SIMD optimizations, a "massive" overhaul to histogram encoding/decoding for more performance too, and a variety of other SIMD optimizations and performance improvements at large.
Downloads and more details on the many libjxl 0.12 changes can be found via [1]GitHub .
[1] https://github.com/libjxl/libjxl/releases/tag/v0.12.0