When browsing JXL files, it seems that you cannot switch between images.

[Huangzhiqiang]

This image viewer is really well-made! Great job!

Besides the issue with switching between images in JXL format, opening files in this format seems a bit slow. It's faster when I use Nomacs, so there's likely an issue somewhere that's preventing the decoder from performing at its full potential.

[justnullname]

Thanks! I'm currently investigating the JXL performance issue. Could you please press the i key to open the Info Panel and share a screenshot with me? I plan to release an update with fixes in the next few days.

[Huangzhiqiang]

Thank you for the quick reply. Below are my screenshots. I suspect the slow JXL decoding is due to the lack of multithreading, as the CPU usage observed during opening is lower compared to Nomacs.

The JXL image switching issue has been fixed in the latest version, but it seems that the update process is broken, as I can't click the update button, which is causing the problem.

[justnullname]

This image viewer is really well-made! Great job!

Besides the issue with switching between images in JXL format, opening files in this format seems a bit slow. It's faster when I use Nomacs, so there's likely an issue somewhere that's preventing the decoder from performing at its full potential.

QuickView 2.1.0 is out! This update brings multi-threaded JXL decoding, 90MP image load time reduced from ~2.8s to ~0.4s (6.7x faster) on 8-core CPUs. Download now to give it a try!

[Huangzhiqiang]

I tested the new version, and JXL decoding is much faster now. Thank you very much!
Additionally, I noticed that if I continuously switch between images, it seems that multiple decoding instances compete for CPU resources, leading to slower decoding and excessive memory usage (especially with formats like JXL that support multi-core decoding). You might consider adding a decoding task cancellation mechanism and serialization control (mutex locks) as needed.

[justnullname]

Thank you for the valuable feedback! I am currently exploring a fix that implements a cancellation mechanism and task scheduling to address the resource contention. This will also involve refactoring the pre-fetching system. Please look forward to the next release!