>
> The speed obtainable by lossless JXL is significantly better than what the
> email above suggests - first of all, jxl_cli's --speedtest option will run
> two decodes by default.


I have verified it and this does seem to be the case. So make it a *15x*
slowdown vs lossless WebP. That is still astronomical, and my point still
stands.

I've re-measured on my laptop with 'taskset -c 0 target/release/jxl_cli
--num-threads 1 --num-reps 1 --warmup-reps 0 --speedtest' and I'm seeing a
20x slowdown in practice. Not sure why it's higher than the theory - the
lack of AVX-512 on this machine is to blame.

Moreover, the encoder can be configured to prioritize decoding speed over
> maximum compression -- for example, setting --faster_decoding 1 does not
> change the compressed size significantly, but doubles decoding speed.


Then perhaps only fast-decoding lossless mode should be supported in web
browsers. A 7.5x slowdown instead of 15x might be excusable, depending on
the use case.

Also, note that the linked image is almost 100 megapixels and strongly
> benefits from multi-threaded decoding.


Multi-threading helps with latency on some systems but does not
meaningfully affect power usage. And given that a webpage contains not only
images and the CPU is already busy processing the rest of it,
multi-threading is less of a boon to latency in a web browser than in an
image viewer.

I do not mean to derail this thread into a discussion of the details of the
lossless mode and what can be done about it. Rather, I believe the lossless
compression mode of JPEG XL has not received enough attention on its
cost-benefit analysis (unlike the lossy mode
<https://cloudinary.com/blog/jpeg-xl-and-the-pareto-front>), and should not
be included in a stable release in web browsers until compelling evidence
for its use cases and trade-offs is provided.

-- 
You received this message because you are subscribed to the Google Groups 
"blink-dev" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to [email protected].
To view this discussion visit 
https://groups.google.com/a/chromium.org/d/msgid/blink-dev/CABaUXi4%2BVHb3VWNrQC0CyCpKWd13FOz2ak%2BaJRPrgTVhW1%2BhyQ%40mail.gmail.com.

Reply via email to