Finally had a bit of time to play with multi-GPU deconvolution. Using @truenorth-ia.bsky.social and @haesleinhuepf.bsky.social previous work. Here it is in action in #BigDataViewer using async and lazy processing. Look at the two poor laptop GPUs going as fast as they can: - ThreadSky

About ThreadSky

nkiaru.bsky.social • 47 days ago

Finally had a bit of time to play with multi-GPU deconvolution. Using @truenorth-ia.bsky.social and @haesleinhuepf.bsky.social previous work.

Here it is in action in #BigDataViewer using async and lazy processing. Look at the two poor laptop GPUs going as fast as they can:

Comments

nkiaru.bsky.social•47 days ago

Hopefully, that's 2 PR away from an 'easy' use in #Fiji.

preibischs.bsky.social•47 days ago

Is the code public already? 😊

preibischs.bsky.social•47 days ago

Very cool!!

nkiaru.bsky.social•46 days ago

Yes, but unmerged. It's the code in here:
https://github.com/NicoKiaru/clij2-fft/tree/multi-gpu
that depends on this PR:
https://github.com/clij/clijx-faclon-heavy/pull/3

nkiaru.bsky.social•30 days ago

1 PR (and a release) away!
https://github.com/clij/clij2-fft/pull/39

nkiaru.bsky.social•27 days ago

Release: one credential away (https://forum.image.sc/t/scijava-maven-deployment-for-clij2-clijx/107907/4) and we still need to build some native lib for mac (https://github.com/clij/clij2-fft/issues/38). The admin cost of modern software + not having WORA.

truenorth-ia.bsky.social•27 days ago

I've been debating for a while whether I should just buy a Mac for testing. I know there are Github CI tools that could do a Mac build, and AWS VMs but the former may not allow you to open up Fiji and verify the linking works, the latter will be more expensive than buying a Mac long term.

Posting Rules

Be respectful to others
No spam or self-promotion
Stay on topic
Follow Bluesky's terms of service

Comments

Posting Rules

Reply