Support devices with limited framebuffer attachments

[micahscopes]

The GPU animation is using something like 8 framebuffer attachments, but a lot of mobile devices (especially iOS ones it seems) don't support more than 2.

In the long term it'd be nice to do all the GPU interpolation with just one big texture, but in the meantime a simple workaround could be to just disable GPU animation for devices with < 8 framebuffer attachments.

[micahscopes]

The fallback workaround is implemented (disabling interpolation for devices with < 8 framebuffer attachments)

[derhuerst]

not sure if this is related, but i can't see the graph with an iphone se 2nd gen, running ios 16.6.1

[serapath]

Also another friend wasn't able to see the visualization on their device.
He's working on cabal chat (another dat-ecosystem project).

Can we somehow collect more details on this problem?
If the framebuffer attachements are the main issue, how much effort would it be to fix this or implement fallback logic when a device does not support it?

We also have a small backlog regarding features and we will very likely receive another tiny grant to be able to define some new projects or upgrades to existing project (like webscape wanderer). Would you be up for something like that?

We just would need to work a little bit on a plan and potential scope.
The new grant might drop in 4-8 weeks from now, so if we start planning now we might be ready by then :-)

[micahscopes]

Hey @serapath! Did you manage to get some information about his device? Cabal chat is awesome by the way ✅

I've been wanting to tackle this. For a while I thought I'd modify the gpu-io library to use transform feedback instead of framebuffer attachments. The other option would be to use transform feedback with three.js custom buffer objects (I forget what they're called off the top of my head).

The point is that I'd like to refactor the GPU compute stuff to use transform feedback instead of relying on all those framebuffer attachments.

That said, there might be some other issues with iOS devices but at this point I'm not sure what those are.

Alongside the compute refactor, I'd like to implement camera interpolation on the GPU, so that even if a node is moving the camera can still track to it. Currently if the layout is still computing the camera will lock to a static position where that node is at the moment but the node will drift along and no longer be centered.

Anyhow, I'll get in touch with you on discord and we can meet and discuss this possible funding 😊

[serapath]

i reached out to @cblgh and they might know what the error message is like.

---

In terms of codebase, it wpupd be cool if webscape wanderer could be 2 things.

a module you can require and a static githubpage here to use the module and provide data to the module call and it returns a dom element, which displays the demo like now.

the opts passed to webscape wanderer to the module call might take additional params, e.g. a callback, which provides an api object to control the visualization maybe?

---

Additionally, do you think you could take a look at dat-garden-rake and run it to acquire a new snapshot of data next to the old one to check if it works with the visualization?

I would not want to lose the previous dataset, but be able to switch between them. a timeline about how the graph evolved from one to the next datasnapshot 🙂

[cblgh]

@serapath @micahscopes i couldn't see it on windows using firefox when it was embedded on the new website nina and serapath have been stewarding. my tab briefly froze and then the frame where the viz should be just displayed i think a blank/black bg?

[cblgh]

some data from when i went to the deployed version of this repo (click for big img?)

and shows a blank canvas