Webgpu based force graph layout and rendering

[sid597]

This is how the current arch is:

• Load/Get all the nodes data
• Set initial offset, render the nodes
• So at this point the data is in memory, the rendering is handled by the webgpu
• On wheel or pan, calculate and reset the offset and zoom level
• since we are watching offset, Re-render based on this new data

For force directed graph

• we have to do some internal calculation for node positions
• occasionally render the node position in calculation step. There is a spectrum from: render every intermediate step and render once whole simulation is done.
• so there are 2 types of calculations that SHOULD be done in parallel to make the above possible: Run compute and render shader parallely, browser is single threaded but webgpu api? .
• The loop would be something like
  • load initial data
  • compute shader
  • update the initial data
  • based on some logic read the data into render

[sid597]

https://developer.nvidia.com/gpugems/gpugems2/part-iv-general-purpose-computation-gpus-primer