Is there a fast way to skip the work in PTCL?

Hello there

When I was using vello, I found that it works very well when the stuff I want to draw is in full screen.

However when I zoom out to make it very small, the speed starts dropping down quickly.

The root cause is that I want to draw too much stuff in the same pixel.

I know there is a lot of pre-culling work been done, such as if the bbox of a path to less than a pixel, just skip it.

But I think there might be some edge cases, a lot of stuff is been queued in a same small area.

The truth is when a pixel is drawing stuff according to the PTCL, most of the later command is an opaque thing, which can override the current one, which means if there is 1000 things to draw, the first 999 is useless.

I am thinking if there is any optimize way to speed this process up.

Thanks a lot!

The issue does not name a file, test, or entry point; begin by locating PTCL processing in vello and profiling the zoomed-out case described. Compare the cost of dense commands in one small area with the existing pre-culling behavior, and define done as a measured reduction in that workload without changing rendered output.

Written by the indexing model from the issue text.