| ID: | 2625 |
| From: | ge...@igor2.repo.hu |
| Date: | Wed, 30 Jan 2019 14:41:41 +0100 (CET) |
| Subject: | Re: [pcb-rnd] poly clipping performance optimization - padstack, |
| in-reply-to: | 2624 from Gabriel Paubert <pa...@iram.es> |
On Wed, 30 Jan 2019, Gabriel Paubert wrote: > >I suspect I'm trying to do things with well more than 10k edges, perhaps >100k, perhaps 1e6. This was one of the reasons for which, in a previous >board with pcb, I implemented rectangular clipping of rectangular pads. >The other reason was that it was more consistent with soldermask, which >was rectangular, and in the end a slightly misaligned solder mask could >result in exposed copper, leading to shorts. With padstacks you don't need a patch anymore for the rectangular solder mask cutout part - although that wouldn't do poly clipping normally. For the cutout shape, I have long term plans to add options, but it's a rather big and non-trivial change if we don't want to shoot ourselves in the foot, like the current code makes us, so it won't happen soon. The preliminary version of the idea is, in short: introduce (optional) negative nominal padstack shapes. If you need a clearance and the padstack has a nominal clearance shape, it is bloated up by the clearance value and used as the cutout shape. Why multiple? To cycle through differen thermal styles. However, that will need some extra thoughts so we don't need to violate the "one object per shape" rule. (Why it won't happen soon: I need to think a lot about the details to get a long term maintainable setup. And I don't want to touch the data model with anything major for a few releases so we can stabilize everything and finish fixing all the pending features/bugs.) HTH, Igor2
Reply subtree:
2625 Re: [pcb-rnd] poly clipping performance optimization - padstack, from ge...@igor2.repo.hu