On Fri, 2 Oct 2026 11:59:18 GMT, Lukasz Kostyra <[email protected]> wrote:
> > By the way, if you use this to measure performance differences between D3D9 > > and 12 I'd be interested to know the results. > > I moved the app to the latest D3D12 version we have and it actually revealed > a problem with D3D12 and multiple 3D lights. Very nice. For correctness, I'd use the Models since you can texture them and they have a volume. > We tested multiple lights on D3D12 at an earlier point, so I suspect > something broke in recent changes. Regardless, the problem seems visual so I > would assume the framerate is indeed higher - avg 212FPS on D3D vs avg 244FPS > on D3D12 (both cases have 10000 meshes and all three point lights enabled > with range 100). Yeah, when there is a visual bug we can't trust the measurements because we're probably not measuring the same thing. > I'll do more testing on D3D12 with this when we merge the update to the repo > =) Great. I have some fixes for D3D9 and it would be good to look if 12 inherited these issues. And if someone can answer my [email](https://mail.openjdk.org/archives/list/[email protected]/thread/VPZB3Q6TJIQNBIEQ2OLSAUZEYTJPFOAU/) about the pipelines it would be helpful too. ------------- PR Comment: https://git.openjdk.org/jfx/pull/1387#issuecomment-5953950860
