PhotonGI cache

Discussion related to the LuxCore functionality, implementations and API.
User avatar
Sharlybg
Donor
Donor
Posts: 3101
Joined: Mon Dec 04, 2017 10:11 pm
Location: Ivory Coast

Re: PhotonGI cache

Post by Sharlybg »

If you want to match view from corona to luxcore. I will suggest you to start by placing a new camera outside for both renderer and try to make them look simillar by playing with light power, considering you are using a linear camera for both without no extra hard coded pre processing phase.
Then when both camera match put them inside your interior.
Support LuxCoreRender project with salts and bounties

Portfolio : https://www.behance.net/DRAVIA
User avatar
Sharlybg
Donor
Donor
Posts: 3101
Joined: Mon Dec 04, 2017 10:11 pm
Location: Ivory Coast

Re: PhotonGI cache

Post by Sharlybg »

About caching and glossy shader. I wonder how we can evaluate before rendering if a shader average reflection roughness will be under or higher than cache setting when we use texture and band node instead of flat number.
Support LuxCoreRender project with salts and bounties

Portfolio : https://www.behance.net/DRAVIA
User avatar
Dade
Developer
Developer
Posts: 5672
Joined: Mon Dec 04, 2017 8:36 pm
Location: Italy

Re: PhotonGI cache

Post by Dade »

Sharlybg wrote: Sat Mar 02, 2019 8:58 pm About caching and glossy shader. I wonder how we can evaluate before rendering if a shader average reflection roughness will be under or higher than cache setting when we use texture and band node instead of flat number.
At the moment, the average roughness texture value is used.
Support LuxCoreRender project with salts and bounties
User avatar
lacilaci
Donor
Donor
Posts: 1969
Joined: Fri May 04, 2018 5:16 am

Re: PhotonGI cache

Post by lacilaci »

Ok, so the biggest problem in the scene that was being compared was tonemapper turned off. I don't know what it does to rendering when it is turned off but I don't think it is a good idea and you should probably use it to get proper exposure before you render final image.

Here is the setup:
Screenshot from 2019-03-03 07-50-39.png
Another thing is that I'm pretty sure you made gi cache pretty much useless with your settings (especially 2m lookup radius). This is gonna keep happening and it's just another proof that those settings will have to be hidden from users as soon as they can.

here are settings I used:
Screenshot from 2019-03-03 07-46-39.png
Screenshot from 2019-03-03 07-46-39.png (11.31 KiB) Viewed 4298 times
Performance is much better now but there is still a lot wrong with this scene, it's a complete mess with random object outside scene and unnecesary clutter, bad shading on objects, null materials(those probably don't do harm but why not use archglass).... I'm not gonna dig further

Try to play with clamping values next to gain better performance if you want...
andreymd87
Posts: 22
Joined: Mon Jun 04, 2018 10:59 am

Re: PhotonGI cache

Post by andreymd87 »

Dade wrote: Sat Mar 02, 2019 8:26 pm Sincerely, pretending to have the same outputs from 2 totally different rendering engines because you are inserting the "same" inputs is a veeeeery long shot.
I mean, are the models for light sources, materials, textures, camera, film response, etc. the same ? Are all the input values expressed with the same physical units ?
I wouldn't loose my sleep over that.
i agree with you. but not totally :) both rendering engines are physically accurate , that makes sense to give almost same result. ofc i used same outputs for light. same diffuse or glossy color. same untonemapped method. that's why i compared the raw exr images. but even this maybe is not a true case to loose your sleep over that :) also let's keep in mind that this version is still alpha. we dont know what changes might be there later. so i wouldnt put a cover on this and will follow for the new builds.
User avatar
lacilaci
Donor
Donor
Posts: 1969
Joined: Fri May 04, 2018 5:16 am

Re: PhotonGI cache

Post by lacilaci »

andreymd87 wrote: Sun Mar 03, 2019 8:29 am we dont know what changes might be there later. so i wouldnt put a cover on this and will follow for the new builds.
If Dade doesn't know what changes might there be later, then only god does :D
andreymd87
Posts: 22
Joined: Mon Jun 04, 2018 10:59 am

Re: PhotonGI cache

Post by andreymd87 »

lacilaci wrote: Sun Mar 03, 2019 8:32 am
andreymd87 wrote: Sun Mar 03, 2019 8:29 am we dont know what changes might be there later. so i wouldnt put a cover on this and will follow for the new builds.
If Dade doesn't know what changes might there be later, then only god does :D
i said about us, users :D
User avatar
lacilaci
Donor
Donor
Posts: 1969
Joined: Fri May 04, 2018 5:16 am

Re: PhotonGI cache

Post by lacilaci »

Talking about changes ,I still wonder if it wouldn't be worth it building cache for all materials for diffuse rays and use glossiness threshold for glossy rays only.
User avatar
Dade
Developer
Developer
Posts: 5672
Joined: Mon Dec 04, 2017 8:36 pm
Location: Italy

Re: PhotonGI cache

Post by Dade »

andreymd87 wrote: Sun Mar 03, 2019 8:29 am ofc i used same outputs for light. same diffuse or glossy color. same untonemapped method. that's why i compared the raw exr images.
How do you know Lux and Corona use the same glossy model ? Lux uses a Schlick's approximation (https://en.wikipedia.org/wiki/Schlick%27s_approximation) while Corona could use a GGX distribution (like Cycles do ?). The end result is totally different and requires different input values.
You are literally comparing apples to oranges without having Corona sources and checking the implementation of every single items (light sources, materials, textures, camera, film response, etc.).
Support LuxCoreRender project with salts and bounties
andreymd87
Posts: 22
Joined: Mon Jun 04, 2018 10:59 am

Re: PhotonGI cache

Post by andreymd87 »

Dade wrote: Sun Mar 03, 2019 10:20 am
andreymd87 wrote: Sun Mar 03, 2019 8:29 am ofc i used same outputs for light. same diffuse or glossy color. same untonemapped method. that's why i compared the raw exr images.
How do you know Lux and Corona use the same glossy model ? Lux uses a Schlick's approximation (https://en.wikipedia.org/wiki/Schlick%27s_approximation) while Corona could use a GGX distribution (like Cycles do ?). The end result is totally different and requires different input values.
You are literally comparing apples to oranges without having Corona sources and checking the implementation of every single items (light sources, materials, textures, camera, film response, etc.).
then make lux to use same GGX :D
Post Reply