Page 1 of 1

CPU BiDir, Metropolis, Different Settings and Systems

Posted: Wed Jan 30, 2019 7:42 am
by FarbigeWelt
i7, 2.93 means intel core i7 with 2.93 GHz (4+4 cores)
R2300U 2.0 means AMD Ryzen 3 2300U 2.0 mobile version (4 cores)
Win 10, H means Home, P means Pro (both with "2018 October" update)

S for Glass Homogen means Scattering size.

Run time, and system can be seen in file name with mouse over picture.

Further system details see signature, Win 10 H details follow in this thread.

Hope all other information are clear so far.
1
1
2
2
3
3
4
4
5
5
6
6
7
7
8
8
9
9

Short notice: background task

Posted: Mon Feb 11, 2019 5:24 pm
by FarbigeWelt
The past days and the following weeks, I‘m collecting notes and items from my former posts to update and recyle them if in my opinion still useful for one or the other user.
One resson is, forum’s search form is not very useful, as already mentioned by other forum members.

And sure, I am looking forward to seeing how latest developments improved luxcorerender output and speed, because my current yet unposted scenes take long times to render.

Re: Short notice: background task

Posted: Tue Feb 12, 2019 2:56 am
by wasd
FarbigeWelt wrote: Mon Feb 11, 2019 5:24 pm One resson is, forum’s search form is not very useful, as already mentioned by other forum members.
Forum's search usefulness is below zero. Because it's phpBB. Maybe "google custom search" could fix that?

Recommend to precise future post subjects

Posted: Tue Feb 12, 2019 7:28 am
by FarbigeWelt
wasd wrote: Tue Feb 12, 2019 2:56 am
FarbigeWelt wrote: Mon Feb 11, 2019 5:24 pm One resson is, forum’s search form is not very useful, as already mentioned by other forum members.
Forum's search usefulness is below zero. Because it's phpBB. Maybe "google custom search" could fix that?
Well, this idea to use an external search engine is a good one. The issue with it is theat I, and most other users too, do not change the subject of their replies. The implied connection of any post to its thread would not be lost in case of altering post‘s subject.

To get better hits with search engine I recommend to precise subject of future replies.

Re: CPU BiDir, Metropolis, Different Settings and Systems

Posted: Wed Feb 13, 2019 12:38 pm
by epilectrolytics
Thanks for your systematic work.
I probably will respond later with questions, but at the moment OIDN & PGI development with new updates every day are completely consuming my time and attention.

Re: CPU BiDir, Metropolis, Different Settings and Systems

Posted: Thu Feb 14, 2019 12:29 pm
by FarbigeWelt
epilectrolytics wrote: Wed Feb 13, 2019 12:38 pm Thanks for your systematic work.
I probably will respond later with questions, but at the moment OIDN & PGI development with new updates every day are completely consuming my time and attention.
Your recognition of my work and your graditude both I really appreciate.

I am looking foreword replying to your future questions.

Re: CPU BiDir, Metropolis, Different Settings and Systems

Posted: Fri Mar 01, 2019 2:11 pm
by epilectrolytics
It hurts me to turn down such a heartfelt invitation but because of the no-SDS-paths issue I decided not to use BiDir any longer.
While the caustics are nice and quick the reflections in and on the glass are all wrong or missing.
I don't know what will become of the caustic cache path tracer but at the moment it has an evil unadjustable blur ruining every attempt at serious work.
Also I realised that I really don't like the biased approach in spite of the speed gain.

Best wishes for your future work, cheers!

Re: CPU BiDir, Metropolis, Different Settings and Systems

Posted: Fri Mar 01, 2019 3:11 pm
by CodeHD
I also meant to respond to this at some point but forgot...

FarbigeWelt, could you elaborate a bit on the Scene and waht you wanted to test exactly?
My interpretation is you set up some glass cubes with various homogeneous internal scattering volumes, and send a laser into them?
All of this to test performance of the BiDir engine?

epilectrolytics wrote: Fri Mar 01, 2019 2:11 pm It hurts me to turn down such a heartfelt invitation but because of the no-SDS-paths issue I decided not to use BiDir any longer.
While the caustics are nice and quick the reflections in and on the glass are all wrong or missing.
I don't know what will become of the caustic cache path tracer but at the moment it has an evil unadjustable blur ruining every attempt at serious work.
Also I realised that I really don't like the biased approach in spite of the speed gain.

Best wishes for your future work, cheers!
Sounds like a fairly harsh opinion ;)
Although I am with you on the fact that there are issues, I don't think we should be expecting too much.
a) because no single rendering approach is appropriate for everything (BiDir, backward path, forward path)
b) there just is a development focus on the cache at the moment, that will be over at some point and other things can be addressed with higher priority again.

I wish I had more time to contribute myself, as I am very interested in all the raytracing stuff. But I have never worked with C++ on this level and I have to work myself into the whole structure of a project first. Too many other hobbies keep me from doing that so far :lol:

Re: CPU BiDir, Metropolis, Different Settings and Systems

Posted: Tue Mar 05, 2019 1:16 am
by FarbigeWelt
epilectrolytics wrote: Fri Mar 01, 2019 2:11 pm It hurts me to turn down such a heartfelt invitation but because of the no-SDS-paths issue I decided not to use BiDir any longer.
While the caustics are nice and quick the reflections in and on the glass are all wrong or missing.

Best wishes for your future work, cheers!
In one of my former posts I noted UniDir and BiDir have both their strengths.
And since appearance of PGIC I can agree in most caseswith lacilaci's opionion about BiDir resp. caustics, that BiDir can be avoided in most cases.

Thank you for the wishes.

At the moment, I am still collecting data in different areas, like programming languages: cpp, Python and some background; libraries: openCL, Vulkan; different API.

Re: CPU BiDir, Metropolis, Different Settings and Systems

Posted: Tue Mar 05, 2019 1:25 am
by FarbigeWelt
CodeHD wrote: Fri Mar 01, 2019 3:11 pm I also meant to respond to this at some point but forgot...

FarbigeWelt, could you elaborate a bit on the Scene and waht you wanted to test exactly?
My interpretation is you set up some glass cubes with various homogeneous internal scattering volumes, and send a laser into them?
All of this to test performance of the BiDir engine?
Purposes of these tests are
a) I like to understand BiDir settings' influence on performance.
b) Homogenous and heterogenous volumes did not work well with UniDir.