[Bf-committers] 转发: GSoC 2018 Proposal for Freestyle Rewrite?

Wu Yiming xp8110 at outlook.com
Wed Feb 28 02:53:25 CET 2018


(Sorry list, the last mail is too big to be sent by the mail list server)
Hello Tamito,

Thank you for your email on your opinions! Here’s my answer to your questions. I have classes this morning so I can’t really write too much detailed things here, but everything including every possible solution to rendering (and including real time), and to what extent I can do will all be posted on my porposal page very soon.

Backward capability:
In current solution, All Freestyle configuration in Blend file can be capactible, tha is because my solution is based on the same principle that Freestyle uses. If we are going to redesign the line layer compositing, the related data may become obsolete. But we can always try to adapt that in do_version.

ViewMap Data Structure Replacement?
My line representation is different than Freestyle’s ViewMap, So it’s likely that this part will be replaced.

Realtime?
Very likely. I used to focus more on legacy vector render solution, but did some researched on real time as well. This morning I had a discussion with a NPR enthusiasm named Wasili, and we found that line style and technical-drawing-like accurancy may not be the usual requirement, In this case real time solution has become very very valid, and most part the implementation is straightfoward.

Performance Numbers?
During the test, numbers are screen captured but very few were documented in text. I will run a specific test again and make a compare table.
(just because it’s so easy to share images in Chinese IM software).

Python sctipting?
Style scripting is likely to be preserved, others may need further investigations, and api is suspected to change.

My proposal page is here (and I haven’t update that yet but very soon) http://www.wellobserve.com/#!NPR/2018Summer/Proposal.md
I will definally notify the list when I finish the detailed update on this. And I’d like to thank all devs and artists who gave their suggestions!

Best regards,
YimingWu

发件人: Tamito KAJIYAMA<mailto:rd6t-kjym at asahi-net.or.jp>
发送时间: 2018年2月27日 22:01
收件人: bf-committers at blender.org<mailto:bf-committers at blender.org>
主题: Re: [Bf-committers] GSoC 2018 Proposal for Freestyle Rewrite?

Hi,

I guess I'm supposed to say something to the proposed plans on Freestyle
improvements.

First of all I honestly appreciate the willingness of addressing the
long-awaited performance boosts in Freestyle.  I believe the community
expectation to the project if accepted would be very high.  I wish you
all the best.


General comments:

- The project plans look good to me.  The main goal of improving the
Freestyle rendering speed is relevant not only in the present 2.7x
development line but also in the new 2.8 series.

- How and to what extent do you preserve the backward compatibility?
Does this project expect to break existing Freestyle setups?

- Is it appropriate to base the proposed project on the 2.8 code base?
That is a kinda high-level question that needs to be addressed by the
2.8 core developers. (Ton?)


Specific comments:

- Do you have a plan to completely replace the so-called view map data
structure of Freestyle?  If you keep using the view map, how would you
integrate face-intersection lines into it?  (Currently feature edges in
Freestyle are edges of mesh faces.  Non-face edges are not considered as
feature edges.  Hence introducing non-face edges like face-intersection
lines is likely to require a major revision to the view map data
structure.)  Any chance to integrate other non-face line elements such
as grease pencil strokes?

- By line style preview, do you mean a real-time preview in the 3D view
port?  Or do you mean a preview icon such as those in the material and
texture properties windows?

- Any idea of a real-time line rendering in the 3D view port as expected
in the 2.8 development series?  How do the present project targets fit
to the 2.8 real-time requirement?

- Do you have any numbers indicating performance factors, such as
speedups, memory footprint reduction rates, and frame rates (FPS), based
on your stand-alone implementation with respect to the present Freestyle
in Blender?


Additional remarks:

- I don't personally mind if the project introduces major backward
incompatible changes (especially in view of the 2.8 development series).

- The Freestyle functionality I like the most is its programmability
with Python scripting.  If the project could preserve it in a backward
compatible way, that would be very much appreciated.


That's it for now.  I stay at your disposal for question-answering and
discussions.  I hope these comments help improve your proposal.

Best regards,

--
KAJIYAMA, Tamito <rd6t-kjym at asahi-net.or.jp>

_______________________________________________
Bf-committers mailing list
Bf-committers at blender.org
https://lists.blender.org/mailman/listinfo/bf-committers




More information about the Bf-committers mailing list