Keep 3do model support
Moderator: Moderators
Re: Keep 3do model support
Question .. would it be feasable to have a auto-converter in spring, that converts 3do models on game-load to s3o?
Re: Keep 3do model support
Feasible yes, sensible I dont think so, This is much better done externally.
This preliminary shot shows 0 performance delta between 3do and s3o. There are some minor bugs and some rendering things I need to work on still.
This preliminary shot shows 0 performance delta between 3do and s3o. There are some minor bugs and some rendering things I need to work on still.
- Attachments
-
- screen01037.jpg
- (1.02 MiB) Not downloaded yet
-
- screen01035.jpg
- (907.4 KiB) Not downloaded yet
Re: Keep 3do model support
Don't worry, even basic step to replace many rendering buffers with single buffer proved to require fixing half of the rendering code. This won't be fast. And until completed, there's no reason to enforce 3do assassination.
Re: Keep 3do model support
Early prototype class implementation shows very significant FPS boost. Something like 500% growth.
Although this number might go change in either direction: the implementation is not yet optimized, so it can go up, but in the same time it's not complete, so it may go down. In any case it's certainly worth the effort.
Although this number might go change in either direction: the implementation is not yet optimized, so it can go up, but in the same time it's not complete, so it may go down. In any case it's certainly worth the effort.
- Attachments
-
- screen01921.png
- (5.74 MiB) Not downloaded yet
-
- Posts: 823
- Joined: 21 Oct 2008, 02:54
Re: Keep 3do model support
Other than gpu performance increases, any other plans for FPS improvements?ivand wrote: ↑10 Apr 2021, 10:02 Early prototype class implementation shows very significant FPS boost. Something like 500% growth.
Although this number might go change in either direction: the implementation is not yet optimized, so it can go up, but in the same time it's not complete, so it may go down. In any case it's certainly worth the effort.
Re: Keep 3do model support
What do you mean specifically?Super Mario wrote: ↑10 Apr 2021, 15:06 Other than gpu performance increases, any other plans for FPS improvements?
Re: Keep 3do model support
The frame rate gains here are from reducing the cpu/driver load from drawing units. Spring rendering is more cpu bound anyway.
-
- Posts: 823
- Joined: 21 Oct 2008, 02:54
Re: Keep 3do model support
Other than porting spring to modern opengl, do you have any other improvements regarding performance that you have in mind?ivand wrote: ↑10 Apr 2021, 21:34What do you mean specifically?Super Mario wrote: ↑10 Apr 2021, 15:06 Other than gpu performance increases, any other plans for FPS improvements?
Re: Keep 3do model support
No. But if I happen to see the part that can be MT-ed, I'll MT it without any specific plan.Super Mario wrote: ↑11 Apr 2021, 19:37 Other than porting spring to modern opengl, do you have any other improvements regarding performance that you have in mind?
That said, right now the biggest perf. hog is rendering.
Re: Keep 3do model support
Well, CUnitDrawer class is something special in spring.
It took me most of the weekend to get from early in-place prototype to something more or less sustainable.
200+ FPS vs ~30-40 FPS as in status quo.
P.S. Don't mind shading, as it only shows model color + teamcolor at the moment.
It took me most of the weekend to get from early in-place prototype to something more or less sustainable.
200+ FPS vs ~30-40 FPS as in status quo.
P.S. Don't mind shading, as it only shows model color + teamcolor at the moment.
- Attachments
-
- screen01926.png
- (5.63 MiB) Not downloaded yet
Re: Keep 3do model support
AlphaPass is something most of the games replace with LUPS transparency, but it's anyway important for drawing ghosted buildings and cannot be left empty.
Status quo:
Work In Progress:
Status quo:
Work In Progress:
- Attachments
-
- screen01931.png
- (6.23 MiB) Not downloaded yet
-
- screen01930.png
- (6.18 MiB) Not downloaded yet
Re: Keep 3do model support
Like every software, there's a limit to the support given to any product/technology, I don't get why you seem to be apologizing yourself or trying to get consent before getting rid of old parts of the code that are too costly to maintain.
I would say: who does the work, decides what is done.
I would say: who does the work, decides what is done.
Re: Keep 3do model support
You are right in the sense: whoever does the job decides what is right and what is not.
In the same time we cannot distance from the community needs too much otherwise a very few or none will be using new engines. Some games use 3do exclusively (I'll leave IPR infringement question out) and the lack of people with shaders/OpenGL programming is real here.
We will take a measured approach so the removal is gradual and migration tools are provided. That doesn't mean engine team will do everything themselves, game devs will need to invest some time too to leverage future perf improvements (akin to FPS numbers seen above).
In the same time we cannot distance from the community needs too much otherwise a very few or none will be using new engines. Some games use 3do exclusively (I'll leave IPR infringement question out) and the lack of people with shaders/OpenGL programming is real here.
We will take a measured approach so the removal is gradual and migration tools are provided. That doesn't mean engine team will do everything themselves, game devs will need to invest some time too to leverage future perf improvements (akin to FPS numbers seen above).