Python S3O module and optimizer - Page 3

Python S3O module and optimizer

Discuss the source code and development of Spring Engine in general from a technical point of view. Patches go here too.

Moderator: Moderators

Muon
Posts: 24
Joined: 23 Jan 2012, 21:42

Re: Python S3O module and optimizer

Post by Muon »

Suggestions to handle these cases? Perhaps a lower limit on the number of vertices before it decides to optimize.
User avatar
Beherith
Posts: 5145
Joined: 26 Oct 2007, 16:21

Re: Python S3O module and optimizer

Post by Beherith »

Yeah, don't optimize under 16 vertices. The gain negligible there anyway.
Muon
Posts: 24
Joined: 23 Jan 2012, 21:42

Re: Python S3O module and optimizer

Post by Muon »

Hm. Is there perhaps any more reliable way of detecting this? Optimizing 16-vertex meshes would still be a performance improvement if they were drawn often.
User avatar
Deadnight Warrior
Posts: 183
Joined: 08 Jun 2009, 17:59

Re: Python S3O module and optimizer

Post by Deadnight Warrior »

I tried it on XTA, since it uses Mr D's core vehicle models, and the result was only 2 or 3 files out of ~20 got optimized. When I tested the performance of those optimized units, there was no FPS gain and I was even sure that in case of CORE Cloakable fusion power plant I saw a FPS reduction.
User avatar
FLOZi
MC: Legacy & Spring 1944 Developer
Posts: 6241
Joined: 29 Apr 2005, 01:14

Re: Python S3O module and optimizer

Post by FLOZi »

Muon wrote:Hm. Is there perhaps any more reliable way of detecting this? Optimizing 16-vertex meshes would still be a performance improvement if they were drawn often.
Ask Nemo what he did when fixing the issue in S44, it was something trivial afaik.
Muon
Posts: 24
Joined: 23 Jan 2012, 21:42

Re: Python S3O module and optimizer

Post by Muon »

Deadnight Warrior wrote:I tried it on XTA, since it uses Mr D's core vehicle models, and the result was only 2 or 3 files out of ~20 got optimized. When I tested the performance of those optimized units, there was no FPS gain and I was even sure that in case of CORE Cloakable fusion power plant I saw a FPS reduction.
Hm, I should probably make the decision whether or not to save the optimized S3O depending on whether or not the ACMR goes down, rather than file size.

EDIT: Okay, updated. Now it commits anyway, because it may have optimized drawing regardless of the change in file size. I should point out that the vertex cache optimizer only works on triangles, but not quads or tri strips.

EDIT EDIT: I've run the optimizer on that model, and the ACMR goes from 2.009901 to 0.982673, which means it *should* yield a performance improvement.
User avatar
Beherith
Posts: 5145
Joined: 26 Oct 2007, 16:21

Re: Python S3O module and optimizer

Post by Beherith »

I would say anything under 4 vertices can definitely be left unoptimized.
Btw I use your s3olib extensively. Its super nice and well built!
Muon
Posts: 24
Joined: 23 Jan 2012, 21:42

Re: Python S3O module and optimizer

Post by Muon »

Beherith wrote:I would say anything under 4 vertices can definitely be left unoptimized.
Btw I use your s3olib extensively. Its super nice and well built!
Anything you could contribute back to the repo?
User avatar
Forboding Angel
Evolution RTS Developer
Posts: 14673
Joined: 17 Nov 2005, 02:43

Re: Python S3O module and optimizer

Post by Forboding Angel »

FLOZi wrote:In what quantum reality is Upspring reliable? :shock:
Upspring 1.54 is very reliable. It has it's quirks, but it is more or less rock solid. I have far more trouble with spring doing stupid shit to models than upspring.
Post Reply

Return to “Engine”