Dev Blog 1/31/2019 - Rendering Optimizations & Texture Packing

[jcsnider]

Rendering Optimizations & Texture Packing

Jan. 31st 2019

 

 

Client Rendering Optimizations

This is really hard to explain, but we've made some massive performance updates to the rendering system in Beta 6.

 

We no longer cache each map into 9 massive background images. 

(Lower/Middle/Upper layers * 3 Autotile frames)

 

Multiplying 9 massive images by 9 on-screen maps = 81 massive images.

Those no longer eat up valuable video ram.

 

The historical alternative for these engines was to draw every tile individually.

While that requires far less ram you kill your fps by telling your graphic card to draw up to 5k tiles every frame for every map. 

 

Now we're using a new method. We generate a set of triangles and vertices which are used to draw our map and ask the graphic card to remember those.

Going forward when we want to draw the map we tell the graphic card to use it's remembered instructions and we gain massive fps by not sending the full set of instructions each frame.

 

TLDR; Lower vRam usage, quicker map loading, higher fps, and other optimizations!

 

 

 

Texture Packing

When rendering graphics in a video game you can 'batch' draw calls. If you draw several consecutive sprites/objects/animations/ui elements/etc from the same texture all of those drawing commands can be combined and sent to the graphic card at one time.  The fewer batches that get sent the higher your fps.

 

Texture packing increases performance by combining lots of your small textures into just a few larger ones. 

 

There is a button in the Beta 6 editor that will do all of the texture packing for you:

 

 

 

Click here for an example of what a packed texture looks like.

 

Texture packing is completely optional, and the client will utilize packed textures if they are found in the resources folder.

 

 

 

The Bottom Line

Let's talk stats, what do these changes mean in practice?  We should see roughly the same or better fps across the board. Laptops with integrated graphics won't see a huge difference while gaming rigs with decent graphic cards might see fps gains in the hundreds.  As far a memory usage goes you should see a sharp decline all across the board, and the client will feel better when traversing map boundaries because there won't be a big stutter when map caching previously occurred.

 

Here are some of the basic performance tests that I did using my laptop with integrated graphics and my gaming rig.

 

 

This shows fps, higher is better. My gaming rig saw huge gains, laptop only climbed about 30fps.

 

 

This is gpu memory usage in megabytes, lower is better. Beta 6 brings huge drops across the board . 

 

 

These updates will be live with the Beta 6 launch! Stay tuned!

[Beefy Kasplant]

Massive improvement! Looking forward to B6 and good to see a Dev Blog again! 

[Sweet Candy]

Excelent work JC!
Im waiting for B6! 

[Ainz Ooal Gown]

Looking awesome mate!

 

Love the Dev blogs, proper highlight of the day when you see an email notification with a new one!

[Worldofjimmy]

13 hours ago, Ainz Ooal Gown said:

Looking awesome mate!

 

Love the Dev blogs, proper highlight of the day when you see an email notification with a new one!

I totally agree! Looking forward to every email notification from Ascensiongamedev