Fliperworld

Results 19 comments of Fliperworld

@gdkchan : I hadn't considered BC1 yet for two reasons: I have no idea how much semi-transparent textures games would use, and mainly because i intend to keep Ryujinx's default...

Yes, it actually helps more who has more VRAM, as the user will not need to apply limitations to the extreme.I assume that just setting cache texture to about 1.5k...

> Unrelated to other comments, the method of reducing the texture resolution this way (manually downsampling) is unnecessary work. Nearly all of these textures come with mipmaps - just decode...

> Can this not be behind a single toggle, perhaps combined with texture recompression, that would: > > 1. Query the users VRAM. > > 2. Use a defined preset...

My appologies, I accidentally closed this PR when i was forcing my visual studio to resync.Anyway... I did some tests with BC1 recompression... in fast mode(using default textures sizes) it...

@DniweTamp I also did it in my tests...to be more precise, i tryed a "texture quality implementation" .... ultra = no recompression high = bc7 medium = smart (start at...

@jduncanator That's why i'm not making any pull request or opening a issue. I'm just sharing some experience/workflow of what i did to solve a issue for me that's can...

@gdkchan Sure, there as some samples on Original ASTC (only at a point where my vram explode) , BC7, BC1 , and my hack at 'limit 512"(the lowelest one): MBs...

To be fair, if i use something like [BCnEncoder.NET](https://github.com/Nominom/BCnEncoder.NET), just for tests, i got same results on fast recompress, and a little less block on Slow mode, at a cost...

> could you share a latest build for testing purposes https://github.com/Fliperworld/Ryujinx/tree/Testing-ForceBC1 https://github.com/Fliperworld/Ryujinx/tree/Testing-ForceMipLevels-Limit512