Update README.md

This commit is contained in:
Darío 2024-04-17 11:20:06 -03:00 committed by GitHub
parent bcbdeeb239
commit 884656c580
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194

View File

@ -63,7 +63,7 @@ Any operation related to reading memory and using it as texture information duri
All vertex transformations by the RSP (e.g. position, lighting, texturing, etc.) are performed by a highly parallel compute shader in the GPU. This saves a lot of CPU processing time and allows for much higher vertex and polygon counts. Since all transformations are deferred to this step, it's very easy for RT64 to patch transformations of the objects in the scene and the camera and produce a new frame very quickly. This effectively removes CPU bottlenecks that are critical for reaching very high target framerates.
## Texture Decoder (Compute)
RT64 does not decode textures on the CPU and instead opts for uploading TMEM (4 KB) directly to the GPU. If possible, a RGBA32 version of the texture will be decoded and cached using a compute shader. If the sampling parameters prove to be too troublesome for that (e.g. giant texture masks with no clamp due to bad configuration), then RT64 can just sample TMEM directly just like the console with a small performance sacrifice.
RT64 does not decode textures on the CPU and instead opts for uploading TMEM (4 KB) directly to the GPU. If possible, a RGBA32 version of the texture will be decoded and cached using a compute shader. If the sampling parameters prove to be too troublesome for that (e.g. giant texture masks with no clamp due to bad configuration), then RT64 can sample TMEM directly just like the console with a small performance sacrifice.
RT64 features one of the most accurate TMEM loaders to date so far which has been directly reverse engineered by observing console behavior with the aid of homebrew test ROMs developed by [Wiseguy](https://github.com/Mr-Wiseguy). All the color conversion formulas for decoding have also been sourced from [Tharo](https://github.com/Thar0)'s excellent RDP research.