docs: fix support order, add table for direct backend compatibility

Author: crimist

README.md

@@ -21,7 +21,7 @@ This is an VA-API implementation that uses NVDEC as a backend. This implementati
 
 # Codec Support
 
-Hardware encoding only, decoding is [not supported](/../../issues/116).
+Hardware decoding only, encoding is [not supported](/../../issues/116).
 
 | Codec | Supported | Comments |
 |---|---|---|
@@ -126,13 +126,19 @@ There's no real reason to run it with mpv except for testing, as mpv already sup
 The direct backend is a experimental backend that accesses the NVIDIA kernel driver directly, rather than using EGL to share the buffers. This allows us
 a greater degree of control over buffer allocation and freeing.
 
-Given this backend accesses the NVIDIA driver directly, via NVIDIA's unstable API, this module is likely to break often with new versions of the kernel driver. If you encounter issues using this backend raise an issue and including logs generated by `NVD_LOG=1`.
+**Generational compatibility**
+
+| Generation | Desktop | Mobile |
+|---|---|---|
+| <=Maxwell (<=9XX) | ❓ | ❓ |
+| Pascal (10XX) | ✅ | ❓ |
+| Turing (20XX/16XX) | ❓ | ✅ |
+| Ampere (30XX) | ✅ | ❓ |
+| Lovelace (40XX) | ❓ | ❓ |
 
-**Verified working generations**
+❓ - Unknown, let us know the test results of an unknown generation [here](/../../issues/126).
 
-* Pascal (10XX)
-* Turing (20XX)
-* Ampere (30XX)
+Given this backend accesses the NVIDIA driver directly, via NVIDIA's unstable API, this module is likely to break often with new versions of the kernel driver. If you encounter issues using this backend raise an issue and including logs generated by `NVD_LOG=1`.
 
 This backend uses headers files from the NVIDIA [open-gpu-kernel-modules](https://github.com/NVIDIA/open-gpu-kernel-modules)
 project. The `extract_headers.sh` script, along with the `headers.in` file list which files we need, and will copy them from a checked out version of the NVIDIA project