Read a bit more about video codecs and wrote AVX2 for 10-bit H.264 pred16x16 in FFmpeg. Plane prediction had no 10-bit SIMD at all, it's now 9.18x over the C reference (453 to 49 cycles), bit-exact via checkasm. Patch series just went to ffmpeg-devel. [1/2]
ZaneHam
11 posts
Here to chat about cool things. Self-taught programmer.
github.com/Zaneham if you want to see some of what I do
New Zealand
Joined March 2026
- Just ran doom on a Quantum Computer with a compiler I wrote. Currently the hardware just isn't quite there yet so the pixels is as close as we can get! You can check out the compiler here: github.com/Zaneham/ernest:
- Currently adding Triton and HIP support to BarraCUDA, my GPU compiler. Theres's limitations but I'm very happy with how it's turning out. It currently works on AMD, Nvidia and Tenstorrent GPU's. Check it out:
- My open source systemverilog and VHDL synthesiser is now available. Feel free to design your own chips!
- I got annoyed that chip design tools cost more than a house, so I made my own synthesiser. It handles SystemVerilog and VHDL. Then I started designing chips with it. Heres a couple examples: github.com/Zaneham/voyage… github.com/Zaneham/dozena…