Fast Cvvdp Implementation in C
3 months ago
- #Benchmark
- #Performance
- #CVVDP
- Fast C implementation of the CVVDP metric from the University of Cambridge.
- Benchmarked on Linux, Core i7 13700k, comparing fcvvdp (single-threaded) and cvvdp (multi-threaded).
- fcvvdp uses 91% less RAM, 88% fewer CPU cycles, and is almost 18% faster in wall clock time.
- Compilation requires dependencies, Zig build system, and produces binaries in zig-out/bin/.
- Supports various display models (fhd, 4k, hdr_pq, etc.) and outputs results in JSON format.
- Apache 2.0 License, developed by Halide Compression, derived from Vship (MIT license).