FFmpeg build times on a large custom codec codebase
We compile FFmpeg from source for custom codec work and video pipelines so build time hits us directly. Clean builds are sitting around 24 minutes on a 16-core Xeon and CI is regularly blocked on compilation during active development.
ccache helped with incrementals but didn't move the needle on clean CI runs. S tripped the build down with --disable-everything and only enabled what we actually need, which helped a little. profiling with ninja -d stats puts compilation at around 80 percent of wall time, linking at 15 percent and mostly serial.
Haven't gone deep on distributed compilation yet. Looked briefly at distcc and icecc but haven't committed to either. Also sitting on a linking bottleneck that I'm not sure how to approach without it just becoming a long serial step.
Anyone running distributed builds on FFmpeg or similarly structured C codebases? Curious wheter it actually produces a step change or wheter something else tends to be the ceiling first.