On Tue, 2 Dec 2025 12:36:21 +0000 (UTC) Benny Siegert <[email protected]> wrote:
> But the thing is, I want to measure build performance taking into account > disk I/O and all of that. Also, the amount of RAM is tight already, so I > cannot afford a ramdisk :/ I would use a different system with enough RAM. I would run first build on disk, then run second build on tmpfs. Compare two values to see the disk I/O overhead. Use only a single CPU to keep timings consistent. Disk I/O has nothing to do with build tool performance. It's just another variable that has an impact on the amount of time it takes to complete build jobs. However with different storage subsystems that will change drastically. You can measure disk performance directly with tools like dd. Some time ago I measured how NetBSD build.sh scales on tmpfs with multiple CPU cores. I think it has a lot of serial sections which inhibit its scalability on large systems.
