> But let's bust this ludicrous idea and show us a test that disproves me
Emotive terminology like "ludicrous" doesn't encourage a constructive response. If I understand correctly, I think you're describing your own proposal with that term, as a rhetorical device. It wasn't ludicrous and Paul didn't describe it like that. He said it would be a good default metric, but explained why it isn't done like that at the moment. I'm not sure I have understood correctly, because you go on to say: > Here is [a] slightly improved test script[.] ... yet I didn't see one. > if you explain what it does You might be left with questions after reading: https://www.gnu.org/software/make/manual/html_node/Parallel.html ... but they'll be more specific. > I'm not the first one to propose that[1], it turns out Far from it, yet Paul, who's been maintaining this project for nearly 30 years, patiently responded. I do appreciate that you're likely not a native speaker, but I suspect you are being deliberately robust, beyond what is due, and being somewhat cavalier about it. Perhaps you've constrained your thinking to the traditional use of Make to build C programs on a single computer. Imagine that it's being used instead to farm out jobs across an industrial-scale compute farm. Most of the elapsed time in each job on the initiating node is spent twiddling its thumbs waiting for the results to come back. If each one takes 0.1 s to initiate but 10 s to complete, it might good sense to start 100 in parallel - instead of 1000 s for the whole run, it takes a little over 20 s. Atypical use could easily use atypical arguments, yes. No one was arguing that, with the current behavior, make -j is often (ever?) useful on its own.
