On Thu, 2024-10-31 at 17:14 -0400, David A. Wheeler wrote: > > On Oct 31, 2024, at 4:48 PM, Paul Smith via austin-group-l at The > > Open Group <[email protected]> wrote: > > Previously I recommended using a special value, such as "-j0", to > > mean "best guess". > > I think using "-j 0" as "best guess" *should* be standardized. > > Selecting the number of processors is often challenging. Having a way > to say "run in parallel, you work out the details" is a helpful thing > to standardize.
I don't disagree, but as I mentioned before choosing a "best guess" is getting increasingly difficult for a tool like make. More compilers and compiler-like facilities (linkers in particular) are becoming multi-threaded in their own right. If make detects an 8-core system and decides to invoke 8 linkers to link 8 different binaries and each linker detects an 8-core system and decides to start 8 threads to manage the link... things can get out of control. GNU Make publishes a "jobserver API" which allows other tools to join into the jobserver implementation and reserve jobs from the -jN value that is provided to make. I have no idea how many, if any, tools are attempting to make use of this.
