Add hint to readme, how to use automatically the optimal number of CPU cores

[SunBlack]

Summary

Usually the best compile performance you get via calling make -j$(nproc) as it is using the number of available CPU cores (more doesn't increase usually the performance but increases the required memory).

Usually you get the best compilation performance by calling make -j$(nproc), since it uses the number of CPU cores available (more usually does not increase performance, but increases the memory needed). Furthermore, the change fixes the incorrect wording of specifying the number of CPUs (which is usually 1 on desktop systems) instead of CPU cores.

[lattner]

$(nproc) isn't set for me in bash, I don't think we should point people to that. I agree "CPUs" is probably the wrong thing but your suggestion doesn't cover SMT systems, maybe "the number of CPU cores/threads you have"?

[SunBlack]

You are right, CPU cores is still not correct. I believed at SMT the virtual cores are also called cores, but apparently they are not, if I look e.g. in the specification of Intel processors. I think the reference to $(nproc) is useful if your system supports it, though a rewording to say that it only exists on some systems would be better.

[SunBlack]

Updates the description. I'm still thinking about writing that the use of `-j$(nproc)` is recommended, if the system supports it, in environments with a number of dynamic CPU threads (e.g. due to different build agents).

[lattner]

LGTM, thanks!

[SunBlack]

Is there any chance that this will be merged? Or do I still have to do something for it?

[lattner]

Oh sorry, I assumed you had commit access. If not, it is easy to get, please read the LLVM Developer Policy and request access according to its instructions, thanks!

README.md

 
         * Running a serial build will be **slow**.  To improve speed, try running a
           parallel build.  That's done by default in Ninja; for ``make``, use the option
-          ``-j NNN``, where ``NNN`` is the number of parallel jobs, e.g. the number of
-          CPUs you have.
+          ``-j NNN``, where ``NNN`` is the number of parallel jobs to run.
+          In most cases, you get the best performance if you specify the number of CPU threads you have.
+          On some Unix systems, you can specify this with ``-j$(nproc)``.
 
       * For more information see [CMake](https://llvm.org/docs/CMake.html)