Hi, In the bug report, Samuel Thibault writes:
parallel(1) says `The default is to run one job per CPU core.' but the code shows parallel uses the result of sysconf(_SC_NPROCESSORS_ONLN); which doesn't give the number of cores, but the number of processors, i.e. of threads in case the cores are multithreaded for instance. The attached patch just removes the "core" word to leave just "CPU". I disagree with his patch, as in my case, it will be the number of cores, but not the number of CPUs: my laptop has a single CPU with two cores and sysconf(_SC_NPROCESSORS_ONLN) returns 2. I believe the more common case will be CPUs that have one or more cores, but don't have multiple execution threads per core. I guess a middle way would be to say something like «The default is to run one job per CPU core(*)» and then have a footnote that reads something like «(*) we are using sysconf(_SC_NPROCESSORS_ONLN), which can return the number of hardware threads multiplied by CPU cores on machines that have execute more than one thread per core simultaneously.» (Or I might just be really nickpickety and you can ignore me. :-) Please Cc me on replies, I'm not subscribed to the bug or the package. -- Tollef Fog Heen UNIX is user friendly, it's just picky about who its friends are -- To UNSUBSCRIBE, email to [email protected] with a subject of "unsubscribe". Trouble? Contact [email protected]
