It makes sense. I managed to mix things up. Should we add documentation that "cpu" parameter is typically coming from odp_cpumask_first().
-Petri > -----Original Message----- > From: ext Robbie King (robking) [mailto:[email protected]] > Sent: Tuesday, January 13, 2015 5:23 PM > To: Savolainen, Petri (NSN - FI/Espoo); [email protected] > Subject: RE: [lng-odp] [PATCHv2 1/2] Convert linux thread/proc helpers to > use core mask > > Hey Petri, I think it needs to be there at least the way I am currently > using it. You begin the process by calling "odp_cpumask_first" to get > your starting CPU, and the "next" function is the continuation. So > the CPU number passed into the function is assumed to be the previously > set CPU, and we are searching for the next one. So the "+1" is necessary > to skip over the CPU passed in (else you would just keep finding the > same CPU). > > If you wanted to only use "odp_cpumask_next", I think it could be > done with passing in "-1" as the CPU to indicate start versus > continuation. > > Hope this makes sense. > > > +int odp_cpumask_next(const odp_cpumask_t *mask, int cpu) > > +{ > > + for (cpu += 1; cpu < CPU_SETSIZE; cpu++) > > > First valid cpu ID is 0. So, "cpu += 1" should not be there, right? > > -Petri > > > > + if (odp_cpumask_isset(mask, cpu)) > > + return cpu; > > + return -1; > > +} _______________________________________________ lng-odp mailing list [email protected] http://lists.linaro.org/mailman/listinfo/lng-odp
