> -----Original Message----- > From: ext Hongbo Zhang [mailto:hongbo.zh...@linaro.org] > Sent: Tuesday, September 08, 2015 2:21 PM > To: Savolainen, Petri (Nokia - FI/Espoo) > Cc: ext Bill Fischofer; ext hongbo.zh...@freescale.com; > stuart.has...@arm.com; lng-odp@lists.linaro.org > Subject: Re: [lng-odp] [API-NEXT PATCH v4 09/10] linux-generic: > cpumask: add API odp_cpumask_available() > > On 4 September 2015 at 15:44, Savolainen, Petri (Nokia - FI/Espoo) > <petri.savolai...@nokia.com> wrote: > > Cpu mask API mainly manipulate the mask. So, odp_cpumask_max can be > mixed to > > return “max” or “last” cpu ID. Same problem with plain > odp_cpumask_all – > > it’s too close to “set all cpu IDs in the mask”. > > > > > > > > I think these “extra” functions which return system preferences > (default > > workers, controllers, all available for ODP, all CPUs on the device > (also > > those unavailable), …) can be a bit longer and thus more > descriptive. > > > Yes "all available" seems more descriptive. > > But let us think in another way, Stuart suggested me to introduce such > a function, further thought would be that this function can replace > the current odp_cpu_count(), I didn't take the further steps into this > patch set till now. > > Currently the only user of this new API is the validation code, I use > this API to iterate each CPU, so is it important enough to introduce > such a new API?
It could replace cpu_count(). The mask and the count are related. User should use both when iterating through CPUs. A count without the mask builds expectation that available CPU IDs are from 0 to count - 1, which may not be the case. User needs cpu count/mask for e.g. allocating resources per CPU and iterating those. -Petri _______________________________________________ lng-odp mailing list lng-odp@lists.linaro.org https://lists.linaro.org/mailman/listinfo/lng-odp
