Hi Maxim, Bill, all, Thank you for adding this to the agenda for the next meeting, thus we can discuss in more detail. I understand the aim of config.h, but I think that, from a platform perspective, this setting should specify the maximium limit of packet headroom/tailroom that it's able to provide and, from an application perspective, the dynamic pool initialization is able to configure a vaild size for packet headroom/tailroom (when I said valid I mean that it doesn't exceed the platform specific limits defined by this config.h file). By this way, an application would be more flexible, and it does not depend on recompiling ODP libraries to increase the headroom/tailroom size when someone needs more space.
As I said, this is only my humble opinion, so we can discuss it during the next meeting. However, a use case could be to keep some metadata inside the packet while it's been processed along a pipeline (e.g, a set of chained operational routines), so that each routine can comunicate some information per-packet to others using the packet itselfs as a "means of transportation". I can't say you enough thanks for this feedback. Best, Genís Riera Pérez. Genís Riera Pérez Software Engineer at StarFlow Networks Edifici K2M, S103 c/ Jordi Girona 31 08034 Barcelona E-mail: [email protected] On Mon, Jul 13, 2015 at 1:42 PM, Bill Fischofer <[email protected]> wrote: > Actually, the original proposal for this had these as part of the > odp_pool_param_t structure, however some platforms have difficulty > implementing varying headroom/tailroom on a per-pool basis. > > But good to revisit this since we'd like to minimize use of odp_config.h > in preference to more dynamic APIs, so I'll add this to the agenda for this > week's call. > > On Mon, Jul 13, 2015 at 6:00 AM, Genis Riera <[email protected]> > wrote: > >> Hi all, >> >> I just wanted to propose an idea for extending the packet >> headromm/tailroom feature of ODP linux-generic platform. >> >> As far as I'm aware, the default values for initialize a packet >> headroom/tailroom can be found at config.h, and if someone wish to modify >> it, then ODP must be recompiled. Since I'm currently taking advantage of it >> to manage some metadata, I think would be interesting to provide some >> mechanism to set the packet headroom/tailrrom size when pool memory is >> initialized inside an application which uses ODP framework. For example, >> having a new parameter on odp_pool_params_t to set it up (when the pool is >> type of packet pool, of course). Does it make sense for you? >> >> So far, I haven't looked in detail the source code of linux-generic >> implementation, but if you think that this proposal is not too difficult to >> implement, I could try to do it. Furthermore, it would be a good way to get >> more involve in the ODP developement. >> >> Please, don't hesitate to make suggestions for this proposal. Does it >> make sense for you? >> >> Thank you so much for your support. Best regards, >> >> Genís Riera Pérez. >> >> Genís Riera Pérez >> Software Engineer at StarFlow Networks >> Edifici K2M, S103 c/ Jordi Girona 31 >> 08034 Barcelona >> >> E-mail: [email protected] >> >> _______________________________________________ >> lng-odp mailing list >> [email protected] >> https://lists.linaro.org/mailman/listinfo/lng-odp >> >> >
_______________________________________________ lng-odp mailing list [email protected] https://lists.linaro.org/mailman/listinfo/lng-odp
