Thanks for the answer, that really clears it up for me. Could you link the previous question you mention for my own information?
So if I really wanted to use a fixed address for TX buffers, I'd basically have to reimplement an allocator that makes use of the uncached zone. Sure sounds inconvenient. Regarding performance, it probably is slower but it also means more predictability at run-time as 1) you don't need to perform additional cache operations from user-space at run-time 2) you avoid additional system calls at run-time. That sounded more elegant for a real-time system. Anyway, we'll use run-time cache primitives then. Thanks a lot for the help, Olivier 2017-11-07 21:17 GMT+01:00 [email protected] <[email protected]>: > Olivier Desenfans wrote: >> >> Regarding the TX side of things, > > > Keep in mind that most TX functions allocate PBUF_RAM pbufs, which are > allocated via mem_malloc(). To move this into pools, you'll need some extra > work. > > Also, are you *sure* constructing packets in uncached memory is faster than > flushing the cache before transmission? > (Wait a minute, I feel like I have asked this question before here? *g*) > > Cheers, > Simon > > > > _______________________________________________ > lwip-users mailing list > [email protected] > https://lists.nongnu.org/mailman/listinfo/lwip-users _______________________________________________ lwip-users mailing list [email protected] https://lists.nongnu.org/mailman/listinfo/lwip-users
