Cyril Wallois wrote: > Hi Jan, > > I solve my problem of GPIO which comes from a bad address. > > So I could measure the time to write a packet on the dm9000, It takes 200 µs > for a packet of 1000 bytes.
...as I was afraid of. You should be aware of the fact that those 200 us add to the intrinsic latency of your platform. Even non-RTnet application will suffer. > > I had cheked if someone has ever try to use dma, but I don't find anything, > maybe there is a reason. I'm afraid it need lots of works to modify it for > dma, isn't it ? Well, it might even be infeasible if the hardware doesn't support it. Check the datasheets of your platform, check latest Linux drivers - if there are no traces of DMA support, you lost. But, while writing a frame into the NIC, do you really need exclusive hardware access? RTnet xmit handler can be synchronized (set NETIF_F_LLTX in rtdev->features), you just need to make sure that there is no conflict with other parts of your driver manipulating the device state in an incompatible way. Jan
signature.asc
Description: OpenPGP digital signature
------------------------------------------------------------------------------
_______________________________________________ RTnet-users mailing list RTnet-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/rtnet-users
