Hi Greg, For Informations, here are some measurements I can remember of different solutions I have tested, measurements where done with iperf in TCP and UDP : - Copyback cache without flush: ~40Mbits/sec in TCP but UDP tests shows a lot of lost frames. - Copyback cache with full data flush : ~9Mbits/sec in TCP, don't remember in UDP. - DMA_ZONE : ~32Mbits/sec, UDP ok. These results are very approximative, if I can I'll try to give you more precise tests results. For the partitioning and used TLB entries, with default settings I map 8MiB with 8TLB entry (on 32). I choose this path as it is recommended by the documentation and I also saw that it was the solution used by Freescale, on their old Linux patch. I agree that cache managing functions should also be improved. But with available cache management possibility on mcf54xx mcu, I don't see how to perform effectively the operations needed to exchange data with built-in DMA. So for my project I will live with this solution for the moment. Regards, -- Stany -- To unsubscribe from this list: send the line "unsubscribe linux-m68k" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
