On Thu, May 03, 2001 at 12:52:21AM +0200, Martin Diehl wrote: > Well, admittedly, I don't have complete understood how the HCD works in > this detail. So what I've tried was somewhat a blind man's approach. > However, tentatively not using uhci_append_queued_urb() is a big success > for me - in combination with depth-first linking!. > > The problem disappeared, nothing seems to be broken, queued bulk with uhci > is now working for me. I've just transfered several GB in a row, all in > 2112 byte queued transfers. Rock solid - withstands even floodpings from > localhost and remote (which saturates eth0 on this box - "too much work in > interrupt"). And oh, btw: for an idle testbox the sustained throughput > averaged over 1 hour was >1.1MB/sec - i.e. same as usb-ohci and about 10% > gain over usb-uhci! The 10% gain is IMHO coming from the depth first linking, the transfer is made in one go and no extra descriptors are fetched from memory. You can try the same with usb-uhci by changing the define USE_BULK_DEPTH_FIRST to 1. -- Georg Acher, [EMAIL PROTECTED] http://www.in.tum.de/~acher/ "Oh no, not again !" The bowl of petunias _______________________________________________ [EMAIL PROTECTED] To unsubscribe, use the last form field at: http://lists.sourceforge.net/lists/listinfo/linux-usb-devel