And performance information too. I noticed that running full speed bulk traffic through a USB 2.0 hub is about 25% slower than running it through a full/low speed host controller, at least for higher volumes,
Unfortunately, when running over HS/TT hub, I didn't manage to get any of the sg_list testcases 5-8 working without getting stuck in D-state sooner
Hmm, odd. Were you still running 2.5.42? I don't think any of the patches would have caused (or prevented) that, but I was using more current code. What stuck, and why? Which firmware?
or later. I had a few occasions where test5 didn't get stuck and then it was completed (default test parameters) in 17.5 sec. This compares to the 17.0 sec it needs on the same port (NEC) without HS/TT, i.e. with companion ohci - no matter with or without hub running at FS in between. On an USB-1.1 ohci port (SiS-7001) it takes 16.0 sec for the same test.
I suspect there's something odd with the interrupt mitigation. If you feel like experimenting, try commenting out the td_fill() lines that mask in a TD_DI_SET(6) ... it's sometimes as if the last TD's 'DI==0' is being ignored. - Dave
but the latency for short requests is much reduced (they can complete before the next SOF).
Yep. Testcases 1-4 suggest with HS/TT hub the transfers are completed taking 5 microframes (-s 512) or even as little as 1 microframe (-s 64).
Martin
------------------------------------------------------- This SF.net email is sponsored by: ApacheCon, November 18-21 in Las Vegas (supported by COMDEX), the only Apache event to be fully supported by the ASF. http://www.apachecon.com _______________________________________________ [EMAIL PROTECTED] To unsubscribe, use the last form field at: https://lists.sourceforge.net/lists/listinfo/linux-usb-devel
