On Tue, Sep 15, 2015 at 02:16:03PM -0500, Felipe Balbi wrote: > Hi, > > with these patches (and, no, the mass storage patch is not extremely important > although it gives some nice extra improvement - about 1 MiB/sec extra in some > cases - at least not for USB2, I tested without that as well), I increased USB2 > on a AM437x board with g_mass_storage using RAM as backing store from 17MiB/sec > to 39MiB/sec as measured by dd (with oflag/iflag=direct depending on whether I > was writing or reading from g_mass_storage). > > I'm yet to test all other gadget drivers to make sure there are no > regressions, but it's looking promissing. > > cheers > > Changes since V1: > - fix regression with isochronous transfers found a regression with g_ether, I'll debug further tomorrow. -- balbi
Attachment:
signature.asc
Description: Digital signature
