Sorry for the delay. Responses below .. On Thu, Oct 17, 2019 at 4:16 PM Kurt Van Dijck <dev.kurt@xxxxxxxxxxxxxxxxxxxxxx> wrote: > > Hey, > > On do, 17 okt 2019 15:37:04 -0500, Elenita Hinds wrote: > > I pulled in the c_can patches from the for-kurt branch (63574e9 thru > > bf01f717) and tested on my device. > > The number of overruns are noticeably fewer; however, the overall > > system performance seems to have slowed down. For example, > > the console response and Bluetooth data rate are noticeably slower. > > You now give more precedence to CAN and less to console etc. ? > > [eh] No. The priorities are the same or however Linux handles interrupt priorities. > > I also noticed that while the number of overruns decreased, the number > > of errors increased: > > You have an interesting setup. I suspect that the FIFO is filled, and > does drop messages, although that is recorded in the statistics? > > Would you want to try branch c_can-v4.9 in github.com/kurt-vd/linux? > I improved the reception a bit, and increased the FIFO size. [eh] I pulled the c_can-v4.9 changes (delta from v4.9) and the behavior is the same -- the number of overruns is reduced (about 0.011% loss compared to about 0.32%) but the number of errors is increased (used to be the same number as the overruns). I still see a slight degradation in system performance. I also tried the c_can-more-objects branch (increased the message objects to 64) and this also improved CAN handling (about to 0.004% overruns). Then I combined both IRQ-rxoffload and more-objects changes together and the number of overruns is further reduced (loss is close to 0%). > Besides that, can you give more details about your system and your load? [eh] System is similar to Beagleboard Black: TI AM335x. >From 'top', running about 0% idle, with sys about 44%, usr about 43%, sirq about 9%. Load average: 4.70 4.88 4.04 Let me know if you need more info. > > Kind regards, > Kurt
