Oliver Endriss wrote: > On Monday 14 July 2003 14:48, Johannes Stezenbach wrote: > > > > There are two things that cause the high load: > > - the I2C driver > > - the frontend thread that uses I2C to talk to the frontend > > > > If you can improve the I2C driver, great. But the load won't go > > unless we change the frontend thread. Which isn't difficult, but it > > is a policy decision. > > According to oprofile the load is caused by the busy-waiting stuff in > the i2c code. The zig-zag scan isn't a problem, because the fe thread > *sleeps* for 200ms after each tuning attempt.
Hm. > BTW, I had a look into dvb-kernel and noticed that the i2c code is > interrupt-driven. A quick test showed that the fe thread caused > virtually no cpu load, even if the sat signal is missing! > > Maybe the guys who use the DVB-S without a dish should simply switch to > dvb-kernel... Last time I checked the dvb-kernel I2C was noticably slower because the mdelay()s were replaced by schedule_timeout()s. But with your patch applied it could be just as fast? Johannes -- Info: To unsubscribe send a mail to [EMAIL PROTECTED] with "unsubscribe linux-dvb" as subject.
