[lm-sensors] Re: I2C block reads with i2c-viapro: testers wanted

Jean Delvare khali at linux-fr.org
Thu Aug 11 18:56:51 CEST 2005


Hi Hinko,

> > Could you try running "i2cdump 0 0x50" and "i2cdump 0 0x50 i" (with
> > the patch still applied), and compare both the outputs and the time
> > each command takes? You should see similar outputs, but the second
> > command should be magnitudes faster. This would confirm that the I2C
> > block mode works as intended on your VT8233 chip.
> 
> Hmm, not really. Here it takes 6 seconds for the first test nad about
> 5 seconds  for the second test (I just read the WARNING - need to
> substract 5s from the  results...).

With a recent version of i2cdump (2.8.8 or later), you can use the -y
flag, which will skip this delay. This is very convenient for timing
tests.

That being said...

> noa xtrm # time i2cdump 0 0x50
> (...)
> real	0m6.033s
> (...)
> noa xtrm # time i2cdump 0 0x50 i
> (...)
> real	0m5.174s

This is 1.033s down to 0.174s. This is just great, I2C block reads work
and allow faster dumps, as expected.

> while simple cat takes a lot less time:
> noa xtrm # time dd if=/sys/bus/i2c/devices/0-0050/eeprom bs=4

This goes through the eeprom driver, which has an internal cache, so the
results are not suitable for timing comparisons.

Thanks a lot for the testing again :)
-- 
Jean Delvare




More information about the lm-sensors mailing list