* Jean Delvare <khali at linux-fr.org> [2004-01-10 13:49:45 +0100]: > Hi all, > > >From what I see in the w83781d driver, it is assumed that the as99127f > extra temperatures are 9 bit signed values with LSB = 0.25 degree C (as > opposed to LSB = 0.5 degree C in other winbond chips and lm75 itself). > > What reason do we have for such an assumption? > > I think we are wrong here, for the following reasons: > > 1* Asus chips are known to be copied from Winbond ones. I don't see any > reason they would change the LSB value. What's more, MMH's asb100 driver > uses standard lm75 conventions. > > 2* For all boards using the as99127f, a conversion is needed for temp2 > and temp3 (which is not that frequent AFAIK), and the default is: > compute temp2 @*2.0, @/2.0 > compute temp3 @*2.0, @/2.0 > Which brings us back to a 0.5 LSB. I know that Asus are the recognized > international champions of senseless temperature conversions (as I have > experienced on my TX97-E) but let's not charge them with something they > probably haven't done. > > So I propose that we use standard lm75 conversions for the as99127f > subclients. This will simplify the code much. Comments, objections? I thought you had an as99127f to test? I don't have one. AFAIR there are multiple different kinds out there. I felt free to change/fix the conversions when going from as99127f to independent asb100 driver because it required an updated sensors.conf for the user anyway. as99127f is a different story - are you sure you want to break everyone's existing config? > BTW, MMH, would it make sense to move the as99127f support from w83781d > to asb100? I haven't looked at the code in deep myself yet, just > wondering if you have an opinion. A step backward, I think. If each driver supported just one chip then that would allow the driver to stabilize and just accumulate bugfixes until it's "done". Someone submitted a patch once to put as99127f into its own driver. I'm sorry I wasn't able to review it at the time. Well, it's in the archives... Regards, -- Mark M. Hoffman mhoffman at lightlink.com
