On Sunday 05 June 2005 13:05, Dominik Brodowski wrote:
> IIRC (from the comment above) several chipsets suffer from this
> inconsistency, namely the widely used PIIX4(E) and ICH(4 only? or also
> other ICH-ones?). Therefore, we'd need at least some sort of boot-time
> check to decide which method to use... and based on the method, we can
> adjust the priority maybe?

To begin with, will the simple proof-of-concept patch like below work? (It 
tests the chipset read in the same do{}while loop - if the loop executes only 
once, it considers the chipset good - in which case it executes the faster 
read_pmtmr_fast function.) Or does it need wider testing under different 
circumstances to conclude that chipset is good?

I tested the patch under Virtual PC which emulates a PIIX4 chipset. Test 
passes there, meaning the do {}while  loop executes only once.

Parag