Li-Ta Lo wrote:
So you still think it is the timer ?
It may not be the only problem. But its one problem. The flaky results I get seem to match up with wierd timeing problems.
I haven't allocated a lot of time on this because we have found that our board has hardware problems with the pci bus. And it's difficult to seperate out all the individual issues.
But testbios dosen't work on an ASUS P2B booting LB or COTS and my modified ATI vbios. Nor does LB+ADLO+BOCHS with the original M1 vbios. X with InitPrimary works both under COTS and LB but that was using using the orginal M1 bios not my "fixed" version.
So see I still have a lot of apples->oranges comparisons. I'll try to scratch out sometime and build a consistent maxrix of everything that works and what dosen't.
This is where having that PCI expansion ROM stuff working would come in _real_ handy.
Did you try to access the timer HW in testbios ?
I think HW access is what you _don't_ want to happen. Rather than allow the vbios to fiddle with the actuall timer hw it needs to provide a software legacy timer interface that increments consistent with the emu "clock". The only hardware IO that should be allowed untrapped is the IO range requested by the card and the Legacy VGA ranges. Currently testbios lets the vbios access any port it wishes. The X emu must be doing this or the ATI M1 bios would not work.
This has to happen for non-x86 platform stuff as well which don't even have a x86 type timer.