RE: PCI 64-bit addressing - problem

[Lame Brooks-G14738 <brooks.lame@motorola.com>]

Raan, how are you expecting to have 64bit support with your 440ZX
motherboard?  If I'm not corrected, the 440ZX is a 32bit PCI chipset.  Is
there a true 64bit slot on the board? -- BrooksL

> -----Original Message-----
> From: Raan Kahn [mailto:raan@ngcable.com]
> Sent: Thursday, 24 February, 2000 10:33
> To: PCI SIG (E-mail)
> Subject: PCI 64-bit addressing - problem
> 
> 
> Hi,
> 
> We have a problem with 64-bit memory addressing on a 32-bit PCI bus.
> 
> We have a chip that is connected to a 32-bit PCI but has an 
> internal pci 
> core that allows the PC to allocate our memory space in a 64-bit area 
> (address 0x10 in the configuration space, which is the memory 
> base address 
> register, is of reset value 0xC).
> 
> I see that the PC allocates us in a 64-bit address space. 
> This I see by the 
> fact that the PC configures two consecutive memory base 
> address registers: 
> the PC configures a value of 0xFFA0000C in memory base 
> address 0x10 and a 
> value of 0xFF in memory base address
> 
> When I perform a memory transaction, I expect the PC to 
> address my chip 
> with a DAC (Dual Address Cycle), because the PC is supposed 
> to know that my 
> memory space is allocated in a 64-bit space. But instead, the 
> PC accesses 
> my chip with a SAC (Single Address Cycle), only passing the 
> base address of 
> 0xFFA000xx, and for this reason my chip does not recognize that the 
> transaction in towards it and therefore does not assert 
> #DEVSEL, causing 
> the PC to do master abort.
> 
> We are using an Intel motherboard BI440ZX, and WIMDOWS NT 4.0 
> service pack 
> 3.
> 
> Why doesn't the PC perform DAC? Is this problem a motherboard 
> problem or 
> the operating system (Windows NT) problem?
> How can this be solved?
> 
> When I use Windows NT 4.0 with service pack 5, but on a different 
> motherboard, the PC does not allocate us in a 64-bit address space 
> (configures the second configuration memory base address, 
> address 0x14, to 
> a value of 0x0), causing our chip to recognize only SAC, so this
> 
> 
> Raan.
>