[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: Crossing the First 4 GB Address Boundary
The PCIX spec. does not really need to have a special implementation note.
Both specs. require that when mastering an address that is above 4 GB, you
must use a DAC so both specs. require that if restarting a disconnected
transaction on the other side of 4 GB, you must come out with a DAC.
Rich Iachetta
IBM Microelectronics Division -- Austin
World Wide Field Design Center
Phone: 512-838-6305 Tie Line: 678-6305
Rob Kellogg
<RKellogg@NetOcta To: "'pci-sig@znyx.com'" <pci-sig@znyx.com>
ve.com> cc:
Subject: Crossing the First 4 GB Address Boundary
05/21/02 01:49 PM
PCIX 1.0a, page 49, top paragraph contains an implementation note that
reads as follows:
"If a burst transaction crosses the first 4 GB boundary (i.e., the boundary
between memory locations for which the upper 32-bit of the address are 0
and locations for which they are not), the Sequence begins with a single
address cycle (see Section 2.12.1). If the Sequence is disconnected and
resumes beyond the first 4 GB boundary, the continuation transaction
requires a dual address cycle..."
I looked for a similar note or specification in PCI 2.3 (I searched the
document) and I cannot find any. I did find in section 3.9 "64-bit
Addressing" several statements to force 64-bit devices and 32-bit devices
to use SAC when the address is >= 32-bit and DAC when the address is <=
64-bit, but nothing at all like the above statement linking a continuation
of a burst following a disconnect. Does anyone know the reason that
statement is there and if it does or does not apply to PCI 2.3?
Thanks in advance,
-rob