Re: Split Completions

["Richard Iachetta" <iachetta@us.ibm.com> (by way of Daniel Weaver <dan.weaver@znyx.com>)]

>1. If due to internal completer error or unpopulated address  range (say
>a 128 byte block with only 8 bytes populated), the backend  application
>stops providing split read completion data in the middle of an  ADQ, is
>the PCI-X logic supposed to insert dummy data and negate Byte  Enables
>till the next ADB and then disconnect?

Once you start a split completion, there is no stopping it in the middle of
an ADB.  Byte enables mean nothing.  Whatever garbage data you deliver will
be delivered so I recommend that before crossing an ADB, that you know that
you have another full ADB worth of real data to deliver (or enough to
satisfy the byte count).

>2. The SCM generated  will have the remaining lower-address and remaining
>byte-count fields  indicating the data not sent starting from the dummy
>data in the previous  split completion?

Once you disconnect on the ADB, if there are more bytes remaining to be
delivered (byte count not yet zero), you can give an SCM for the
"remaining" bytes that you haven't yet delivered.  Again, you have
delivered the last ADB worth of bytes even if it wasn't all good data.

Rich Iachetta
IBM Microelectronics Division -- Austin
World Wide Field Design Center
Phone: 512-838-6305   Tie Line: 678-6305