A slight clarification of a note in the attachment re RDMA Read outstanding in iWARP. There's an ORD and IRD, Outgoing Read Depth and Incoming Read Depth. The...
As you stated it is primarely for configuration management. In Linux, IA drivers (that include IT API implementation) are not provided by the OS vendor but by...
Jim, I understand you answer. For all query operation we do report all the parameters that users pass in, so why not here? I suggest we do report API version...
Jim, yes, that is what DAT defines. The version number is not quite IA version number but really implementation library version number. DAT does also expose IA...
Peter, DAT does not have one right now. DAT can define one. But I think DAT Consumers do want to see real numbers that is why it is in the DAT spec to begin...
Thanks Jim. I will pass it to DAT along with Peter's suggestion for discussion. Until DAT decides how to deal with for DAPL-1.2 I will mantain my (DAPL-1.0)...
Hello, This email message is a notification to let you know that a file has been uploaded to the Files area of the dat-discussions group. File :...
dat-discussions@yahoo...
Dec 9, 2002 7:21 pm
1640
Hello, This email message is a notification to let you know that a file has been uploaded to the Files area of the dat-discussions group. File :...
dat-discussions@yahoo...
Dec 9, 2002 7:21 pm
1641
I finally merged the two spec documents into one. I deliberately highlighed with colors differences between kDAPL and uDAPL specs. uDAPL specific text is...
Joe, I would like to add one more proposal: Proposal C: RMR triplet definition: All fields in the RMR triplet to be in network format (BIG ENDIAN, network byte...
Enclosed, please, find the first draft of new DAPL-1.1 APi call for allocating connection qualifier and creating public service point on it. Caitlin, you still...
I am confused by something in Arkady's email. Please excuse my ignorance but why is the RMR_context in network byte order format by definition? If this is the...
Hi Ellen, I think that by stating that the RMR_context is always returned and specified in network byte order, we make clear to the consumer that no ...
Peter Hochschild
phoch@...
Dec 11, 2002 2:08 am
1650
[Apologies for my mail server wrapping your quoted message...] Specifying the byte order of the triplet ensures that the RMR_context is transmitted to the peer...
... Four bytes are four bytes, so I can't see any run-time gotchas that could result from this. However, Ellen's fundamental point is totally correct. The RMR...
... Having just defended the divisibility of advertised buffers on the rddp mailing list, I will turn right around and agree that the overwhelming majority of...
Cailtin, Just to make it clear what we are arguing is do we need to expose rmr_triplet or just use rmr_context and rmr_ta instead. This will effect RDMA...
Due to recent developments in IB, see IBM Microelectronics Kills InfiniBand Chip, Byte and Switch <http://www.byteandswitch.com/document.asp?doc_id=25633>, I...
... Thanks Arkday, I assume that the note above is to call attention to the fact that semantics are undefined until process termination is complete. Is this...
Peter Ogilvie
peter.ogilvie@...
Dec 12, 2002 1:17 am
1658
Arkady, Where does this leave the HCA silicon abstraction layer? Didn't you folks choose the IBM spec for the reference implementation? Is that still going...
Having problems with message search?
Fill out this form to ensure your group is one of the first to be migrated to the new message search system.
