This posting is an attempt to capture the rationale used in extending DAPL to support SRQs. I have not attempted to determine where to insert this, hence the...
To be added under Section 5.5 Memory Management, 4a. vi) uDAPL supports platforms with non-coherent memory in areas between - cpu cache and memory - I/O...
Sherman Pun
sherman.pun@...
May 4, 2004 12:27 am
2920
I have provided some editorial comments to the 1.2 draft specification to Arkady. Arkady will post those with other changes in his next revision. However, I...
Sherman Pun
sherman.pun@...
May 4, 2004 10:48 pm
2921
Isn't "outstanding buffers" (at least from the consumer's perspective) defined as the total number of buffers which have been submitted to the SRQ counting...
We will have a DAT Collaborative conference call this Wednesday May 5, 12:00-1:00pm EDT (10:00am-11:00am PDT). Moderator: Arkady Kanevsky Phone: 888-827-8686 ...
... My interpretation had been that you couldn't shrink an SRQ if doing so would force it to yield buffers back. Not that it *shouldn't* be able to give...
I think we are going too deep here. A simple statement that SRQ cannot be shrank below the number of outstanding buffers is enough. Remember that both creation...
Actually both uDAPL and kDAPL 1.2 drafts have been uploaded. The major changes: 1. errata 56 incorporated 2. bunch of editoral changes based on Sherman's...
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...
May 6, 2004 12:17 am
2935
We have two items for iWARP/IB extension verbs that relates to mx IOV. iWARP only supports of max IOV for RDMA Read of 1. For RDMA Write iWARP splits max IOV...
... IOV limitations are attributes of the local interface, not of the wire protocol. The distinction between Send, Write and Read IOV lengths will be nearly...
... Oops too. I didn't actually *read* the meeting notice until I was getting ready to look up the international number. And by then there wasn't anybody...
The issue is not whether to have 3 separate attributes but where to put them. I propose that all of them are attribute of IA rather than Provider. The main...
... Yes, they need to be IA attributes. But they are all a local interface issue, the wire protocol dictates nothing. The "stag" is an rdma read response MUST...
... We need to make that a bit more flexible to deal with SRQ implementations that forget about a buffer after it is transferred to an EP. The essential rule...
This changes the defition of outstanding DTO. Without it Consumer cannot size work queues including SRQ. Consumer can count each buffer as outstanding until...
I don't see how it makes much difference from the consumer's point of view. Either way they have to be aware of the limitations of what is beneath them. It...
I see that we have a consensus on IA attributes. I will add it to the next version of the drafts and schedule a vote on it for next week. Arkady Arkady...
Fellow DAPLers, We are getting closed to having a full draft of DAPL-1.2. For next week meeting I want to have full proposals for all outstanding items. ...
... We also need to decide whether the matching changes are made to QP attributes, or if we draft inference rules to avoid requiring that any existing...
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.
