1 / 7

Transmit and Receive Contexts

Transmit and Receive Contexts. Sean Hefty. Local Queue Management. Traditional view Fixed number of entries Each entry a fixed size. Memory optimized Fixed allocation Entries variable size. Local Queue Data. Application request Specified operation 0 or more IO vectors (SGEs).

deliz
Download Presentation

Transmit and Receive Contexts

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Transmit and Receive Contexts Sean Hefty

  2. Local Queue Management • Traditional view • Fixed number of entries • Each entry a fixed size • Memory optimized • Fixed allocation • Entries variable size www.openfabrics.org

  3. Local Queue Data • Application request • Specified operation • 0 or more IO vectors (SGEs) • Provider limits • Size of transmit queue • Maximum number of IOVs www.openfabrics.org

  4. Local Queue Data Size to record operation Size to record IO vector Max IOVs Size in bytes Alignment restriction www.openfabrics.org

  5. Local Queue Management • Queue attributes • Total size, in bytes • IOV limit – max number of IOVs for a single operation • Operation alignment, in bytes • Calculate padding with each operation • Domain attributes • Operation size, in bytes • Amount of space consumed to track each request • IOV size, in bytes • Amount of space consumed for each IOV www.openfabrics.org

  6. Local Queue Management • Generic calculation done by app: needed = ((op_size + iov_size * nsge) +op_alignment- 1) & ~(op_alignment - 1) • For queues with fixed sized entries: • needed = op_alignment • For queues with variable sized entries • needed = op_size + iov_size * nsge www.openfabrics.org

  7. Injected (inline) Data • Queue attribute • Maximum injected data size • Queue space calculation needed = op_size + inject_size • (Subject to alignment) www.openfabrics.org

More Related