- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Hi,
We have a use case where we wanted to use the Descriptor controller of A10 AVMM DMA with pipelined accesses on Write DMA Avalon-MM Master port. Actually we need a huge delay to respond with the data for read accceses from wr_dma interface. So we wanted to pipeline the read addresses from the DMA descriptor controller Write DMA Avalon-MM Master port to optimize the read performance. The AVMM DMA userguide for A10 does not talk about this. Can you please confirm if the above usage is feasible with AVMM DMA or not.
Thanks,
RamaMohan
Link Copied
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
There are few different DMA IPs in the Platform designer. Are you referring to this PCIE specific https://www.intel.com/content/dam/www/programmable/us/en/pdfs/literature/ug/ug_a10_pcie_avmm_dma.pdf?
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Yes, I'm referring specifically to the the PCIe AVMM DMA pointed by you(Arria10). We are using version 17.1 of the Platform Designer.
Thanks,
RamaMohan
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
If you are using the internal descriptor controller of AVMM DMA PCIe IP, I don’t see there is an option to control the delay of descriptor controller, therefore, I afraid that this is not feasible to do this by using the internal descriptor controller.
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Hi GNg,
I'm sorry for the confusion. The delay is required for the read-transactions from Read data mover to application, where data is moved from application to the system memory. We do not need any delay for the descriptor controller transaction as these are always Descriptor table writes. We don't have any specific requirement for Descriptor controller instance.
Please clarify if pipe-lining is possible for read requests from PCIe Read DMA Data Move of AVMM DMA.
Thanks,
RamaMohan
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
I afraid that the pipe-lining is not an option as well from Read DMA mover
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Hi GNg,
Thanks for the clarification. Will let you know if I have any further questions.
Regards,
Ramamohan

- Subscribe to RSS Feed
- Mark Topic as New
- Mark Topic as Read
- Float this Topic for Current User
- Bookmark
- Subscribe
- Printer Friendly Page