12-22-2018 04:55 PM
I have several megabytes in DRAM that are batch loaded. Once done I need to walk sequentially through the addresses, retrieving the data. I am working in a 156.25mHz SCTL. It works fine and I get all the data, but very slowly (relative). I do a very tight loop of two states. One does the request and the other does the Retrieve and coverts the data and puts it into a FIFO. The Retrieve takes hundreds of clocks for each valid data. There are no other actors associated with the DRAM, I am the only one accessing it.
I found it by putting a counter on the number of clocks with valid/non-valid data. I am in a very time (aka throughput) application.
01-02-2019 11:30 AM
Hello,
Are you using architecture similar to the example "Getting Started - External Memory.lvproj". you can find that in the NI Example Finder in LabVIEW at Help>>Find Examples.
The architecture in this would be good to study and potentially emulate to achieve faster DRAM access.
Brandon
01-03-2019 08:57 AM
You need to quque up multiple requests to get anywhere near good throughput. This is a typical "problem" with DRAM, the first address takes a lot longer than the subsequent addresses when queued.