Hi,
Currently we have a TI C66x demo card exercising SRIO to move data back and forth to our custom card that has a big FPGA on it. The interface is SMC on the demo card side and XMC on the FPGA side. The high speed data comes from the FPGA and we want to transfer it to DDR3 on the C66x card.
We want the FPGA to initiate Ftype 6 packets and fill these with data and send them back over the SRIO to the C66x and then have it automatically move that data into DDR3. We are not sure how this automatic movement should work for best operation... We think MAU to DDR3?
The basic packet forming and bi-direction data transfer over SRIO is working, but the step where the Ftype 6 packets moving data to the DDR3 is not working. Currently we have the data going directly into L2 SRAM rather into DDR3. We do not have the DDR3 core initialized and aren't sure how to do this. We just found out about the Platform Library. Is this what we should be using to configure the DDR3?
We think that by using direct I/O and the MAU, we should be able to take the incoming data and make it automatically transfer into DDR3 rather than L2 Cache which we think is happening now. The crux of our problem is how to get incoming data into DDR3.
We have looked at the following e2e thread and are confused:
http://e2e.ti.com/support/dsp/c6000_multi-core_dsps/f/639/t/215645.aspx
The Demo code TI has (that we have found) uses Ftype 9 packets and is of limited use.
Does TI have any source code that exercises the MAU or shows MAU configuration. Also, does TI have code that sets up the DDR3 core. The eval doesn’t appear to do this? Typically eval bds supply the initialization code and for the DDR3 on the eval? Is this in the Platform Library. Is this what we should be using to configure the DDR3?
We are going to try and see if data can be moved automatically by the MAU to DDR3 based on writing to the DDR3 address range alone, but aren’t certain if this is the correct approach (or if more automatic configuration is needed).