OK, I think that I have solved it. A vector size of 2048 seems to not
work in RFNoC, no matter who wrote the block. When I dial back to 256
that seems to work. I changed my block to output a tlast every 256
samples, and data passes through fine now.
Is there a way to make 2048 work in RFNoC? Is it an Ethernet frame
issue? I feel like somewhere back in the day I heard that all the data
needs to fit into a frame size, so maybe 1024 is a max frame size in
this case...
On 09/11/2017 03:06 PM, Jason wrote:
One other piece of information. It seems like I always get hung up on
the same output sample, 6905860. I noticed that that is almost
exactly 3372 packets of 2048 sample vectors. Is there a set number of
samples that can be buffered before the system stops? I set the MTU
to be 11 in the axi_wrapper parameter list, but no dice.
When I get stuck in this state (and it is always is the same way),
i_tvalid is high, and i_tready is low (because I am stalled waiting on
my output to catch up). o_tvalid is high and o_tready is low. For
some reason o_tready never goes high at this point.
On 09/11/2017 02:05 PM, Jason wrote:
I have a block that I've been working on and it seems like I am
missing something vital to get it to work properly. My block takes
in a vector of 2048 samples and outputs 2048 samples as well. I
simulated it and it passes simulation fine.
But, if I look at the output as it goes to a file, nothing goes
through. I opened up wireshark and I see 11-12 packets comes from my
PC to RFNoC, and then 1 60B packet going RFNoC to host, 1 58B from
host to RFNoC, and then another 60B from RFNoC to host
(wash-rinse-repeat).
The packets to the host seem to be incrementing as they should, but
are obviously not full packets. An example of two packets from RFNoC
to host are:
00:00:00:05:00:00:04:dd:00:00:60:18:59:b6:cb:0b
00:00:00:05:00:00:04:de:00:00:60:1c:ca:77:b6:11
I have noticed that if I don't use my block and just have data pass
through the FIFO, I still don't see data pass through, so I am
wondering if the crossbar is getting screwed up somehow.
_______________________________________________
USRP-users mailing list
USRP-users@lists.ettus.com
http://lists.ettus.com/mailman/listinfo/usrp-users_lists.ettus.com
