Question about FPP and DDP messages.

[cschill2020]

Hi,

I am working on a side project building a webapp to send realtime ddp to a kulp64D BBB board running FPP.  I am sending DDP via UDP.   My project is using ws2811 12v strips with 3 leds per individually addressable pixel.  Currently I have ~10k individually addressable pixels.  

I have spent some time in the FPP code, but not to a point where I am rebuilding/testing/debugging.  I have a question about the DDP code in FPP.  Specifically this line:
https://github.com/FalconChristmas/fpp/blob/3cf77f0b8a9edc1e0cd336070dabc690d5c4f66d/src/e131bridge.cpp#L63C1-L64C1
"#define MAX_MSGS 64"

Does this mean that FPP will process at most 64 messages per frame?  This would limit DDP to ~32,000 individually addressable pixels.  I am assuming 500pixels per message.   

Maybe more likely is that this is just the max number of messages that can be stored in a buffer.  But I am curious if this is limiting the number of pixels that can be controlled over UDP.  

Thanks to the community for all the support!  It's so cool to be able to imagine such big projects. 

[dkulp]

When we get a notification from Linux that there is data on the socket to be read, there are a couple of ways we could retreive the data.   The simple method would be to just call recvmsg or similar to pull one packet off the socket, process it, call recvmsg again, etc... until all the packets are read off and processed.   That involves a bunch of trips into the linux kernel.   Instead, we use the recvmmsg call to grab up to 64 packets in one single call.  We then process the 64 packets and then immediately call recvmmsg again to grab more packets if there are more there.   It has nothing to do with frame rates or number of packets per frame or anything like that.  It's just the number of messages/packets to transfer from Linux's network buffers  into the application space in one call.   

[cschill2020]

Oh got it, super helpful.  Thank you so much for the quick answer.  Ill take a look a closer look at the code FPP is using to send packets and try to mimic it in my app.  I am sure there are improvements I can make to optimize how I am sending the pixel data to FPP...