I’m sorry, but it was two years ago, and I’ve been lucky enough not to deal with Displayport since. So I don’t remember much about it at the moment.

Great blog, great article. What I find peculiar in this pixel packing approach, is that in the blanking regions we’re basically transmitting dummy symbols (not even packed in TUs if I am not mistaken) and the whole M_vid counting is rather theoretical / arbitrary there. For example, should we still “mimic” the stuffing or should we increment M on every link clock cycle?

Moreover, if I understand correctly, things need to be set such that at each scan-line the number of stream pixels will end before the link symbols so that some margin will be left for the asynchronous corrections. What’s the size of this margin? Where would it affect the receiver? (a large margin would mean that M_vid will not change over these intervals of every line – is that an issue?)

Perhaps I am confused about who this control-flow mechanism work in the async mode..