No. Inside the code for ecdf.m, you can see the line that explicitly creates an error message if the input is not a vector. I won't post the actual code, but you can do
to see for yourself. (The line is near the top of the file, and commented.)
If you wanted to, I suppose you could create your own "parallelized" version of ecdf (by copying and editing ecdf), in which you loop through your matrix internally within your new function. But that has its own inelegancies.