parallel, splitting input files and starting

[Christian Meesters]

Hi and a happy new year to everyone reading this list,

Well, I would like to call a program with "parallel" where each instance 
of the program takes a certain chunk of the work of two input files. One 
condition is that the input files or the created chunks need to have a 
multiple of 4 lines (e.g. 1e6 lines each), because each record to work 
on consists of 4 lines. There are two big input files and the first 
solution I came up with was using "split" and then creating a file list 
and starting the tasks with "parallel". The second, slightly faster 
solution was to use "split" to split the input files and inotifywait and 
bash-only to trigger the concurrent tasks keep track of the running tasks.

My first question is: Is "parallel" able to read a given number of lines 
from two input files and distribute the work without using "split", 
whilst having the advantage of inotifywait.

I tried

parallel --no-notice --max-lines=200 -j x cmd 
some/path/to/a/global/reference {1} {2} ::: ./input1 ::: ./input2

which starts one instance of cmd and not x instances.

Hm, I tinkered with --pipe and the options in this thread: 
http://unix.stackexchange.com/questions/66463/using-gnu-parallel-with-split 
- to no avail. With --pipe and --max-lines "parallel" starts two perl 
processes to read the buffers and does not proceed.

Any ideas on how this could be achieved with two input files?

The second question would be: If this more direct approach does not 
work. Could "parallel" support inotify? If so, how?

And a last unrelated question is: If I start several tasks with 
"parallel" on several hosts of a cluster, supplying -S with a host list 
from a batch system - does parallel copy the input files to the hosts or 
- supplied with a full path - can it use a parallel file system? Well, I 
think I need to supply a bash function, which extracts the path from a 
file list, which is copied, but perhaps there is an idea how a parallel 
file system could be used?

I hope I was understandable.

Best regards,

Christian Meesters