There is no concept of row "numbering" within Teradata, so you can't get the odd or even numbered rows.
If you want to get every other row from a particular query, you can do something like:
select * from tablex qualify sum(1) over (rows unbounded preceding) mod 2 = 1;
This will give you all odd-numbered rows that would have been returned from the query without the "qualify". However, every time I execute this, I could get a different set of rows (although the number of rows would stay consistent).
A record is first distributed to an AMP based on the rowhash of the primary index column values. And with in an AMP it's sorted logically in the order of Partition (if it has PPI) and then by the rowid (which is a combination of rowhash + uniqueness value).