In Hawq we had a functon that could do a fast row-count on AO tables based on querying the catalog versus issuing 'select count ' SQL. Was a lot faster for larger tables. I found the script and tried using it with HAWQ but one of the underlying Postgres functions that gets called (get_ao_distribution) bombs since it doesn't seem to be able to find the table files when they are stored in HDFS.
gpadmin=# select sum(tupcount) from get_ao_distribution('public.foo');
ERROR: could not open relation 1663/24731/24740: No such file or directory (seg0 localhost.localdomain:40000 pid=82964)
DETAIL: Database directory "base/24731" does not exist
CONTEXT: SQL statement "select gp_segment_id, sum(tupcount) from gp_dist_random('pg_aoseg.pg_aoseg_24738') group by (gp_segment_id)"