Nodupkey returns different outputs

[MarkNicholas]

Dear Team,

I have this issue with nodupkey option in proc sort.It seems to return different results on different servers.

proc sort data=comtest.x2 nodupkey; by comp_id; run;

My source data sample with 2 records with same comp_id:
comp_id-  business - prod_id  -PTY_ID
1           -  ABC -     123A       -   ASHVIK
1           -          -     245CA     -   ASHVIK

After removing duplicates on one server:
comp_id-  business - prod_id  -PTY_ID

1           -  ABC -     123A       -   ASHVIK

After removing duplicates on another server:

comp_id-  business - prod_id  -PTY_ID

1           -          -     245CA     -   ASHVIK

I want to know why we are getting different results on different servers .

Cheers
-Mark

[gergely_batho · Accepted solution]

PROC SORT is multi-threaded (by default). The allocation of data to different threads and the actual conditions can effect which observation comes first to the output.

Results can be different even on the same system.

If you want to get identycal results use the equals option.