UsenetSearch/thread_timing.txt

29 lines
1.2 KiB
Plaintext

time to process 100026 headers
batch size | threads | time
-----------+---------+-------
1 | 1 | 3m 44s
5 | 2 | 4m 51s
25 | 2 | 3m 25s
50 | 2 | 3m 4s
1000 | 2 | 2m 29s
5000 | 2 | 2m 26s
25006 | 4 | 2m 3s (batch size = #headers/#threads)
3125 | 8 | 1m 19s
6251 | 16 | 1m 18s (batch size = #headers/#threads)
3125 | 16 | 1m 9s (batch size = #headers/#threads/2)
1562 | 16 | 1m 8s (batch size = #headers/#threads/4)
781 | 16 | 1m 6s (batch size = #headers/#threads/8)
500 | 16 | 1m 5s
390 | 16 | 1m 9s (batch size = #headers/#threads/16)
125 | 32 | 1m 7s
250 | 32 | 1m 3s
500 | 32 | 1m 5s
781 | 32 | 1m 5s
781 | 50 | 1m 3s
1000 | 100 | 1m 21s
781 | 200 | 1m 44s
(this is on a 16 core machine)
in any event - now that i've got that working - i should probably go back to either optimizing the database i have or just switch to pgsql or some shit
or both