I wonder if somebody could explain what the figures for hashing performance mean. I’m a bit baffled.
Raspberry Pi model B at mine journalctl --since=04:03:03 --until=04:03:10 --no-pager /usr/bin/syncthing – Logs begin at Mon 2017-05-15 11:36:12 BST, end at Fri 2017-06-23 14:12:51 BST. – Jun 23 04:03:03 pi nice[481]: [3YBZS] INFO: Single thread SHA256 performance is 2.5 MB/s using minio/sha256-simd (2.5 MB/s using crypto/sha256). Jun 23 04:03:05 pi nice[481]: [3YBZS] INFO: Hashing performance with weak hash is 2.67 MB/s Jun 23 04:03:08 pi nice[481]: [3YBZS] INFO: Hashing performance without weak hash is 2.65 MB/s Jun 23 04:03:09 pi nice[481]: [3YBZS] INFO: Weak hash enabled, as it has an acceptable performance impact. Jun 23 04:03:09 pi nice[481]: [3YBZS] INFO: Local networks: 127.0.0.1/8, 192.168.1.1/24, ::1/128, fe80::ba27:ebff:feea:d01b/64 root@pi:~#
Raspberry Pi Zero W at son’s
journalctl --since=04:22:49 --until=04:22:56 --no-pager /usr/bin/syncthing
-- Logs begin at Mon 2017-05-15 11:42:38 BST, end at Fri 2017-06-23 14:13:13 BST. --
Jun 23 04:22:49 pie nice[453]: [B3IQ5] INFO: Single thread SHA256 performance is 1.7 MB/s using crypto/sha256 (1.7 MB/s using minio/sha256-simd).
Jun 23 04:22:53 pie nice[453]: [B3IQ5] INFO: Hashing performance with weak hash is 2.24 MB/s
Jun 23 04:22:55 pie nice[453]: [B3IQ5] INFO: Hashing performance without weak hash is 3.74 MB/s
Jun 23 04:22:55 pie nice[453]: [B3IQ5] INFO: Weak hash disabled, as it has an unacceptable performance impact.
root@pie:~#
Laptop at mine
journalctl -b --no-pager /usr/bin/syncthing|grep -i performance
Jun 21 18:07:44 monocerus nice[586]: [IPWVZ] INFO: Single thread SHA256 performance is 122 MB/s using minio/sha256-simd (109 MB/s using crypto/sha256).
Jun 21 18:07:45 monocerus nice[586]: [IPWVZ] INFO: Hashing performance with weak hash is 99.24 MB/s
Jun 21 18:07:45 monocerus nice[586]: [IPWVZ] INFO: Hashing performance without weak hash is 117.15 MB/s
Jun 21 18:07:45 monocerus nice[586]: [IPWVZ] INFO: Weak hash enabled, as it has an acceptable performance impact.
Jun 22 10:06:37 monocerus nice[19450]: [IPWVZ] INFO: Single thread SHA256 performance is 130 MB/s using minio/sha256-simd (112 MB/s using crypto/sha256).
Jun 22 10:06:37 monocerus nice[19450]: [IPWVZ] INFO: Hashing performance with weak hash is 111.49 MB/s
Jun 22 10:06:38 monocerus nice[19450]: [IPWVZ] INFO: Hashing performance without weak hash is 126.95 MB/s
Jun 22 10:06:38 monocerus nice[19450]: [IPWVZ] INFO: Weak hash enabled, as it has an acceptable performance impact.
Jun 22 10:37:57 monocerus nice[21510]: [IPWVZ] INFO: Single thread SHA256 performance is 127 MB/s using minio/sha256-simd (109 MB/s using crypto/sha256).
Jun 22 10:37:58 monocerus nice[21510]: [IPWVZ] INFO: Hashing performance with weak hash is 111.10 MB/s
Jun 22 10:37:58 monocerus nice[21510]: [IPWVZ] INFO: Hashing performance without weak hash is 125.45 MB/s
Jun 22 10:37:58 monocerus nice[21510]: [IPWVZ] INFO: Weak hash enabled, as it has an acceptable performance impact.
Jun 22 10:46:26 monocerus nice[22057]: [IPWVZ] INFO: Single thread SHA256 performance is 116 MB/s using minio/sha256-simd (89 MB/s using crypto/sha256).
Jun 22 10:46:26 monocerus nice[22057]: [IPWVZ] INFO: Hashing performance with weak hash is 111.54 MB/s
Jun 22 10:46:27 monocerus nice[22057]: [IPWVZ] INFO: Hashing performance without weak hash is 125.95 MB/s
Jun 22 10:46:27 monocerus nice[22057]: [IPWVZ] INFO: Weak hash enabled, as it has an acceptable performance impact.
Jun 23 10:24:43 monocerus nice[24863]: [IPWVZ] INFO: Single thread SHA256 performance is 130 MB/s using minio/sha256-simd (111 MB/s using crypto/sha256).
Jun 23 10:24:44 monocerus nice[24863]: [IPWVZ] INFO: Hashing performance with weak hash is 111.49 MB/s
Jun 23 10:24:44 monocerus nice[24863]: [IPWVZ] INFO: Hashing performance without weak hash is 126.29 MB/s
Jun 23 10:24:44 monocerus nice[24863]: [IPWVZ] INFO: Weak hash enabled, as it has an acceptable performance impact.
root@monocerus:~#
The results on the Pis make sense to me.
On pi, weak hash is enabled when it gives more throughput with weak hash. On pie, weak is disabled when it gives less throughput.
All looks reasonable so far, but it all changes when we look at monocerus. Weak hash degrades performance, but it is disabled.
I’m looking at the hashing performance lines here. Maybe these are only looked at when the single thread line gives equal numbers??? But if so, why bother to run the further test?
I don’t understand 