oh yay, undocumented flags required for the surf miner, to the source code cave

``` selected_field=8 waves=1 hash_threads=1 work_unit_mem_gb=1 ``` let us see :stuck_out_tongue:

i guess i have a hunch, but from the horse's mouth, what's an 'unit' in the context of the surf miner

`LVM | vg0-home | busy 95% | read 49992 | write 72 | KiB/r 127 | KiB/w 216 | MBr/s 104.1 | MBw/s 0.3 | avq 1.98 | avio 1.13 ms |` it is maxing out the raid1 it is reading from

so i guess it will not have the time to complete a scan enough many times to be useful at all

at 100MB/s

some explanation on the three required tweakables would be welcome to try to reason about it, what does this enable for the non-ram and potentially even sata ssd crowd

and this could use a comment https://github.com/snowblossomcoin/snowblossom/blob/master/miner/src/surf/SurfMiner.java#L471-L474 ``` private int getRecordSize() { return 4+1+8+12+32; } ```

so i guess waves are sequential scans, and i do not know why i'd want to have more than one on a spinny disk

nor more threads doing hashing either as nothing is feeding them

but cannot figure out what the work unit size is, seems to just feed into whatever the targets juggling is

also what is up with the average calcuations? :smile: ``` INFO: Shares: 0 (rejected 0) (blocks 0) tammikuuta 31, 2019 12:24:14 AP. snowblossom.miner.surf.SurfMiner printStats INFO: 1-min: 0,000/s, 5-min: 0,003/s, hour: 0,000/s tammikuuta 31, 2019 12:24:14 AP. snowblossom.miner.surf.SurfMiner printStats INFO: No hashing, and we have a good work unit from the pool. So probably something else wrong. ``` ``` INFO: Shares: 0 (rejected 0) (blocks 0) tammikuuta 31, 2019 12:27:14 AP. snowblossom.miner.surf.SurfMiner printStats INFO: 1-min: 0,017/s, 5-min: 0,007/s, hour: 0,001/s ```

i suppose it can actually have had found a candidate

It takes a while to get started

and slowly rising, i'll leave it overnight, gonna be entertaining to see if it breaks one hash per second

Like 5 min

but am i correct in assuming what the 'wave' is?

so there is just some equally spaced offset for the sequential scans when you have more than one?

and one should only bump the hashing threads if one saturates a cpu?

but what's the work unit sizing stuff

ok, so the idea is it builds up some crescendo towards the end of the block interval and jams a shitton of shares in

so, not for spinny disks