Andy, I did another comparison of Funakoshi vs Dstm-0.6
on a Ubuntu 16 cloud server with gtx 1070 GPU.

First I run Dstm for 2 hours and after that Funakoshi for 2 hours.
2 hours is too short time (but I wanted to test both solvers in
as close as possible period of time).

Dstm:
cpu: 6%
I/s: 237
Sol/s avg: 442 (reported by miner)
Flypool avg sol/s: 399


Funakoshi:
cpu: 16%
I/s: 240
Sol/s avg: 446 (reported by miner)
Flypool avg sol/s: 401

In the following graph from flypool,
Dstm 2 hours recording is on the left-side
and Funakoshi 2 hours recording is on the right:

    https://imgur.com/JQNKKo3

In this configuration both solvers seams to have similar
performance. I also noticed that there are many large
jumps in flypool graph in both directions.

Maybe different configurations gives completely different results.
Maybe different hours of day in flypool gives different results.

However I will continue to try to find ways to improve the performance.
Thanks for sharing your experience with me.