Quote from: dcct on November 06, 2014, 10:31:28 AM

Quote from: bsunau7 on November 06, 2014, 02:18:36 AM

Anyone else been keeping an eye on testnet?

Difficulty is rising pretty fast, which means some serious mining from a very small base is happening.

Regards,

--
bsunau7

I see a difficulty below 17. Thats <100k pps total.

Let me show you how to optimize the miner:

My I7-4770k with 8 threads, stock-miner:
pps: 31821 / 31160 10g/h 4473 / 3429 15g/h 0 / 0

Now go on and build your own libgmp from source. When done, run the tune utility for optimum results. Copy your final libgmp.so.10 to /usr/local/lib/.

Now we're at:
pps: 34039 / 34041 10g/h 6077 / 4479 15g/h 0 / 0

Optimizing sieve size:
open src/main.cpp

line 252:
1048576);

to
8 * 1048576);

and line 256
500000);

to
15000000);

make sure to also change the shift, otherwise you wont be able to submit most blocks! You could specify it on command line, but this is safer:

line 128:
uint16_t shift = (opts->has_shift() ? atoi(opts->get_shift().c_str()) : 20);

change to:
uint16_t shift = (opts->has_shift() ? atoi(opts->get_shift().c_str()) : 23);

Now its mining at
pps: 38276 / 37068 10g/h 5156 / 5265 15g/h 0 / 0

Lets also add the tuning for Haswell AVX2:
open Makefile add set

OTFLAGS = -march=core-avx2 -mtune=core-avx2 -O3
(if you have a haswell, of cause)

Results:
pps: 39691 / 39553 10g/h 6059 / 5635 15g/h 0 / 0

Oh yeah. 20% for free.
Thanks dcct.