2 to 5 solutions problem or not

[Zwilla]

./cuda29_GTX_1080 -h 324D6F373754576B3546684B772F493865493343416B71497531492B624A4B6155567A6543424B3278414D3D -n 18335313330172576564 -r 120 -g
...
2 total solutions of 120 nonces

---

<0.2010.13>@aec_pow_cuckoo:generate_int:203 Executing cmd: "./cuda29_GTX_1070 -h 324D6F373754576B3546684B772F493865493343416B71497531492B624A4B6155567A6543424B3278414D3D -n 18335313330172576564 -r 120 -g -d 1"
...
2018-12-12 14:27:15.201 [debug] <0.1778.13>@aec_pow_cuckoo:parse_generation_result:481
5 total solutions of 120 nonces

---

two solutions on cuda29_GTX_1080 uses the original code

five solution on cuda29_GTX_1070 uses an modified code

Header and start nonce always the same just the solution differs.

I can repeat the same on any nonce and any header, my modified code, which is not a part of the uploaded binaries produces by factor 2,5 to 5 more solutions as the original driver.

Is there any chance to test if I can get with such a modification a valid AE sol?

[tromp]

I cannot reproduce the above runs as my solvers use a different nonce location.
What is the difference in source code that produces the above difference?
Do the additional solutions pass the verification?

Ok; let me clone your repo and try to reproduce...

Hmm, not very build friendly:-(
How to reproduce?