Some questions about solution finding rates

Hello dear, I have succesfully compiled and managed some points on your nice project. Like replacing blake by sha256 and some other stuff.

I am running cuda29 modified to cuda23. cuda23: ../crypto/siphash.cuh sph_sha2.h sha2.c mean.cu Makefile $(NVCC) -o $@ -DEDGEBITS=23 -arch sm_75 mean.cu $(SHA256_SRC)

header : 00000070c71b9991827e7da2a7cd1bf845b4b5f0d8eded475ebd958ce664e8a5351f000092eca17ab0529552d0b8dd7d936460fafa8afa338906d4f9dff0d6106b5ad95859364b66710b231e2f84ce1f Nonce : 533627951

During my tests i saw that same computing was producing different results : Sometimes i find proofsize solution and sometimes not.

Using 90672KB of global memory. Headernonce : 2af24a90 nonce 533627951 k0 k1 k2 k3 29fc5401ddb4646b 66ef83343c6d6cac 13040bee013412c7 dd09f6fd012d771 nonce 533627951 k0 k1 k2 k3 3025385416677418091 7417291371677838508 1370233303477785287 995470820229961585 Seeding completed in 1 + 1 ms 62-cycle found 10-cycle found 42-cycle found findcycles edges 114 time 0 ms total 7 ms Time: 8 ms Solution 5981 12c4f 64e08 8a422 1398d3 14c7c6 15d246 1b729a 1faa0d 211305 211c37 23289d 247b48 25bb1d 27c923 2a6a6a 2c1104 2cbadc 309a14 37eb35 3ab099 3f141c 434ff0 4ef4af 533d94 5a9a73 5b1337 5c64d3 5cdeda 5f2bf3 65cc0e 67fc65 68afdc 6b7fdb 6dca11 6e42fc 702e68 703692 72a18c 73b7af 7ccd4d 7d18ed Verified with cyclehash 9018d734c2729271e3717a344ce86d373fd5437bfc2c2cf53b5de81cd98de9a0 1 total solutions Using 90672KB of global memory. Headernonce : 2af24a90 nonce 533627951 k0 k1 k2 k3 29fc5401ddb4646b 66ef83343c6d6cac 13040bee013412c7 dd09f6fd012d771 nonce 533627951 k0 k1 k2 k3 3025385416677418091 7417291371677838508 1370233303477785287 995470820229961585 Seeding completed in 1 + 1 ms 10-cycle found 42-cycle found findcycles edges 52 time 0 ms total 8 ms Time: 9 ms Solution 5981 12c4f 64e08 8a422 1398d3 14c7c6 15d246 1b729a 1faa0d 211305 211c37 23289d 247b48 25bb1d 27c923 2a6a6a 2c1104 2cbadc 309a14 37eb35 3ab099 3f141c 434ff0 4ef4af 533d94 5a9a73 5b1337 5c64d3 5cdeda 5f2bf3 65cc0e 67fc65 68afdc 6b7fdb 6dca11 6e42fc 702e68 703692 72a18c 73b7af 7ccd4d 7d18ed Verified with cyclehash 9018d734c2729271e3717a344ce86d373fd5437bfc2c2cf53b5de81cd98de9a0 1 total solutions Using 90672KB of global memory. Headernonce : 2af24a90 nonce 533627951 k0 k1 k2 k3 29fc5401ddb4646b 66ef83343c6d6cac 13040bee013412c7 dd09f6fd012d771 nonce 533627951 k0 k1 k2 k3 3025385416677418091 7417291371677838508 1370233303477785287 995470820229961585 Seeding completed in 1 + 1 ms 10-cycle found 42-cycle found findcycles edges 52 time 1 ms total 9 ms Time: 9 ms Solution 5981 12c4f 64e08 8a422 1398d3 14c7c6 15d246 1b729a 1faa0d 211305 211c37 23289d 247b48 25bb1d 27c923 2a6a6a 2c1104 2cbadc 309a14 37eb35 3ab099 3f141c 434ff0 4ef4af 533d94 5a9a73 5b1337 5c64d3 5cdeda 5f2bf3 65cc0e 67fc65 68afdc 6b7fdb 6dca11 6e42fc 702e68 703692 72a18c 73b7af 7ccd4d 7d18ed Verified with cyclehash 9018d734c2729271e3717a344ce86d373fd5437bfc2c2cf53b5de81cd98de9a0 1 total solutions Using 90672KB of global memory. Headernonce : 2af24a90 nonce 533627951 k0 k1 k2 k3 29fc5401ddb4646b 66ef83343c6d6cac 13040bee013412c7 dd09f6fd012d771 nonce 533627951 k0 k1 k2 k3 3025385416677418091 7417291371677838508 1370233303477785287 995470820229961585 Seeding completed in 1 + 1 ms 10-cycle found findcycles edges 82 time 0 ms total 8 ms Time: 8 ms 0 total solutions Resultats ok : 40 / 100 ----------------------------------------------------------

about 40/60% of my runs of same header and same nonce find solutions.

Is it a normal behaviour of this computing? Is there a way to improve that even if i loose time?

I use these parameters :

params.genablocks= 256; params.genatpb = 256; params.genbtpb = 512; params.trimtpb = 512; params.tailtpb = 1024; params.recoverblocks = 1024; params.recovertpb =1024;

and i modified #define FLUSHA 32

to be honnest without knowing what that will affect.

I am working on 23 edgebits.

I hope you will be able to guide me a bit.

Regards

tromp / cuckoo

Some questions about solution finding rates #112