Just for fun: managed to run it on my gtx 750. It gives 650 khs instead of 880khs with sp_'s ccminer. Not bad but cpu usage is huge.
In order to make it work I needed to change long long constants in Lyra2cl to unsigned long as long long is not supported in nvidia opencl
__constant static const sph_u64 blake2b_IV[8] = { 0x6a09e667f3bcc908UL, 0xbb67ae8584caa73bUL, 0x3c6ef372fe94f82bUL, 0xa54ff53a5f1d36f1UL, 0x510e527fade682d1UL, 0x9b05688c2b3e6c1fUL, 0x1f83d9abfb41bd6bUL, 0x5be0cd19137e2179UL };
Then it was an error in keccak1600.cl with
inline void keccak_block(ulong *s) {
I added static inline static void keccak_block(ulong *s) {
and the problems seems to vanish.
Then it was an error with 2d array init skein256.cl
__constant static const ulong ROT256[8][4] = { 46, 36, 19, 37, 33, 27, 14, 42, 17, 49, 36, 39, 44, 9, 54, 56, 39, 30, 34, 24, 13, 50, 10, 17, 25, 29, 39, 43, 8, 35, 56, 22, };
I've added more {} to form this
__constant static const ulong ROT256[8][4] = { {46, 36, 19, 37}, {33, 27, 14, 42}, {17, 49, 36, 39}, {44, 9, 54, 56}, {39, 30, 34, 24}, {13, 50, 10, 17}, {25, 29, 39, 43}, {8, 35, 56, 22} };
And it finally got compiled. It works and receives "accepted" from the pool )))
I used "intensity" : "14", "worksize" : "512", "gpu-threads" : "1" to run ... gpu_z shows 100% load for gpu engine and memory, the same is with ccminer )))
|