Topic: SILENTARMY v5: Zcash miner, 115 sol/s on R9 Nano, 70 sol/s on GTX 1070 - page 77. (Read 209309 times)

#define PARAM_N				200
#define PARAM_K				9
#define PREFIX                          (PARAM_N / (PARAM_K + 1))
#define NR_INPUTS                       (1 << PREFIX)
// Approximate log base 2 of number of elements in hash tables
#define APX_NR_ELMS_LOG                 (PREFIX + 1)
// Number of rows and slots is affected by this. 20 offers the best performance
// but occasionally misses ~1% of solutions.
#define NR_ROWS_LOG                     19

// Setting this to 1 might make SILENTARMY faster, see TROUBLESHOOTING.md
#define OPTIM_SIMPLIFY_ROUND			0

// Make hash tables OVERHEAD times larger than necessary to store the average
// number of elements per row. The ideal value is as small as possible to
// reduce memory usage, but not too small or else elements are dropped from the
// hash tables.
//
// The actual number of elements per row is closer to the theoretical average
// (less variance) when NR_ROWS_LOG is small. So accordingly OVERHEAD can be
// smaller.
//
// Even (as opposed to odd) values of OVERHEAD sometimes significantly decrease
// performance as they cause VRAM channel conflicts.
#if NR_ROWS_LOG == 16
#define OVERHEAD                        3
#elif NR_ROWS_LOG == 18
#define OVERHEAD                        3
#elif NR_ROWS_LOG == 19
#define OVERHEAD                        8
#elif NR_ROWS_LOG == 20 && OPTIM_SIMPLIFY_ROUND
#define OVERHEAD                        6
#elif NR_ROWS_LOG == 20
#define OVERHEAD                        9
#endif

#define NR_ROWS                         (1 << NR_ROWS_LOG)
#define NR_SLOTS            ((1 << (APX_NR_ELMS_LOG - NR_ROWS_LOG)) * OVERHEAD)
// Length of 1 element (slot) in bytes
#define SLOT_LEN                        32
// Total size of hash table
#define HT_SIZE				(NR_ROWS * NR_SLOTS * SLOT_LEN)
// Length of Zcash block header, nonce (part of header)
#define ZCASH_BLOCK_HEADER_LEN		140
// Offset of nTime in header
#define ZCASH_BLOCK_OFFSET_NTIME        (4 + 3 * 32)
// Length of nonce
#define ZCASH_NONCE_LEN			32
// Length of encoded representation of solution size
#define ZCASH_SOLSIZE_LEN		3
// Solution size (1344 = 0x540) represented as a compact integer, in hex
#define ZCASH_SOLSIZE_HEX               "fd4005"
// Length of encoded solution (512 * 21 bits / 8 = 1344 bytes)
#define ZCASH_SOL_LEN                   ((1 << PARAM_K) * (PREFIX + 1) / 8)
// Last N_ZERO_BYTES of nonce must be zero due to my BLAKE2B optimization
#define N_ZERO_BYTES			12
// Number of bytes Zcash needs out of Blake
#define ZCASH_HASH_LEN                  50
// Number of wavefronts per SIMD for the Blake kernel.
// Blake is ALU-bound (beside the atomic counter being incremented) so we need
// at least 2 wavefronts per SIMD to hide the 2-clock latency of integer
// instructions. 10 is the max supported by the hw.
#define BLAKE_WPS               	10
// Maximum number of solutions reported by kernel to host
#define MAX_SOLS			500
// Length of SHA256 target
#define SHA256_TARGET_LEN               (256 / 8)

// Optional features
#undef ENABLE_DEBUG

/*
** Return the offset of Xi in bytes from the beginning of the slot.
*/
#define xi_offset_for_round(round)	(8 + ((round) / 2) * 4)

// An (uncompressed) solution stores (1 << PARAM_K) 32-bit values
#define SOL_SIZE			((1 << PARAM_K) * 4)
typedef struct	sols_s
{
    uint	nr;
    uint	likely_invalids;
    uchar	valid[MAX_SOLS];
    uint	values[MAX_SOLS][(1 << PARAM_K)];
}		sols_t;

Total 281.9 sol/s [dev0 44.4, dev1 50.3, dev2 48.8, dev3 45.2, dev4 46.4, dev5 44.2] 92 shares
Total 281.7 sol/s [dev0 43.4, dev1 49.4, dev2 48.2, dev3 47.0, dev4 47.8, dev5 43.2] 92 shares
Total 282.1 sol/s [dev0 43.4, dev1 51.4, dev2 47.0, dev3 45.2, dev4 46.6, dev5 43.1] 92 shares
Total 282.8 sol/s [dev0 42.8, dev1 52.5, dev2 47.6, dev3 45.8, dev4 47.4, dev5 44.2] 92 shares
Total 281.7 sol/s [dev0 43.0, dev1 49.6, dev2 47.8, dev3 45.0, dev4 47.8, dev5 44.5] 93 shares
Total 282.1 sol/s [dev0 43.6, dev1 50.4, dev2 46.8, dev3 45.7, dev4 49.3, dev5 42.3] 93 shares
Total 281.0 sol/s [dev0 44.2, dev1 48.6, dev2 47.0, dev3 46.1, dev4 48.5, dev5 44.4] 94 shares
Total 282.7 sol/s [dev0 43.8, dev1 48.6, dev2 48.7, dev3 48.4, dev4 47.7, dev5 45.5] 95 shares
Total 283.0 sol/s [dev0 44.5, dev1 49.1, dev2 46.5, dev3 47.7, dev4 49.8, dev5 47.1] 95 shares
Total 282.3 sol/s [dev0 44.3, dev1 48.1, dev2 47.0, dev3 47.4, dev4 49.7, dev5 50.0] 96 shares
Total 281.3 sol/s [dev0 43.3, dev1 47.6, dev2 45.4, dev3 48.2, dev4 50.4, dev5 50.0] 96 shares
Total 283.6 sol/s [dev0 43.6, dev1 51.2, dev2 46.9, dev3 47.4, dev4 50.8, dev5 50.4] 97 shares
Total 284.4 sol/s [dev0 43.5, dev1 48.3, dev2 47.4, dev3 48.2, dev4 50.1, dev5 50.3] 97 shares
Total 284.2 sol/s [dev0 42.8, dev1 47.0, dev2 47.4, dev3 50.0, dev4 48.1, dev5 49.4] 98 shares
Total 283.5 sol/s [dev0 42.4, dev1 50.2, dev2 46.8, dev3 48.8, dev4 49.8, dev5 48.6] 98 shares
Total 283.6 sol/s [dev0 43.4, dev1 49.8, dev2 49.3, dev3 47.4, dev4 47.8, dev5 51.5] 99 shares
Total 283.4 sol/s [dev0 41.3, dev1 50.4, dev2 49.6, dev3 48.5, dev4 48.7, dev5 50.9] 100 shares
Total 282.2 sol/s [dev0 39.5, dev1 50.9, dev2 46.1, dev3 48.0, dev4 48.7, dev5 52.7] 100 shares

#define NR_ROWS_LOG                     19

#elif NR_ROWS_LOG == 19
#define OVERHEAD                        8

#define MAX_SOLS                        500

LD_PRELOAD="./libtime.so" ./silentarmy --instances=4 --use=0,1,2,3,4,5

Total 212.6 sol/s [dev1 41.3, dev2 44.3, dev3 38.4, dev4 46.5, dev5 39.0] 21 shares
Total 213.8 sol/s [dev1 43.3, dev2 44.2, dev3 41.6, dev4 45.0, dev5 38.7] 21 shares
Total 213.9 sol/s [dev1 43.6, dev2 43.7, dev3 41.6, dev4 43.6, dev5 37.2] 22 shares
Total 213.4 sol/s [dev1 45.2, dev2 41.4, dev3 41.5, dev4 41.7, dev5 37.1] 22 shares
Total 213.6 sol/s [dev1 44.3, dev2 42.6, dev3 41.4, dev4 42.7, dev5 37.9] 22 shares
Total 212.6 sol/s [dev1 45.4, dev2 41.6, dev3 41.1, dev4 43.1, dev5 38.8] 22 shares

Total 44.7 sol/s [dev1 46.8] 1 share
Total 44.4 sol/s [dev1 47.1] 1 share
Total 45.2 sol/s [dev1 48.0] 1 share
Total 44.7 sol/s [dev1 46.8] 1 share
Total 44.9 sol/s [dev1 46.7] 1 share
Total 45.1 sol/s [dev1 47.4] 1 share
Total 45.3 sol/s [dev1 46.7] 1 share