In the next hard fork, should we tweak RandomWOW to improve verification time? #47

RandomWOW proposed changes

RANDOMX_DATASET_BASE_SIZE  268435456 // 256 MB
RANDOMX_DATASET_EXTRA_SIZE 16777216 // 16 MB
RANDOMX_PROGRAM_SIZE       384
RANDOMX_PROGRAM_ITERATIONS 400
RANDOMX_PROGRAM_COUNT      8
RANDOMX_SCRATCHPAD_L3      262144 // 256 KB
RANDOMX_SCRATCHPAD_L2      131072 // 128 KB
RANDOMX_SCRATCHPAD_L1      65536 // 64 KB

RANDOMX_FREQ_FSWAP_R        4
RANDOMX_FREQ_FADD_M         4
RANDOMX_FREQ_FSUB_M         3
RANDOMX_FREQ_FSCAL_R        3
RANDOMX_FREQ_FDIV_M         8
RANDOMX_FREQ_FSQRT_R       12

./randomx-benchmark --verify --auto

WOW v1	WOW v2	Time Decrease
17.02 ms/h	3.23 ms/h	~81%

CPU miners get ~130% hashing boost

If it cuts verification time by 80% I think we should probably go for it. I don't think the decreased complexity or hardware requirements are going to entice anyone to build an ASIC or FPGA implementation in the near term.