pow-bench

This repository contains the source code of the paper "Evaluating Memory-Hard Proof-of-Work Algorithms on Three Processors", by Zonghao Feng and Prof. Qiong Luo.

Build

The build instructions for each algorithm are as follows. There is also a build script available in the scripts folder.

cpuminer-multi (Hashcash and Cryptonight, CPU and KNL)

./build.sh

ccminer (Hashcash and Cryptonight, GPU)

./build.sh

ethash (Ethash, CPU)

mkdir build
cd build
cmake ..
cmake --build .

ethminer (Ethash, GPU)

mkdir build
cd build
cmake .. -DETHASHCL=OFF
cmake --build .

cuckoo (Cuckoo Cycle, CPU, KNL and GPU)

cd src/cuckoo
make mean29x8
make cuda29

Run

Run performance tests for each algorithm with run.py in the scripts folder (requires Python 3.3+).

cpuminer-multi

# -t [num_of_threads]
./cpuminer --benchmark -a sha256d -q -t 40
./cpuminer --benchmark -a cryptonight -q -t 40

ccminer

# -d [device_id]
./ccminer --benchmark -a sha256d -d 0
./ccminer --benchmark -a cryptonight -d 0

ethash

# -b [block_num] -i [iterations] -t [num_of_threads]
./ethash-fakeminer -b 10000 -i 50 -t 40

ethminer

# [block_num] [device_id]
./ethminer --benchmark 10000 -U --cu-devices 0

cuckoo

# -t [num_of_threads] -r [rounds]
./mean29x8 -t 64 -r 10 
./cuda29 -r 10

Profiling

Requirement: perf (CPU, KNL), nvprof (GPU)

Power consumption monitoring: s-tui (CPU, KNL) and nvidia-smi (GPU)

CPU

perf stat -e L1-dcache-load-misses,mem_uops_retired.all_loads,mem_uops_retired.l1_miss_loads,LLC-loads,l2_requests.miss,mem_uops_retired.l2_hit_loads,mem_uops_retired.l2_miss_loads,unc_e_rpq_inserts,unc_e_wpq_inserts,unc_m_cas_count.rd,unc_m_cas_count.wr

KNL

perf stat -e mem_load_uops_retired.l1_hit,mem_load_uops_retired.l1_miss,mem_load_uops_retired.l2_hit,mem_load_uops_retired.l2_miss,mem_load_uops_retired.llc_hit,mem_load_uops_retired.llc_miss,uncore_imc/data_reads/,uncore_imc/data_writes/

GPU

nvprof -m sm_efficiency, achieved_occupancy, dram_read_throughput, dram_write_throughput, shared_efficiency, shared_load_throughput, shared_store_throughput, l2_read_throughput, l2_write_throughput, sysmem_read_throughput, sysmem_write_throughput, stall_inst_fetch, stall_exec_dependency, stall_texture, stall_sync, stall_other

Sample Results

	i7-3770	Xeon Gold 5115	Xeon Phi 7210	GTX 670	Tesla K80	GTX 1080 Ti	Tesla V100
Hashcash (MH/s)	25.21	190.26	191.18	159.31	577.74	1832.95	2724.92
CryptoNight (H/s)	128.513	552.95	1005.2	91.21	478.57	780.14	1541.10
Ethash (MH/s)	0.75	2.31	3.09	16.30	27.97	32.63	94.21
Cuckoo Cycle (G/s)	0.24	0.67	0.54	-	0.85	3.92	5.39

Acknowledgement

This project is based on the following open-source projects:

• cpuminer-multi
• ccminer
• ethash
• ethminer
• cuckoo