Skip to content

A benchmark suite for exploring the micro-architecture of GPGPU based on blockchain workloads

License

Notifications You must be signed in to change notification settings

ChaseLab-PKU/BcBench

Repository files navigation

BcBench

BcBench is for evaluating blockchain workloads on throughput processors(like GPGPU) at micro-architecture level. BcBench contains a collection of blockchain-based cryptocurrency apps with different characteristics. BcBench mainly focuses on CUDA platform and can be run on top of real hardware as well as simulators (like GPGPU-Sim).

BcBench makes several modifcations to support blockchain workloads with GPGPU-Sim. The modified GPGPU-Sim is in accel-sim-framework-1.2.0.tar.gz. You can build GPGPU-Sim following the GitHub page of accel-sim-framework.

The BcBench workload repo contains two folders of workloads, namely blockmark and equihash. blockmark conatins workloads NeoScrypt, CryptoNight, X11, and equihash contains workload equihash.

Build

We recommend to put blockmark and equihash in the /root folder, or in a docker container environment.

To build the workloads of BcBench, please execute the following commands with corresponding workload:

./blockmark/make test_cryptonight
./blockmark/make test_neocrypt
./blockmark/make test_all_hash
./equihash/make eqcudah

Run

To run the workloads of Bcbench inside GPGPU-Sim, please execute the following commands:

./blockmark/test_neoscrypt.sh
./blockmark/test_crypto.sh
./blockmark/test_hash.sh
./equihash/eqcudah

You may need to set the correct path of GPGPUSIM_ARCH_PATH as well as modify the corresponding parts of the test script.

Data Prepration

For fast evaluation on simulators, we pre-computed some data of CryptoNight. The binary data file can be downloaded from Google Drive and then put in the path BcBench/blockmark/data.bin.

Publication

The BibTex is shown below:

@inproceedings{10.1145/3555776.3577701,
author = {Pan, Xiurui and Chen, Yue and Yi, Shushu and Zhang, Jie},
title = {BcBench: Exploring Throughput Processor Designs Based on Blockchain Benchmarking},
year = {2023},
doi = {10.1145/3555776.3577701},
booktitle = {Proceedings of the 38th ACM/SIGAPP Symposium on Applied Computing},
pages = {88–97},
numpages = {10},
location = {Tallinn, Estonia},
series = {SAC '23}
}

Contact

Feel free to contact panxiurui at outlook dot com if you have any questions.

About

A benchmark suite for exploring the micro-architecture of GPGPU based on blockchain workloads

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published