Skip to content
/ HarkDB Public

Multi-backend GPU query engine written with Futhark

Notifications You must be signed in to change notification settings

philass/HarkDB

Repository files navigation

HarkDB

Build Status

HarkDB is a SQL Query Engine accelerated with the GPU. HarkDB is written in Futhark, which allows it to compile down to a CUDA, OpenCL, or sequential C backend. HarkDB aims to run extremely fast on GPU's leveraging the code optimization of the futhark compiler.

Due to compiler limitations the query engine works with homogenous numeric data. It is SQL compatible and supports the following Clauses

  • Select
  • From
  • Where
  • Group By
  • Having
  • Sort By

HarkDB aims to have very similary calling semantics to BlazingDB.

Requirements

Installation and Building

Installing and building this git repository is as simple as

git clone https://github.com/philiplassen/HarkDB.git && cd HarkDB
./setup.sh

Usage

An example program using HarkDB

from FutharkContext import FutharkContext
import time

fc = FutharkContext()
fc.create_table('game_1', 'data.csv')
sel = fc.sql("select col1, col3 from game_1")
print(sel)

About

Multi-backend GPU query engine written with Futhark

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published