/poggers

A library of high performance data structures for cuda. This library is header only and can be added to a project using the C package manager.

Primary LanguageCuda

POGGERS: Portable Optimized GPU Generic Structures

POGGERS is a header-only library of high-performnace CUDA data structures.

This library is composed of template objects that allow you to quickly build and modify a modular table while maintaining high performance.

The end goal of the library is to add simple-to-use data structures that simplify the use of CUDA and bring a syntax similar to mainline C++.

For now, the only features that are supported are the hash table and filter components.

FEATURES

  1. Hash Tables and Filtering (TCF)
    • A set of high-performance template components for building key-value data strucutures.
    • Modular components to specify data layout, access patterns, and false-positive rate.
  • Recursive template structure lets you build efficient multi-layer tables with minimal effort.

Adding POGGERS

To add poggers to a project, you can use the CMake Package Manager

To add CPM, add

include(cmake/CPM.cmake)

to your cmake file.

To add poggers, include the following snippet and select a version.

If you remove the version tag, CPM will pull the most up-to-date build.

CPMAddPackage(
  NAME poggers
  GITHUB_REPOSITORY huntermBerkeley/poggers
  GIT_TAG origin/main
  VERSION v1.0
)

To cache the library, specify a download folder:

set(CPM_SOURCE_CACHE "${CMAKE_CURRENT_SOURCE_DIR}/downloaded_libraries")

Building Tests

There are a series of optional tests that can be included with the build.

To build the tests, specify the -DPOGGERS_BUILD_TESTS=ON flag to CMake.

FUTURE WORK

  1. Dynamic GPU Allocation - In Progress!
  2. Host-Device pinned device-side communication
  3. Sparse ML components.

PUBLICATIONS