Skip to content

Version 4.2.0

Choose a tag to compare

View all tags
@tgrogers tgrogers released this 11 Feb 21:21
· 9 commits to dev since this release
v4.2.0
0e39753
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Verified
Learn about vigilant mode.

GPGPU-Sim Release Notes

Major Features

SST Integration

  • Added support for integrating GPGPU-Sim with the Structural Simulation Toolkit (SST)
  • Implemented SST-specific CUDA API functions and memory handling
  • Created an SST-specific GPU simulation class (sst_gpgpu_sim) for seamless integration
  • Added callbacks and synchronization mechanisms for SST communication

Stream-based Statistics

  • Implemented per-stream statistics tracking for more granular performance analysis
  • Modified cache statistics to support stream-specific data collection
  • Updated statistics printing functions to allow filtering by stream ID

Improved GCC Version Detection

  • Enhanced GCC version detection to support newer compiler versions (e.g., GCC 11+)
  • Updated version detection logic in setup scripts and makefiles

Performance Improvements

  • Modified L2 bandwidth calculation to account for different core and interconnect frequencies
  • Optimized cache statistics collection and reporting

Bug Fixes

  • Fixed SEGFAULT issue related to kernel tracking in shader cores
  • Corrected stream ID handling in various simulator components
  • Addressed issues with cache statistics aggregation and reporting

Other Enhancements

  • Added support for indirect texture access in PTX simulation
  • Improved handling of CUDA events and synchronization
  • Enhanced power model integration and statistics collection
  • Updated CUDA version detection for compatibility with newer CUDA toolkits

Developer Tools

  • Added GitHub Actions workflow for SST integration testing

Notes for Developers

  • The codebase now supports C++11 by default
  • Several internal APIs have been modified to accommodate stream-based statistics and SST integration
  • Developers should review changes to statistics collection and reporting when working with cache and memory subsystems

This release significantly enhances GPGPU-Sim's capabilities, particularly in the areas of integration with other simulation frameworks, performance analysis, and support for modern development environments. Users and developers are encouraged to review the detailed commit history for a complete understanding of all changes.

Assets 2
Loading