gpgpu-sim/gpgpu-sim_distribution

GPGPU-Sim provides a detailed simulation model of contemporary NVIDIA GPUs running CUDA and/or OpenCL workloads. It includes support for features such as TensorCores and CUDA Dynamic Parallelism as well as a performance visualization tool, AerialVisoin, and an integrated energy model, GPUWattch.

GPGPU‑Sim – Cycle‑level GPU Architecture Simulator

What it is – GPGPU‑Sim is an open‑source, cycle‑accurate simulator that models modern NVIDIA‑style graphics processing units (GPUs). It can run programs written for CUDA or OpenCL and produces detailed timing, functional, and power‑consumption results.

Why it matters – Researchers use it to study how GPU hardware works, to explore new architectural ideas, and to evaluate the performance and energy use of deep‑learning workloads without needing expensive physical GPUs.

Key components

  • Core simulator – Executes PTX/SASS instruction streams cycle by cycle, modelling warp scheduling, memory hierarchy, and compute units.
  • AccelWattch power model – An extensible framework that estimates GPU power draw, built on the McPAT model and validated against real Volta GPUs.
  • AerialVision – A visualisation tool that plots simulation statistics for easier analysis.
  • Integration with Accel‑Sim – Allows trace‑based simulation using NVIDIA’s NVBit SASS traces and supports execution‑driven modes for PyTorch/CuDNN.

Typical workflow

  1. Install dependencies – Linux system, CUDA toolkit, optional OpenCL driver, and a few development packages (gcc, make, bison, etc.). Docker or VS Code dev‑container setups are provided for convenience.
  2. Build – Choose a debug or release configuration and compile via CMake or the legacy Makefile.
  3. Run – Point the simulator’s library path to your compiled CUDA (or PyTorch) binary, optionally enable the AccelWattch power model, and launch the program. Results are written to log files and can be visualised with AerialVision.

Who it’s for – Computer‑architecture researchers, graduate students, and engineers who need a detailed, configurable GPU model for performance or power studies, especially when evaluating new hardware features such as Tensor Cores or dynamic parallelism.

How to get started – Clone the repo, follow the “Step 1‑3” instructions in the README, or pull the pre‑built Docker image (ghcr.io/accel-sim/accel-sim-framework:ubuntu-24.04-cuda-12.8). The manual at http://gpgpu-sim.org/manual/ provides deeper documentation.

Related

  • Project
    BeaconBay/ck

    ck is a Rust‑based CLI/TUI tool that builds local embeddings of source code and lets you search by meaning (semantic search) as well as by traditional regex. It supports incremental indexing, multiple embedding models, AI‑agent integration via an MCP server, and outputs results in JSON/JSONL for automation.

  • Project
    bespokelabsai/curator

    Bespoke Curator is a Python library for bulk LLM inference and synthetic data generation. It provides a `curator.LLM` base class for prompt/parse logic, supports many providers (OpenAI, Anthropic, Gemini, Ollama, vLLM, etc.), offers batch mode and caching to cut costs, includes a code‑execution engine, and integrates with the Tinker SDK for LoRA fine‑tuning. Ideal for building large‑scale, structured datasets for training or evaluation.

  • Project
    datastax/jvector

    JVector is a Java library that implements a fast, scalable graph‑based approximate nearest‑neighbor index (combining HNSW and DiskANN ideas). It supports two‑pass searches with compressed and full‑precision vectors, incremental construction, and native SIMD acceleration, enabling large‑than‑memory vector search for AI‑driven semantic search and recommendation systems.

  • Project
    NeoLabHQ/context-engineering-kit

    A plug‑in marketplace for LLM‑based coding assistants (Claude Code, Gemini, Antigravity, etc.) that supplies token‑efficient, context‑engineering patterns—reflection loops, sub‑agent judges, specification‑first development, code‑review agents, and Git helpers—to make AI‑generated code more reliable and easier to integrate into CI pipelines.