gpgpu-sim/gpgpu-sim_distribution

Health v2 · maintenance only

Flags: no_license

Adoption not part of the score

1701 stars · 663 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-30, confidence not recorded

GPGPU-Sim is a cycle-level simulator that models contemporary NVIDIA GPUs running CUDA and OpenCL workloads, including support for TensorCores and CUDA Dynamic Parallelism. It bundles the AerialVision performance visualization tool and the AccelWattch/GPUWattch energy model, and integrates with the Accel-Sim trace-based framework.

Use cases

• simulate CUDA kernels on a modeled GPU without hardware
• run OpenCL workloads on a detailed GPU simulator
• estimate GPU power consumption with AccelWattch
• evaluate GPU microarchitecture design changes
• visualize GPU performance dynamics with AerialVision
• simulate deep learning workloads with Tensor Cores
• run trace-based SASS simulation via Accel-Sim

When to choose

• you need cycle-accurate GPU architecture research without owning hardware
• you are studying power/performance tradeoffs of NVIDIA GPU designs
• you want to experiment with Tensor Core or dynamic parallelism behavior
• you need a validated, citable GPU simulator for academic publications

When to avoid

• you just need to run CUDA code fast on real GPUs
• you want a lightweight or functional-only GPU emulator
• you need simulation of non-NVIDIA or modern post-CUDA-11 hardware features out of the box
• you need a maintained production tool rather than a research codebase

Facets

library · maturity maintenance

simulation benchmarking gpu-computing developer-tools gpu-computing hardware performance cpp gpu-simulator cuda opencl cycle-level-simulation power-modeling accelwattch aerialvision academic-research tensor-cores accel-sim research computer-architecture linux

Member repositories

For agents

markdown · JSON · MCP: product_card(name="gpgpu-sim/gpgpu-sim_distribution")