Library/impForked

kekzl/imp

imp

High-performance LLM inference engine in C++/CUDA for NVIDIA Blackwell GPUs (RTX 5090)

View on GitHub↗Upstream kekzl/imp↗

Builder

kekzl

kekzl • individual

Stars

Using upstream star count

Forks

Using upstream fork count

Open Issues

Activity Score

0/100

0 commits in 30d

Created

Feb 23, 2026

Project creation date

README Summary

IMP is a high-performance large language model inference engine written in C++/CUDA specifically optimized for NVIDIA Blackwell GPU architecture, particularly the RTX 5090. The engine focuses on delivering maximum throughput and low latency for LLM inference workloads by leveraging the latest GPU capabilities.

AI Dev Skills

Unmapped

GPU ComputingCUDA ProgrammingLLM Inference OptimizationLow-level Systems ProgrammingNeural Network AccelerationMemory ManagementGPU Kernel Development

Taxonomy

AI Trends

On-device AI Hardware-specific Optimization Local AI Inference

Deployment Context

Self-hosted On-premise

Modalities

Text

Skill Areas

GPU Computing CUDA Programming LLM Inference Optimization Low-level Systems Programming Neural Network Acceleration Memory Management GPU Kernel Development

Use Cases

High-speed Local LLM Inference GPU-accelerated Text Generation Research Computing

Recent Activity

Updated 28 days ago

7 Days

30 Days

90 Days

Quality

prototype

Quality: low
Maturity: prototype

PM Skills

Developer Platform

Languages

Cuda100.0%

Timeline

Project created: Feb 23, 2026
Forked: Mar 12, 2026
Your last push: 28 days ago
Upstream last push: 7 days ago
Tracked since: Mar 17, 2026

Similar Repos

pgvector cosine similarity · $0

Loading…

imp