Glenfly HUCA

Heterogeneous Unified Compute Architecture

A Unified Software Foundation — Glenfly HUCA

A Unified Software Foundation — Glenfly HUCA is a unified computing and AI software stack that Glenfly has built around its self-developed Glenfly HMUA GPU architecture. Spanning drivers, compiler, runtime, operator acceleration libraries, and development toolchain, it provides complete software infrastructure for graphics rendering, parallel computing, video processing, and AI training and inference.

A Complete Execution System — Glenfly HUCA

A Complete Execution System — Glenfly HUCA: The Glenfly HUCA software stack comprises core modules including Glenfly HUCA Runtime, Glenfly HUCA Compiler, Glenfly HUCA Driver, the Glenfly HUCA acceleration libraries, and the LLM Toolkit SDK. It provides a complete execution path from Runtime to Driver, enabling efficient management of video memory, context, execution queues, and compute resources.

Offline optimization for LLMs

Offline optimization for LLMs: The Glenfly HUCA LLM Toolkit SDK supports model conversion, quantization, graph optimization, subgraph replacement, operator fusion, pruning, sparsification and model compression, improving LLM deployment efficiency on Glenfly GPUs.

Low-cost ecosystem migration

Low-cost ecosystem migration: Through unified APIs and supporting development tools, Glenfly HUCA helps developers migrate existing applications, algorithms and models to the Glenfly platform and continuously optimize runtime performance.

Glenfly HUCA Layered Architecture

Balancing ecosystem compatibility with optimized performance, powered by a proprietary full-stack framework and fine-grained tuning to unlock the full potential of hardware.

Glenfly HUCA software stack layers
Glenfly HUCA Layered Architecture