High-Performance Portable GPU Primitives for Arbitrary Types and Operators in Julia

TL;DR

KernelForge.jl is a Julia library that provides high-performance, portable GPU primitives for arbitrary data types and operations, matching vendor-optimized libraries across different hardware architectures.

Contribution

It introduces a two-layer architecture enabling portable, high-performance GPU primitives in Julia, bridging the gap between portability and vendor-level efficiency.

Findings

Matches or exceeds CUB kernel execution time on scan and mapreduce on NVIDIA A40

Matches cuBLAS throughput on matrix-vector operations across tested configurations

Demonstrates portable abstractions can achieve vendor-level throughput without sacrificing generality

Abstract

Portable GPU frameworks such as Kokkos and RAJA reduce the burden of cross-architecture development but typically incur measurable overhead on fundamental parallel primitives relative to vendor-optimized libraries. We present KernelForge.jl, a Julia library that implements scan, mapreduce, and matrix-vector primitives through a two-layer portable architecture: KernelIntrinsics.jl provides backend-agnostic abstractions for warp-level shuffles, memory fences, and vectorized memory access, while KernelForge.jl builds high-performance algorithms exclusively on top of these interfaces. Evaluated on an NVIDIA A40 and an AMD MI300X, KernelForge.jl matches or exceeds CUB kernel execution time on scan and mapreduce on the A40, and matches cuBLAS throughput on matrix-vector operations across most tested configurations-demonstrating, as a proof of concept, that portable JIT-compiled abstractions can achieve vendor-level throughput without sacrificing generality.