Tuesday, 26 January 2010

High-performance parallel programming with HLVM

The OCaml Journal just published an article about HLVM:

"This article describes how a global adjustment of the calling convention brings the performance of the multicore-capable HLVM within 17% of its original single-threaded performance. A simple parallelization of the new ray tracer benchmark is then described that obtains a substantial 6.3× speedup on 8 cores, demonstrating the potential of the new design for high-performance parallel computing..."

To read this article and more, subscribe to The OCaml Journal today!

No comments: