Download Now Free registration required
Decades of parallel computing practice have proven that highly parallel code runs efficiently only when it has uninterrupted access to the hardware. The authors report on a project whose goal is to support compiling Data Parallel Haskell code into bootable disk images. Their toolchain layers the Data Parallel Haskell runtime system on top of the GeekOS operating system and Newlib C library. They explain how their toolchain, combined with a virtual machine monitor, can allow optimized low-level parallel code to be run in cooperation with a traditional OS.
- Format: PDF
- Size: 134.3 KB