KStars GSOC: data pipelining and OpenCL
by Henry de Valence for KDE
KStars has many powerful features, but it's slow and does not deal well either with drawing large catalogs, or with normal-sized catalogs on low-end machines suitable for field use. The goal of the project is to rewrite KStars' data-processing code to both be parallel and to make efficient use of the hardware available, whether they have a single x86 core, multiple cores, a discrete or embedded GPU, or a system-on-a-chip like the Exynos 5. There are three main areas of improvement: 1. Computation speed; 2. Memory access patterns; 3. Parallelism. Currently, KStars does poorly at all of these; by the end of the GSoC project, all of them will be improved.