Intel UXA Acceleration Performance
Subsequent to the introduction of the Graphics Execution Manager earlier this year, Intel had introduced a new acceleration architecture. UXA, or the UMA Acceleration Architecture, was developed as a temporary solution based upon the EXA architecture but with support for the kernel-driven GEM memory management. How though does the UXA performance compare to that of EXA? In this article we have ran some benchmarking looking at the Intel graphics performance.
UXA was introduced in early August to allow 2D pixmaps to become GEM objects. The EXA API was used as the basis of forming UXA but the code was then stripped away and replaced with code to support the Graphics Execution Manager for managing its memory. A month later it was clarified by Keith Packard at the X Developers' Summit with his UXA intentions. At that time he stated the UMA Acceleration Architecture will be discontinued and the GEM changes merged back into EXA once they decide how to split the pixmap management and acceleration architecture components. Keith hoped to have this revised EXA work done for X Server 1.6, but that hasn't been completed in time.