I'm about to be assigned to develop an embedded system running over Linux that will necessarily deal with the uC built-in GPU to perform some real time 3D trigonometric calculations - more exactly, to make fast conversion between cartesian to polar coordinates without consuming CPU processing resources for that, due this board will also be tasked with other hard workloads. The choosen hardware platform is the ARM Cortex A8, due there are some open designs for that application already available on the Web that could be used as template, but none of them implements access to their respectives graphic accelerators, hence I should install their libraries, but I never did something like that before by myself.

Both kit options are based either on the Allwinner A10 (gpu=MALI400), or the Sitara AM3359 (gpu=SGX530), and apparently the second choice seems to have more support on the Web, I mean, a bigger audience, which seems appropriate for me. However, I could not find anywhere anyone which have been compared both solutions in terms of usability and easiness to install and to use. Most information available concerns just on performance aspect rather than in the developer perspective, but everything I saw so far steer me to the TI's option.

So, does anybody have any opinion if I'm apparently on the right direction ?