I'm a bit naive at times, and it looks like this is one of them. I had the impression we were having an honest discussion about possibilities, but it's now obvious to me that there is an agenda here, based on the fact that it's pretty clear that you found this thread on Google, and joined the forum solely to advocate for something you have vested interest in.
These processors and GPU's in tablets, handhelds, etc.. are simply advanced microcontrollers. Someone who programs solely in high level languages, like C, Basic, Python, Perl, etc... would of course see things the way you see them. To folks like this, no pre-existing high level libraries means no possibility of accessing features built into hardware. Well, let me tell you that it doesn't work that way. There are still programmers out there who code at the hardware level, and that's where these libraries originate. There are high-level compilers for standard microcontrollers these days as well, but it is not uncommon for code which is timing sensitive, or performance oriented, or provides access to features for which there are no standard libraries, to be written in the native assembler. At that level, functions and features of the hardware are directly accessible. You state that they cannot do this, because there is no information available to do it. I disagree that this is the case. It does not at all seem unlikely to me that the core functions of the GPU are well known by Ingenic, and, of course, the GPU's creators, Vivante.
A little research into Vivante yields some valuable information. First, while it has global headquarters located in California, it's obviously based in China. The CEO's name is Wei-Jin Dai, and they have locations in Shanghai and Chengdu. Their products are in many of the low-cost tablets and other devices originating in China, so there is obviously a strong alliance with chip manufacturers there. What is more important is that they support OCP (no, not the one in Robocop.) Here's the statement;
OCP
In 2008 Vivante became a member of the OCP International Partnership (OCP-IP). OCP-IP is a non-profit semiconductor industry consortium formed to administer the support, promotion and enhancement of the Open Core Protocol (OCP) specification. OCP is the only fully supported, openly licensed, complete interface socket for intellectual property (IP) cores. OCP addresses design, verification and testing issues common to IP core reuse in "plug-and-play" system-on-chip (SOC) products. Additional information is available at
www.ocpip.org.
I encourage you to read the white paper and provide your thoughts. It appears to me that OCP is an open, unified interface which allows hardware developers access to core functionality of integrated components, while protecting IP. As such, it is portable across multiple hardware platforms. It looks to function in a similar, but more advanced, way to something like I2C, where there is a common communication methodology, regardless of the task the hardware on the other side of the communication protocol is designed to perform. Membership and access to the documentation is free, even for commercial use, but requires paid membership to gain access to support and libraries.
So your claims about IP and inaccessible hardware do not seem to be rooted in reality. What the folks at GCW need to do, and are probably already doing, is write low level routines which provide whatever subset of the OpenGL protocol the system is likely to require. This isn't something new, and projects like Mesa do something not so different for other hardware. Their comment about their OpenGL implementation not being open source is telling. It's likely that they are using a library for which they do not have, and cannot get the source code. But regardless of the open-source status, this statement indicates that something is being implemented.
Based on the above, it's nowhere near as impossible, or even improbable, as you have claimed it to be.