* OK
So here's my underlying motivation
My Impossible Dream, my silly idea that I realize I will not achieve but I want to try to push in the direction of, is I want there to exist a "GPU equivalent" of RISCV.
Right now RISCV is open and free, you can run it in a softcore if you have an fpga, you can build it in IC as a "good as ARM" CPU if you are in China and have those facilities and expertise.
And that means there is the potential for a platform where you're beholden to no one. Because if the company in China building your RISCV CPU goes bad. You can just go to the next building over in Shenzhen and say, okay, make me a RISCV chip. And if you can pay for it, they'll do whatever design and fab work the last company did, and it will be a drop in replacement.

But the same is not true for GPUs. The same is not even true for GPUs in the RISCV ecosystem.
Every RISCV board I have yet seen that has _any GPU_, it's some proprietary thing designed by whatever company in China designed the CPU mask. And that means that if that company goes bad, you can't get a drop in replacement. It might even be harder to get a drop in replacement than it would be if your GPU vendor had been NVidia and that relationship went bad.

So, there are lots of reasons it is harder to make a GPU than a CPU.
But I still think it is a shame there is _nothing_ there in the open hardware stack. So I want to make a "seed" of a GPU, something which might not be _good_ but exists and therefore has the potential to grow into something good.