Nothing 2+2=5 has said is wrong. It's just besides the point.
The point, which I sought to clarify, was why the Photoshop brush is slow. The comparison between the simple brush and the complex brush was never the point. Ever. I wanted the readers to fix their attention to this snippet from the article I linked:
It's a short and concise point which clarifies why something seemingly simple is slow. To me, this does answer the question "Why is the Photoshop brush slow?". The answer, in case you missed it, is "because the algorithm used to fill the brush is order n-squared complexity, where n is brush radius" (I've extrapolated this conclusion from the article, so it's not necessarily true, but to me it doesn't seem unthinkable). You can disregard the rest of the text in the article as it doesn't lead to another conclusion.
As for what I'm suggesting; None of the above. We already know Photoshop has GPU accelerated rendering, so what's the point of suggesting otherwise? However, we also know that O(n²) scales poorly regardless of what silicon it is being executed on. As for suggesting that the chosen algorithm is poorly chosen, valid, and/or complex; If I were suggesting any of those alternatives, would that somehow make it not an O(n²) algorithm? Would it modify the original point I was trying to clarify? The way I see it, the answer would remain the same regardless, which is why I refrain from such value judgements.