This recurrence relation
$$ T(n) = T(n - \log(n))+cn $$ is one I came up with from a computer science problem I was studying, but I'm not sure what a closed form or precise time complexity would be. This isn't a homework problem or anything and I'm not looking for hints to keep working on it, I'm just curious if anyone could provide a solution or answer with an argument. Thank you.
EDIT: As a commenter has kindly pointed out, it would make more sense if floor or ceiling was used, so I'll adjust it to
$$ T(n) = T(n - \lfloor\log(n)\rfloor)+cn $$
The original scenario for context: Consider quick sort where the average element (or element closest to the average) is always selected (assume that this pivot can be found in $O(1)$ time).
I've been wondering what the time complexity would be if the input was a list of exponentially growing numbers, e.g. $[2, 4, 8, 16, ..., 2^n]$. I've determined that the pivot would split the largest subarray in one level of recursion into two subarrays of size roughly $n-\log(n)$ and $\log(n)$. And hence I came to the described recurrence relation.
EDIT 2: Well, I suppose more accurately there would also be a $T(\lfloor \log(n)\rfloor)$ in the recurrence. I don't know how much difference it makes. I guess equivalently for this question, what I'm wondering is what the recursion height would be for this example scenario.