sr (nobody.delete@this.nowhere.com) on September 25, 2021 12:07 pm wrote:
>
> And about memory allocation - for very small allocations fine granularity is acceptable but for
> larger ones isn't it wise to allocate to page limits even if hardware allows finer allocation?

Since the only reason to do this is for pointer checking, granularity of checking actually kind of matters.

And in many ways having a fixed granularity limit is ok - a lot of perfectly good algorithms do things like "over-read the end a string by one word to find the final NUL character" because they do vectorization (I mention the C string case, but that kind of thing happens in many other situations).

If you do pointer checking, you need to be able to deal with those kinds of benign pointer overruns.

That's fairly easy to do when you have a fixed allocation alignment checking limit.

But it's actually problematic if you sometimes get faults on a byte boundary, and sometimes at a page boundary. Then you effectively cannot use the optimistic version at all, or have to have some ridiculously complicated version.

It really sounds like you haven't actually ever used any of these kinds of things, and have never actually seen the problems they have in practice.

Linus