improve performance

[agostino64]

This PR improves performance in:

Paging: Faster virtual mapping with cleaner loops and better use of const/inline.

Scheduler: More efficient task switching and reduced overhead.

Graphics: Optimized drawRect with pointer arithmetic.

Improves speed and readability across core systems.

[ghost]

Greetings,
Thanks for taking interest in cavOS, I really appreciate that. The PR you submitted supposedly has performance improvements/optimizations, but I don’t see anything that would necessarily help performance. On the contrary, I see a bunch of my comments removed, as well as inline structs(?) added for no clear reason. Compiler optimizations aren’t used, hence adding keywords like “const” on function definitions is more so for general correctness and not optimization (same with “restrict” which I saw somewhere) Generally speaking my coding style doesn’t use “const” much, so such a change would alter the whole kernel tree without any improvements at all. The layout also looks AI-generated-ish, although I’m unsure about that. Also have you tested your changes and saw any improvements anywhere? Cause from a quick look I can’t pinpoint much. Please help me understand what is done here, because I’m left more so confused by the whole thing.

Regards,
MalwarePad

[agostino64]

Hi, thanks for your response. The code was mostly generated based on suggestions from Microsoft Copilot after analyzing multiple source files. I really don't know about const.

The performance gains from this patch are minor. The most noticeable improvement likely comes from removing the redundant if (length == 0) return;, which avoids unnecessary calculations in some edge cases.

That said, modern compilers like GCC and Clang often detect simple addition patterns inside loops and apply optimizations such as strength reduction or even auto-vectorization. Because of this, some of the manual changes (like avoiding multiplications) may already be handled automatically during compilation.

If you're asking for my opinion, I’d keep the paging: Optimize memory code for performance patch but with slight adjustments—for example, removing const and restrict to stay consistent with the existing code style and design philosophy.

[ghost]

Where in the codebase is it attempted to use the lowest level page map function with a length of 0? The if check is unnecessary and I’d be surprised if it hits even once during runtime. (Even on bad userspace mmap() operations with a zero’d out length, EINVAL is returned) Anyways, please refrain from using AI to contribute here, or in any project in general.