daniel-cortez / pawn-3.2-plus Goto Github PK

Proposal/Question

Ever since I saw your improvements to the runtime, specifically the address checking, I wondered, whether you would consider adding/pulling memory-mapping to the runtime. The way I imagine it could work is, that when an instruction gets invalid address, it would first check, whether it is a mapped address before returning an error.

So for example, the upper 1GB of the amx address space could be used for mapping. So the mapping code would first chceck the two most significant bits and if they are not 11, the address is not in the upper 1 GB. So it would return an error.

Otherwise it would check the third most significant bit, if it was 0, it would use segmentation. So it would strip the three most significant bits from the amx address, compare it to the size of the mapped memory and if it is lower, it would add base pointer and use that address as physical adress.

If the third most significant bit was 1, it would use paging. It would use pages up to 64 kB large, allowing for 8192 pages. Each page would have 8 byte entry in its page table. 4 bytes would be the base address, 2 bytes would be flags and 2 bytes would be the size of the page. So the page table would take up 64 kB. 13 most significant bits (exluding the 3) would be the number of the page in the page table, the remaining 16 bits would be offset from the base of the page.

This would help with a lot of functionality, such as dynamic memory, shared memory, allowing better passing of arguments between amx files (instead of copying an array/string, you would temporarily map the array to the other amx), etc.

The flags would at least contain read permission, write permission, execute permission and possibly more advanced stuff, such as copy-on-write. They could also contain technical information, such as if this is the first page in a mapping.

See pawn-lang/sa-mp-fixes#98
Desired behavior: remove only properties created by the unloaded script.

Hello everybody,
isn't there any plan for implementing these changes into a Pawn 4.0(or at least 3.10) fork ??