Anonymous
9/2/2025, 2:40:25 PM
No.106462364
>>106462218
>But the way you describe it, it's not really different from what any modern OS does (correct me if I'm reading you wrong).
Mezzano doesn't have files. It only has virtual memory and the entire partition is used for paging. The "file system" is just arrays and hash tables on the heap. To call a program, you just call the address of a function associated with it (usually through a symbol). It's like everything is on the heap all the time.
>This seems one of the cases where higher language machines are capable of a paradigm shift entirely, with some pros and some cons, but instead choose to just build a layer of abstraction over the existing method.
That's because Mezzano isn't a new machine, it's an OS designed for x86 and ARM so it has to support what those CPUs provide.
>Memory management on lisp machines, even if ultimately slow, could (and in some machines, to some extent, was) have very powerful features like memory defragmentation and runtime memory access optimization(think what the compilers do today at compile time but adapted to runtime insights) that address translation does not provide at the cost of performance. In my opinion, kind of pointless, but it's still something that higher level abstractions permit with ease and thus more valuable than a slower layer of abstraction.
This is right.
>First of all, no, theoretically the C/asm simple approach is significantly more efficient.
Lisp only needs the exact bytes you need, no overhead. All the other memory can be used by something else. C has a lot of overhead. If you have a 1 GB heap, that memory can't be used by another process even if only 1% of that heap is actually being used.
>But the way you describe it, it's not really different from what any modern OS does (correct me if I'm reading you wrong).
Mezzano doesn't have files. It only has virtual memory and the entire partition is used for paging. The "file system" is just arrays and hash tables on the heap. To call a program, you just call the address of a function associated with it (usually through a symbol). It's like everything is on the heap all the time.
>This seems one of the cases where higher language machines are capable of a paradigm shift entirely, with some pros and some cons, but instead choose to just build a layer of abstraction over the existing method.
That's because Mezzano isn't a new machine, it's an OS designed for x86 and ARM so it has to support what those CPUs provide.
>Memory management on lisp machines, even if ultimately slow, could (and in some machines, to some extent, was) have very powerful features like memory defragmentation and runtime memory access optimization(think what the compilers do today at compile time but adapted to runtime insights) that address translation does not provide at the cost of performance. In my opinion, kind of pointless, but it's still something that higher level abstractions permit with ease and thus more valuable than a slower layer of abstraction.
This is right.
>First of all, no, theoretically the C/asm simple approach is significantly more efficient.
Lisp only needs the exact bytes you need, no overhead. All the other memory can be used by something else. C has a lot of overhead. If you have a 1 GB heap, that memory can't be used by another process even if only 1% of that heap is actually being used.