Has any work been done on page size ? I had another scan and found nothing in the documentation .
I was hoping if someone could post regarding the following questions.
1) Is 4K pages really a good model for a Micro Kernel ? Larger pages would result in fewer OS calls and a much smaller table helping the TLB. If the memory is a concern for small apps you could have an a process which allocates small pages by grabbing a big
page and sub allocating it . In a 16 Gig system of the future id rather have 10.000 4 K pages and 16,000 1 Meg pages than 4,000,000 4K pages.
2) Why are pages assigned through the ABI and not via the channel mechanism ? This means the memory manager becomes trusted but cant the Memory Manager be consider like a device driver ( ie access the pages via the HAL) ? i seem to recall Minix having a
Memory manager that was a seperate process.
BTW i like how you removed paging i dont think its really relevant these days. Besides the bugs and OS issues nothing is more annoying than say when windows has decided an app isnt being used for a while and allocates the memory to cache (starving it) and
then when you switch to it it takes 10* longer to load then a new copy . Prob due to having to load all the pages I note SDN18 states there is no paging but there seems to be a bit code based on this.