OSDev.org

started to design the kernel.

1. malloc and free is ready, it was simply becouse the lack of mmu. not yet tested however, but i will test it soon. ~70 code line, and no security check yet for freeing by the proper process. memory size detected dinamically, no need to help by ,,bios'', and its hotplug (ram can be added dynamically to the system.) hardware mapped areas are automatically skipped.

2. process control block ready, process loading not yet ready. its very simply, becouse there is no hardware acceleration for process switch, and no interrupt, so no interrupt vectors to play around. its around 100 lines.

none of the above is tested yet, but will tested soon

sound output is almost implemented

mallocing and freeing 1 million 1 byte sized data with malloc: ~0.1 second (running my malloc algo on x86)
memory footprint: 25 mbyte

mallocing and freeing 1 million 10 byte sized data with malloc: ~0.1 second (running my malloc algo on x86)
memory footprint: 34 mbyte

mallocing and freeing 1 million 1..100 byte (random) sized data with malloc: ~1(best)-20(worst) second (need very very strict rewind for the malloc algo becouse of the missing mmu)
memory footprint: 74 mbyte




could be optimized later a bit, will do it for now

ok, i optimised the memory demand by 30%, it also speeded up 30% from this.

also, luckily, after investigating the partitions in memory, it turns out that the brutal memory holes featured by operating system books are unexist in reality.

i am in 50% of implementing fopen