Why do you use Linux when MINIX is superior in every way?

Just use OpenBSD NIGGERS

Sorry, meant to say *niggers(not capitalized)

The clever people have been using a hybrid kernel. Combines the best features of both Micro and Monolithic.

>turning on caps lock to write three letters

Fuck
was meant for I feel like a fucking retard now

That diagram is misleading. The scheduler doesn't go through the file system to communicate with the application, nor do device drivers go through the scheduler.

Have you actually compared performance of microkernels vs monolithic kernels? Microkernels theoretically have better core security, but monolithic kernels always have better performance.

NT started a microkernel. But it was too slow so they moved programs into kernel space. But for the last ten years they've been moving various components back to user mode.

Similar story for Apple's XNU, a microkernel degenerated to hybrid kernel for performance.

to be fair redox is pritty gud.

I was typing on mobile senpai

Trapping to kernel for literally everything ends up killing performance. The theory and design of microkernels is really neato. I'm not sure what CPU designers could add that would make microkernals more practical.

Attached: 1452571964658.png (353x387, 106K)

You already are if you run on intel M.E runs a modified minix os

Software get's fucked by hardware optimization.

Binary trees are slower that simple lists because of that for quite a large N for example because the hardware is optimized for lists as it tries to prefetch data to be faster and that doesn't work well with a binary tree that acts as random access and thus generates lots of cache misses when the hardware just tries to prefetch the next element instead of the leaves in the tree. TLDR: Whenever hard&software don't work in tandem shit get's fucked and so the smart optimizations can turn worse than you would expect.

I would expect the linked list to run faster in most cases just because it is an inherently simpler data structure to both analyze and parse. In practice, a heavily fragmented list can perform just as poorly as any other data structure kluge. You can balance your binary trees to keep access times consistent, but that incurs performance penalties in and of itself.

Well to be precise the cutoff where trees beat lists was higher due to that. Of course a small list with 100 elements would always be faster than a tree because random access is inherently slower and so trees shine only when the tree can speedup the search beyond the cost of random access. Due to the cpu optimization you however had the cutoff where trees beat lists moved to over 100000 elements as it increased the cost for trees while speeding up lists.

I'd just like to in- oh, nevermind.

Attached: 1524161389870.jpg (1024x1200, 311K)

> not realizing that Intel chips with Meltdown are basically running Linux as microkernels

Checkmate, atheists.