Q1-kernel / Blog: Recent posts

Metric (Lower is Better)	Linux (PREEMPT_RT)	FreeRTOS	seL4	L4Re/Fiasco.OC	DragonFly BSD (Est.)	Q1-Kernel
1. Inter-Process Communication (IPC)	2.0 – 4.5	1.5 – 3.0	0.5 – 1.0	0.3 – 0.7	1.0 – 2.0	0.80 – 1.3
2. Process Context Switch	1.5 – 3.0	1.0 – 2.5	0.8 – 1.8	0.7 – 1.4	1.5 – 3.0	1.5 – 2.0
3. Worst-Case Interrupt Latency (WCIL)	10 – 50	5 – 15	5 – 20	10 – 30	5 – 15	20 – 30
4. Process Creation Time	50 – 200	N/A	20 – 50	15 – 40	50 – 150	40 – 60
5. Syscall Overhead (Null Call)	0.2 – 0.4	N/A	0.1 – 0.3	0.1 – 0.3	0.2 – 0.4	0.2 – 0.3

The latest nightly update of Q1-kernel has the final version of Q1-kernel. This version is still in beta but everything that is part of the main code is finished. The image compiles and boots with no errors.

We have internet support trough a VM that uses virtual ether.
We have 60 linux syscalls and 28 posix syscalls.
Libc on the new OS that will come out.
And the most important context switcher that is proofed in Isabella, but more like a theory proofed so nothing very very complex but still that means its safe fast and smaller, only needing at lowest 2kb cpu buffer/log on every "Node", but that means more resets of the buffer/log but still fast. And at max 256kb cpu buffer/log on every "Node".... read more

For the last couple of days i was thinking for a fast and safe contex switcher, scheduler, thread scheduler.
As you may know Q1-kernel has a very new kind of core and thread control. The kernel runs on one core plus 2 threads (or 1 based on threads count) and the rest of cores and threads are separated in Nodes, more like a cluster with very fast communication. In this way every core has his own job and that means no core conflict.... read more

Ok so i just completed some things from the roadmap

1. Networking (still only ether) complete
2. Process management to have a new arhitecture planning
3. moreeee syscalls and support for more software complete
4. An OS complete still beta

Special thanks to Robert Claus for helping me with the OS and porting musl to it

I want to start with special thanks:
to MJ_jakk for helping me with the network driver
to Gabriel Nix for helping me with porting lasm to ARM, and giving me tips for assembly
to Harry-Youth for helping me with management the kernel
They are some skilled programmers that helped in redox, linux and ARM development

For the next big update here is the roadmap:
1. Networking (even if only ethernet) complete
2. Process management to have a new arhitecture planning
3. Moreeeeee syscalls and more support for software
4. An OS... read more

Kernel	Syscall Latency (µs)	Context Switch (ns)	IPC Round-trip (ns)	Minor Page Fault (ns/page)
Q1-kernel	0.2 – 0.3	230 – 250	800 – 900	2000 – 2500
Linux	0.2 – 0.4	150 – 250	700 – 1200	1000 – 3000
DragonFly BSD	0.3 – 0.5	200 – 300	800 – 1300	1500 – 3500
XNU (macOS)	0.6 – 0.9	400 – 600	1200 – 2000	2000 – 4000
Windows NT	1.0 – 1.5	500 – 700	1500 – 2500	2500 – 4500

With lasm 0.2 comes some big improvements with Q1-kernel. With lasm 0.2 we have xt and con instructions. Those 2 instructions can be used in register manipulation like instead of 4 lines with "block" instruction plus "main" manipulation you can easily do that xt. At the same time with con you can access cpu cores in separate registers without advance "block" instruction.

Enough with lasm lets start with Q1-kernel improvements in multi coring. As you may know Q1-kernel threats every separate core like a separate pc. So everything runs in a "cluster" with high speed messaging, The OS runs in on core then it gives instruction to each core about userspace processing. In this way there are no core conflicts. In this test every OS runs on the Qemu 10.1.1 version. We have linux vs DragonFly BSD vs Q1-kernel. The test is how fast can each one compile Q1-kernel... read more