[coyotos-dev] Hi + some stuff I found researching

[Sandro Magi]

The Synthesis Kernel
http://citeseer.ifi.unizh.ch/22153.html

Synthesis: An Efficient Implementation of Fundamental Operating System
Services
http://citeseer.ist.psu.edu/massalin92synthesi.html

Threads and Input/Output in the Synthesis Kernel
http://citeseer.ist.psu.edu/massalin95threads.html

Follow-up projects to Synthesis?
http://www-static.cc.gatech.edu/~calton/os.html

I'm not sure how much Synthesis could add to Coyotos or microkernels in
general. Microkernel paths are already so short and highly optimized by
hand, that I'm not sure code-generation could help.

And as shap as pointed out before with Pebble OS
(http://www1.bell-labs.com/project/pebble/), code generation in the
kernel makes it vulnerable to DoS (if the code generation is  unbounded
in any circumstance).

Sandro

Haplo wrote:
> Hello to coyotos-dev. I suppose I'll start with a bit of unnecessary  
> background, but at least it can give you some idea of my level of  
> understanding of all this.
> I'm a 20 year old computer engineering undergrad currently living  
> outside charlotte (however this is soon to change), and really I  
> stumbled across this while randomly researching related topics. I was  
> aware of KeyKOS a while back after I got an interest in kernel  
> architecture, and actually planned to reimplement it with L4  
> optimizations some time in the undisclosed future. It was only  
> recently that I came across coyotos and found it had already been  
> done. So, accordingly I read through most of the documents found here  
> and expanded my research to better understand the topics being  
> discussed.
> 
> The useless information aside, I do actually have something  
> potentially useful to add here. I very recently was doing some  
> unrelated research (on objective-C) which led me to self-modifying  
> code and/or dynamically generated code (don't bother asking how those  
> things got chained together), and eventually came across a rather  
> obscure kernel I'd never heard of before. This being the Synthesis  
> kernel. The general description of it was "very fast, using  
> dynamically generated code and almost object-oriented design", which  
> I took critically, but decided to investigate further. This actually  
> turned out to be harder than I assumed since very little -useful-  
> information on synthesis is floating around. I did, however, manage  
> to obtain an outline specification written by the original designer.  
> After reading through it numerous times, doing research on some of  
> the obscure terms he uses, and comparing it to coyotos, I believe not  
> only that the techniques used could be applied to coyotos, but that  
> they might offer significant performance improvement. Furthermore, he  
> faced similar SMP scaling issues that you are currently having, and  
> his solution may work quite well for you, as well.
> 
> With that said, I confess that I'm fairly naive in the field, perhaps  
> much less so than the average person, but I am by no means an expert.  
> As such most of this is to a point speculative, however if you AREN'T  
> familiar with the design, I would at least ask that you review the  
> document and decide this for yourself. I can provide it (it's a pdf  
> scan of some original typed text) if necessary.
> _______________________________________________
> coyotos-dev mailing list
> coyotos-dev at smtp.coyotos.org
> http://www.coyotos.org/mailman/listinfo/coyotos-dev