Сообщество

It seems, that there is no technical limitations to implement high-frequency version of the architecture. But there are, of course, financial :) And this should give performance gains. As to toolchains, we cannot just port GCC or LLVM, because of their abstractions for target machine. It must be register machine, which means: compiler assumes that after instruction changing register named X all consequent references to the register X will read the written register. MCp does not behave like this. And it will cost too much to rewrite GCC or LLVM in order to adapt these tools to our architecture. We will need in that case to dig into massively complex internals of these systems.

As too Linux it is too early to judge. POSIX may be just too limiting API for our architicture, and it is known (from our supercomputing community at Russian Academy of Science) that there are far better models for next-generation HPC, than currently widely used MPI/OpenMP. May be we should evolve our software stack in that direction. Who knows?

P.S. Excuse me my english.

Our existing MCU has 64-bit internal organization(data paths, on-chip memory, logic operations). Next version also will have the same one.
There are no problems to reach 1GHz, but it require extra finances.

Like I said on other thread, I hope that open source comunity will eventually port GCC and Linux when MultiClet shows market success (and I have no doubts it will, it is great technology - it is just matter of time). Is your compiler open source? Maybe it could be extended to understand GNU C and thus compile Linux. If you are harware company and have no plans of seling software, maybe you could opensource compiler and set up open source project of making your compiler compile Linux, with maybe some prize development boards for top contributrs. What do you think of that?

For me, architecture without Linux support is... well, uninteresting for development.

As for supercomputers, there might be better way to scale than Linux in the future, but in server space, Linux and it's repository of available software are irreplacible in datacenter. Linux is there to stay. Similar is true for mobile phones, which run android and Linux. I think Linux is must have for every CPU architecture which aims to be widespread. That is not to say that every workload is best run on Linux, but Linux have to be available.

I understand you are a small company and cant do porting alone, I just hope you swamp the market with your awesome chips so somebody scratches his itch and ports stuff so you don't have to bother too much.

I am curious, what POSIX stuff are limiting for multyclet? Linux today is not only POSIX...

As for frequency scaling... again optimism: I think you have great technology and my humble opinion is that it will make money. It take some years but it will.