refactoring core service as a driver

[franalta]

Hi, I'm been asking it's possible convert the source code of core as driver to increase performance and efficiency... what do you think? it's require many efforts?

[franalta]

nobody?

[Jalle19]

I have a feeling you don't really know what you're asking for? As far as I know there aren't any performance issues, and even if there were it would probably not make sense to move anything to kernel-space.

[franalta]

yes, there aren't performance issues, my opinion is that get data from driver tuner and convert it it's however more efficient... if I will have enough knowledge I'll try to make this porting with a fork

[catalinii]

I would not bother with it. I think for satip axe which is a very slow cpu, minisatip can still push very high bw because of the fact that the input buffer is used for output and multiple packets are chained together in 1 syscall.

The real bottleneck could be decrypting but doing that on the kernel side will not help much from decrypting perspective.

Not sure what is your use case but never saw minisatip without decryption going over few percentages on modern cpus.

[franalta]

question is: if I can make it more efficient, why not make it?

[catalinii]

Think about it this way: What do you gain from it?

• if your server will use 5% cpu instead of 10% and now is 95% idle instead of 90% ... You still pay the costs of the server.

What is your use case?

[franalta]

I don't have a case... it's simple a suggestion to increase efficiency (both for decrypted that not) getting data from a more raw layer (from driver)... However if you don't want make it when I have some free time and enough knowledge I'll try to make it

[franalta]

whay standard ANSI C has been used to develop the server?

[catalinii]

Now it compiles even with C99