Symmetric multiprocessing...

Nagging hardware related question? Post here!
User avatar
Dave
SandySuperQDave
Posts: 2411
Joined: Sat Jan 22, 2011 6:52 am
Location: Austin, TX
Contact:

Re: Symmetric multiprocessing...

Postby Dave » Tue Jun 18, 2019 9:12 pm

That goes in the "asymmetric processing" thread, I think. In which case, I would point you to the CM3 version of the Pi, which has far superior GPIO. And they're the size of a laptop SIMM module. I hopefully will have something interesting to post photos of in a few weeks on that front.


User avatar
Peter
Aurora
Posts: 845
Joined: Sat Jan 22, 2011 8:47 am

Re: Symmetric multiprocessing...

Postby Peter » Wed Jun 19, 2019 9:38 am

Dave wrote:Do we find a way to fix the OS, or do we throw hardware at the problem to make it go away?

Given the software situation, I tend to say let's throw hardware. But in my examples, higher singlecore performance/throughput is needed.
Multiprocessing, which is technologically easier to achieve, can't help there.

Assuming I had a quadcore Q68, I don't see many cases where it would actually speed up my work.
Yes, I could run jobs in parallel on different cores. But are parallel computing jobs slowing me down at the moment? Almost never.

Recent experiments have shown that achievable CPU clock frequencies do not rise much, if I move from the decade-old Q68 FPGA to a recent chip of the same manufacturer.
Maybe from 40 MHz to 50 MHz at similar speedgrade, but that's no quantum leap.
While on-chip block RAM and available logic size increased about 3..4 times for the same money. That makes symmetric multiprocessing interesting. I just fear it could become a "solution in search of a problem" while the primary issues remain.


User avatar
Dave
SandySuperQDave
Posts: 2411
Joined: Sat Jan 22, 2011 6:52 am
Location: Austin, TX
Contact:

Re: Symmetric multiprocessing...

Postby Dave » Wed Jun 19, 2019 8:52 pm

If you have 4 cores, and the jobs are evenly distributed by load, wouldn't that make for a much more responsive machine? Would changes in polling strategy help, eg have executables elect to be fast or slow polling and be allocated to a fast or slow polling CPU as requested?



Who is online

Users browsing this forum: No registered users and 7 guests