Email: Password: Remember Me | Create Account (Free)

Back to Subject List

Old thread has been locked -- no new posts accepted in this thread
???
11/10/09 16:37
Read: times


 
#170644 - I think system-on-a chip is more imminent than clusters
Responding to: ???'s previous message
The problem is that todays programmers using todays methodology have a hard time creating working software for one processor. Creating software for clustered hardware is not easier. You get new set of problems with synchronization, consistency of data and state, recovery logic, ...

When possible, it really is advantageous to use one bigger/faster processor than many small for a large set of problems. And on the PC, will will see more and more failing programs from bad interprocess communication when a single application has to make use of many processor cores instead of just one very fast core.

Few systems needs the extra security from majority-voting redundant systems, and in that case you would still want each voting system to be complete, and let the voting be handled by discrete hardware that merges the output states from the individual computers.

In reality, a very large percentage of all problems we see are software errors or the result of bad power or noisy communication links.

Software bugs should not be overcome by adding more hardware.
Noisy power should not be overcome by adding redundant computation units.
Noisy communication links will just be even more critical if used in a distributed system.

The 8051 is (with a few exceptions) a very small controller for handling state information for a limited number of signals. What is happening more and more is the addition of high-end controllers with hardware capabilities similar to older PC machines combined with better peripherials optimized for real-time-critical performance. We get systems-on-a-chip solutions. But with the simplicity of simulating macro blocks, we will get even more powerful system chips - custom-designed for very narrow and well-defined problems.

We get networking chips where the traffic is handled in hardware, while an integrated processor core analyzes the operation and performs firewall checks for new connections.

We get signal-processing chips where a low-end processor core is augmented with enough DSP capability for creating tiny vibration-detectors etc - a sugar cube with a connector on one side to be glued to a machine and report an imminent bearing failure.

We get cellular base stations where everything is processed by software and combined with antenna arrays supporting dynamic lobe control.

List of 30 messages in thread
TopicAuthorDate
Everyone has a Favorite 8052... what is yours and why?            01/01/70 00:00      
   SAB 80C535            01/01/70 00:00      
      Mine too            01/01/70 00:00      
         what guild is it then?            01/01/70 00:00      
   favorites and a bit of nostalgia            01/01/70 00:00      
   P89V51RD2            01/01/70 00:00      
   Triscend E5            01/01/70 00:00      
      Triscend E5 successor?            01/01/70 00:00      
         Triscend E5 successor(s)            01/01/70 00:00      
            and Zylogic?            01/01/70 00:00      
               Zylogic mystery!            01/01/70 00:00      
   Believe it or not..mine is AT89C52 / AT89S52            01/01/70 00:00      
      Mine too...            01/01/70 00:00      
   C8051F310            01/01/70 00:00      
      C8051F340            01/01/70 00:00      
         What Jack says            01/01/70 00:00      
         Future for 8051s            01/01/70 00:00      
            the REAL future for the '51 as i see it ....            01/01/70 00:00      
               SOMETHING with a processor            01/01/70 00:00      
                  I think...            01/01/70 00:00      
                     multi-microcontrolling            01/01/70 00:00      
                     I think system-on-a chip is more imminent than clusters            01/01/70 00:00      
                        Fish swarms and bird swarms...            01/01/70 00:00      
                     for automotive applications, for example            01/01/70 00:00      
                     not just the future            01/01/70 00:00      
                        Kind of..            01/01/70 00:00      
                           nope            01/01/70 00:00      
                     This is true future.            01/01/70 00:00      
                        there are two ways            01/01/70 00:00      
                           Co-processors yes, clusters no            01/01/70 00:00      

Back to Subject List