| 1 |
Peter Stuge wrote: |
| 2 |
> wireless wrote: |
| 3 |
>> Or for another (mythical) example: |
| 4 |
>> We used 2 processors, one running a state machine for directly |
| 5 |
>> reading an array of 20 bit D/A in sub microsecond fashion, and then |
| 6 |
>> passed the data (like this) to another processor running RT_linux |
| 7 |
>> (windRiver or embedded-gentoo)..... |
| 8 |
> |
| 9 |
> People plug FPGAs into HyperTransport (FSB used by AMD64) - does that |
| 10 |
> count? |
| 11 |
|
| 12 |
|
| 13 |
Yes. Any thing you can share is great. Timing/latency data, |
| 14 |
issues, pseudo code, system architecture...... what decisions |
| 15 |
where made and why..... |
| 16 |
|
| 17 |
|
| 18 |
My focus is on the "low latency" aspect and what can be done |
| 19 |
(how fast) directly under embedded linux (rt, rtai, xeno... |
| 20 |
whatever tricks) control or by bypassing the rtos all together |
| 21 |
and bringing in the data via another port/interface. why? |
| 22 |
|
| 23 |
and also where folks have had to bypass the rtos because of |
| 24 |
critical timing/bandwidth issues, or just simple to build |
| 25 |
a system that is difficult to reverse engineer (FPGA |
| 26 |
or SOC or asic).... |
| 27 |
|
| 28 |
|
| 29 |
Again, I'm not after sensitive information, just performance |
| 30 |
numbers and any technical details you can or are willing to |
| 31 |
share. |
| 32 |
|
| 33 |
tia, |
| 34 |
James |
Archives