| |||||||||||||
The 2012 realtime minisummitThe 2012 realtime minisummitPosted Oct 24, 2012 22:16 UTC (Wed) by jeff_marshall (subscriber, #49255)Parent article: The 2012 realtime minisummit
>Paul McKenney noted that many processors have lots of floating point registers which can add "multiple hundreds of milliseconds" to save or restore.
Surely this was either misspoken or misquoted? Even substituting microseconds for milliseconds doesn't add up - you can do a lot of work in that amount of time when your CPU clock runs in the Ghz.
Am I missing something here?
(Log in to post comments)
The 2012 realtime minisummit Posted Oct 25, 2012 22:55 UTC (Thu) by PaulMcKenney (subscriber, #9624) [Link] The correct value is hundreds of microseconds, but it is entirely possible that I said “milliseconds” by mistake. It would not be the first time.
The 2012 realtime minisummit Posted Oct 25, 2012 23:05 UTC (Thu) by jake (editor, #205) [Link]
> it is entirely possible that I said “milliseconds” by mistake
it's also entirely possible that I mis-heard you ... in any case, i struck milliseconds and added microseconds ...
thanks,
jake
The 2012 realtime minisummit Posted Oct 26, 2012 14:22 UTC (Fri) by PaulMcKenney (subscriber, #9624) [Link]
Either way, thank you!
Of course, if FPGAs and and GPGPUs were to have large private memories that needed to be saved and restored on context switch, then we could see truly huge context-switch times. ;-)
|
|||||||||||||