sdy@ssc-bee.boeing.com (Ugly Bag of Mostly Water) (04/12/90)
<jamesth@microsoft.UUCP writes:> ... >Each system had hard real-time requirements >which varied from 20 milliseconds to one second. Ada tasking was not >used in any of these systems, due to high overhead and a poor fit of >Ada tasking primitives to the system requirements. No constructs were >used which would have required garbage collection. We have recently flown 2 missiles that have had real-time requirements of 10ms, 20ms and 100ms. We managed to do this by using a vendor supplied (Ada source) run-time system and "normal" Ada tasking (rendezvous et al). We were unable to use a rendezvous attached to a hardware interrupt and utilized the RTS for that interface but all other tasking interactions were done in standard Ada. While the speed was not the greatest we were able to run fast enough to not have any problems related to timing delays. This was on a custom MC68020 running Verdix Ada and RTS. Design wise we have been quite happy to use the Ada tasking model in that when we recently tried to switch RTS it was fairly painless compared to other programs who have tried to change their custom executive. This is not to say that Ada is yet perfect but we have been more than pleased with the progress that vendors have been making in the near past. Now if they keep up the progress we may even get a blazing RTS/Tasking model in time for our next program. Steven D. Yee SEA LANCE Software Design Group BOEING Aerospace and Electronics -- Steven D. Yee >>> my employer does not share my opinions <<< uw-beaver!ssc-vax!ssc-bee!sdy >>> (that's because I'm always right! ;-) <<< I don't get even ... I get ahead.