* Phil Clayton:
>> WIth 5.5, I get:
>>
>> real 0m0.016s
>> user 0m0.003s
>> sys 0m0.010s
>>
>> With the latest version (SVN rev 1838), I get:
>>
>> real 0m0.416s
>> user 0m0.003s
>> sys 0m0.010s
>>
>> The ProofPower build runs Poly/ML about 180 times. The results I am seeing
>> are compatible with an overhead of about 400ms on each of these runs.
>
> I think this delay is actually when exiting Poly/ML. I have just this
> moment hit Ctrl+D to exit poly in a terminal but, as nothing happened
> immediately, I hit it again. In fact, both key strokes registered and
> I zapped the terminal. (It seems I really am that impatient!)
>
> The delay can still be seen with just
>
> time echo | poly
I pressed ^D quickly followed by ^C in GDB, and ended up with this on
the call stack:
// Now release schedLock and wait for a thread
// to wake us up. Use a timed wait to avoid the race with
// setting exitRequest.
initialThreadWait.WaitFor(&schedLock, 400);
I guess that's where those 400 milliseconds come from. :-/
Surely there's a better way to do this.
_______________________________________________
polyml mailing list
[email protected]
http://lists.inf.ed.ac.uk/mailman/listinfo/polyml
