One thing: could you please add a performance comp...
# coroutinesv
voddan
02/22/2017, 4:38 PMOne thing: could you please add a performance comparison of different techniques for reference? Statements like
This is the fastest solution for this particular probleme
elizarov
02/22/2017, 5:30 PMvoddan: I will. I’ll need to make sense of all the results myself, first. I can show them, but I cannot yet fully explain them.
v
voddan
02/22/2017, 6:51 PMYou just roused my curiosity with that comment. How am I supposed to sleep tonight without knowing the results??
voddan
02/22/2017, 6:54 PMOn a more serious note, I think it would be right to share the benchmarks even without an interpretation.
voddan
02/22/2017, 6:56 PMWho knows, maybe some of the people in the crowd could help with the investigation itself 😉
e
elizarov
02/22/2017, 7:13 PMThe real problem is that I need to write/setup a test harness around them with a proper warmup and stuff, before presenting any timing results. I just had no time to do it yet. Just running them as-is produces quite pseudo-random results with respect to time.
v
voddan
02/22/2017, 10:34 PM👍
c
cedric
02/22/2017, 10:58 PM@elizarov The overhead should be minimal, shouldn’t it? Just the state machine overhead and the context switching (which you need to pay for with a regular threaded approach anyway)
3 Views