Re-add set waitfor back to the documentation so i don't have to worry about it going away.

ID:1992635
 
Dec 4 2015, 6:20 am (Edited on Dec 4 2015, 6:35 am)
Applies to:DM Language
Status: Open
Issue hasn't been assigned a status value.
Doing set waitfor = 0 in the target proc is massively faster than using spawn()

Noticeable if you do a loop that calls something that sleeps, and you want to make it happen simultaneously rather then one after the other, calling a lot of spawn()s adds up quickly.

Proof:
                             Profile results (total time)
Proc Name                              Self CPU    Total CPU    Real Time        Calls
----------------------------------    ---------    ---------    ---------    ---------
/mob/proc/testwaitforspawn3               2.473        2.474    28438.984        20000
/mob/proc/testwaitfor3                    2.018        2.019    29180.518        20000
/mob/proc/testwaitforspawn2               1.002        3.479    28439.971        20000
/mob/proc/testwaitfor2                    0.031        2.049    29180.541        20000
/mob/verb/testwaitforspawn                0.023        0.976        0.976            1
/mob/verb/testwaitfor                     0.021        2.059        2.059            1
/mob/verb/testwaitforprocoverhead         0.020        0.035        0.035            1
/mob/proc/testwaitforprocoverhead2        0.015        0.015        0.020        20000
/mob/proc/testwaitforprocoverhead3        0.000        0.002        0.003        20000


mob
    verb/testwaitfor()
        for (var/i = 0, i < 20000, i++)
            testwaitfor2(i)
    proc/testwaitfor2(i)
        set waitfor=0
        testwaitfor3()
    proc/testwaitfor3()
        sleep 1

    verb/testwaitforspawn()
        for (var/i = 0, i < 20000, i++)
            testwaitforspawn2(i)
    proc/testwaitforspawn2(i)
        spawn
            testwaitforspawn3()
    proc/testwaitforspawn3()
        sleep 1


    verb/testwaitforprocoverhead()
        for (var/i = 0, i < 20000, i++)
            testwaitforprocoverhead2(i)
    proc/testwaitforprocoverhead2(i)
        testwaitforprocoverhead3()
    proc/testwaitforprocoverhead3();



The reason this is faster, seems to be because it doesn't have to copy the vars to the spawn()'ed section.

Related: https://github.com/tgstation/-tg-station/pull/13443
http://web.archive.org/web/20060722121013/http:// developer.byond.com/docs/notes/225.html
Dec 4 2015, 7:56 am
The reason it's faster is because the procedure returns when it hits a sleeping procedure or function and then continues to execute.
Dec 4 2015, 10:23 am
I mean it has less overhead then spawn does...
Dec 4 2015, 10:23 am
the profiles speak for themselves...
Dec 4 2015, 11:17 am
+1 on this-- it probably also helps that you're not explicitly forcing a context switch anymore.
Dec 4 2015, 11:29 am
In response to MrStonedOne
MrStonedOne wrote:
the profiles speak for themselves...

Yeah, I'm not arguing, but they're not doing the exact same thing either way
Dec 4 2015, 11:43 am
In response to Super Saiyan X
I think that's the confusing part. He wasn't wishing spawn() to go away, or wanting anyone to necessarily advocate using waitfor instead. He just wants it added back to the documentation so it doesn't just get silently deprecated while he's using it.
Dec 4 2015, 12:04 pm
functionally speaking, in almost every case you'd want to use spawn without a time argument, waitfor = 0 does the same thing..

Ya, internally they are very different, syntax wise they are very different, but spawn is and has always been, primarily used to prevent a sleep from hanging the stack.

set waitfor at the proper place in the stack does the exact same thing, with the bonus of having significantly less overhead.
Dec 4 2015, 7:11 pm
spawn is very expensive, so avoiding using it is a massive +
Dec 23 2015, 1:48 pm
You can set waitfor for procs and the like---waitfor functions similarly to spawn(), but has considerably less overhead if it encounters a sleep and none at all if it doesn't---yet it's completely and totally undocumented.

It would be great if this was fully documented on the DM reference.
Dec 23 2015, 1:50 pm
In response to Fox P McCloud
Oct 8 2017, 10:54 pm
Bump.