~dr|z3d
@RN
@StormyCloud
@T3s|4
@T3s|4_
@eyedeekay
@orignal
@postman
@zzz
%Liorar
%acetone
%cumlord
+DirtyHarry
+FreefallHeavens
+Xeha
+ardu
+bak83_
+mareki2p
+onon_
+profetikla
+r00tobo
+radakayot
+segfault
+uop23ip
AHON1
Arch2
Danny
DeltaOreo
FreeB
Irc2PGuest53192
Irc2PGuest59581
Irc2PGuest64993
Irc2PGuest70134
Irc2PGuest96449
Meow
Onn4l7h
Onn4|7h
b4dab00m
boonst
duck
maylay
not_bob_afk
phobos_
pisslord
poriori_
qend-irc2p
shiver_1
simprelay
solidx66
thetia
u5657
zer0bitz_
dr|z3d
running ant exec sub-tasks in parallel makes for a useful improvement in build times for pack200 target. around 25% or so faster.
zzz
unfortunately, I call b.s.
dr|z3d
how would you know, zzz?
dr|z3d
*** laughs. ***
dr|z3d
you abandoned pack200 years ago. you're clueless :)
zzz
i was curious, so I looked at your rev, and I RTFM'ed the ant spec. that's how
dr|z3d
well, you're still clueless in that case. can do better :)
zzz
I'm saying your change is a noop, your 'results' are phantom
dr|z3d
well, you're wrong.
dr|z3d
demonstrably faster, reproducibly.
dr|z3d
feel free to run your own tests, I think you'll be pleasantly surprised.
zzz
canon 35s; cannon 43s
dr|z3d
that's without the optimizations?
zzz
current heads
dr|z3d
it'll be slower on cannon because I'm not omitting any jars from the process.
dr|z3d
without the optimizations, I was seeing a consistent 30s.
dr|z3d
with optimizations, 22s.
zzz
your <parallel> does nothing. reread the ant spec
zzz
the added & is what does it, but without any thread limit
zzz
I believe what you are grasping for is xargs -P
dr|z3d
sure, backgrounding the pack jobs helps.
dr|z3d
maybe that's the optimization then. :)
dr|z3d
yeah, looked at xargs, that's the next stage.
zzz
ant doesn't know anything about shell for loops. there is only one task that ant sees: the single <exec> for the OS. There's nothing for ant to parallelize. reread the spec
zzz
don't do two things hoping one of them works
zzz
and guessing wrong
dr|z3d
that wasn't quite the approach I was taking, but thanks for the input all the same.
dr|z3d
curiously, xargs -P yields exactly the same time.
dr|z3d
gnu's parallel does slightly better, but not by much.
dr|z3d
<arg value="find pkg-temp/lib/*.jar pkg-temp/webapps/*war -type f | parallel -j 8 'echo Pack200 compression for: {} && mv {} {}.jar && pack200 --effort=6 --modification-time=latest --segment-limit=-1 -g {}.pack {}.jar && rm -f {}.jar'" />
dr|z3d
anyways, the reasoning might have been slightly awry, but the outcome is the same. faster pack200.
dr|z3d
so apologies for calling you clueless :)