Talk:Mapping/Compile time

From UFO:AI
< Talk:Mapping
Revision as of 16:11, 21 August 2008 by BTAxis (talk | contribs)
(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to navigation Jump to search

Linux time

for "time make clean maps" do i have to add up the times?

real    8m23.553s
user    15m39.927s
sys     0m2.248s

means 24min? --Blondandy


To be sure I roughly stopped my time with a clock as well and the first value "real" seems to be the correct time. Haven't studied the manual for the time command yet though :) In your example you either have a very fast computer (unless you have very good compile-time optimizations I guess) or not all of the maps were compiled. I think none of these is the correct value. To be sure all my maps were compiled I actually manually removed all .bsp files and simply ran "time make maps" (at the time I didn't know one could do "clean maps" as well). --Hoehrer 07:00, 20 August 2008 (UTC)

it does seem insanely fast. I will double check. i am not 100% that make clean maps does compile all maps regardless. --Blondandy 11:47, 20 August 2008 (UTC)

Also to make _really_ sure how long it took you could do something like this I guess:

date >> time.txt;time make clean maps;date >> time.txt

Then check time.txt for start&end time :D --Hoehrer 15:36, 20 August 2008 (UTC)

Threads

How do I activate threads for map compiling? Also, if it is active, how is the number of threads determined? --Hoehrer 07:07, 20 August 2008 (UTC)

between us a mac user (can't recall who) Richlv and I modified the makefile. make maps detects number of cores and uses them all. you can verify this by looking at output. the Threads:#2 lines are from ufo2map. --Blondandy 11:42, 20 August 2008 (UTC)

Nice, never noticed that :) On a related note: I always hear/read that one should always use one more thread than available cores/CPUs for multi-threading stuff. I'm no expert on that matter so is that an urban legend, only true in certain cases or a general good idea? --Hoehrer 15:40, 20 August 2008 (UTC)

why not tweak your makefile, to use cores+1 threads, and see if it is faster?--Blondandy 16:54, 20 August 2008 (UTC)

Compiling bunker.map

hoehrer (r18463 / Intel Core2 Quad 2.66GHz / 64bit Ubuntu 8.04 Linux 2.6.24-19-generic)

  • time ./ufo2map -t 1 base/maps/bunker.map -> 2m25.677s
  • time ./ufo2map -t 2 base/maps/bunker.map -> 1m3.230s
  • time ./ufo2map -t 4 base/maps/bunker.map -> 0m28.762s
  • time ./ufo2map -t 5 base/maps/bunker.map -> 0m28.406s
  • time ./ufo2map -t 8 base/maps/bunker.map -> 0m28.870s
  • time ./ufo2map -extra -t 1 base/maps/bunker.map -> 12m5.027s
  • time ./ufo2map -extra -t 2 base/maps/bunker.map -> 4m50.602s
  • time ./ufo2map -extra -t 4 base/maps/bunker.map -> 2m0.922s
  • time ./ufo2map -extra -t 5 base/maps/bunker.map -> 2m4.148s
  • time ./ufo2map -extra -t 8 base/maps/bunker.map -> 2m1.766s

btaxis (r18467 / Intel Core2 Duo 2.13GHz [E6400] / Windows XP Pro Corporate 32bit SP3)

  • ufo2map -t 1 base/maps/bunker.map: sum: 149 seconds elapsed
  • ufo2map -t 2 base/maps/bunker.map: sum: 172 seconds elapsed
  • make -j2 using Makefile.win, -extra enabled, all bsps deleted beforehand
    • starttime = 14:00:24.06
    • endtime = 15:41:32.12