HCE - 2009.027

==============

- Chg:0106 GE Implementing a new eff115 (Autonomous Sensor Intelligence) that

determines when the AI should turn on active sensors and when

to prompt the player to do the same. Added a new log type

option to the command line for this, emcon_ai . Also added

a new command line option to switch back to a fixed version

of the old eff115 that will properly notice if a group's

radars are already on. The option is -e

i.e. c:\games\HCE\Winharp32.exe -e

would revert to the fixed old eff115 that is much more apt

to turn on radars than the new code.

(thanks Lots of People)

http://harpgamer.com/harpforum/index.php?a...amp;showbug=106

 

go to it, more detail in the Issue Tracker. I expect more than a "yes it works" for this.

Interesting Performance results with 2009.030 running HDS III IOPG Scenario 4.0.

 

Old eff115 (using -e switch on command line) (it shows as eff 15 in the data below, not eff115)

185999 command.c:247 - 86000 cycles

eff 0 264.79 seconds 3078.903 microsecs/call, 325 calls/sec

eff 1 0.80 seconds 9.298 microsecs/call, 107554 calls/sec

eff 2 0.51 seconds 5.941 microsecs/call, 168335 calls/sec

eff 3 0.12 seconds 1.440 microsecs/call, 694644 calls/sec

eff 4 38.19 seconds 444.061 microsecs/call, 2252 calls/sec

eff 5 183.96 seconds 2139.090 microsecs/call, 467 calls/sec

eff 6 127.09 seconds 1477.767 microsecs/call, 677 calls/sec

eff 7 0.12 seconds 1.368 microsecs/call, 731005 calls/sec

eff 8 0.14 seconds 1.590 microsecs/call, 629050 calls/sec

eff 9 0.13 seconds 1.531 microsecs/call, 653065 calls/sec

eff 10 1.98 seconds 23.055 microsecs/call, 43375 calls/sec

eff 11 0.55 seconds 6.406 microsecs/call, 156109 calls/sec

eff 12 13.47 seconds 156.591 microsecs/call, 6386 calls/sec

eff 13 25.49 seconds 296.444 microsecs/call, 3373 calls/sec

eff 14 90.56 seconds 1052.977 microsecs/call, 950 calls/sec

eff 15 1884.63 seconds 21914.335 microsecs/call, 46 calls/sec

eff 16 0.12 seconds 1.366 microsecs/call, 732287 calls/sec

eff 17 0.16 seconds 1.845 microsecs/call, 541922 calls/sec

eff 18 44.18 seconds 513.740 microsecs/call, 1947 calls/sec

eff 19 0.16 seconds 1.844 microsecs/call, 542198 calls/sec

eff 20 20.55 seconds 238.984 microsecs/call, 4184 calls/sec

 

New eff115:

185999 command.c:247 - 86000 cycles

eff 0 254.38 seconds 2957.902 microsecs/call, 338 calls/sec

eff 1 0.76 seconds 8.786 microsecs/call, 113817 calls/sec

eff 2 0.47 seconds 5.443 microsecs/call, 183709 calls/sec

eff 3 0.12 seconds 1.356 microsecs/call, 737218 calls/sec

eff 4 43.07 seconds 500.843 microsecs/call, 1997 calls/sec

eff 5 229.57 seconds 2669.368 microsecs/call, 375 calls/sec

eff 6 82.31 seconds 957.136 microsecs/call, 1045 calls/sec

eff 7 0.12 seconds 1.342 microsecs/call, 745036 calls/sec

eff 8 0.13 seconds 1.517 microsecs/call, 659030 calls/sec

eff 9 0.13 seconds 1.498 microsecs/call, 667657 calls/sec

eff 10 1.86 seconds 21.656 microsecs/call, 46177 calls/sec

eff 11 0.53 seconds 6.143 microsecs/call, 162800 calls/sec

eff 12 361.18 seconds 4199.721 microsecs/call, 238 calls/sec

eff 13 23.56 seconds 273.916 microsecs/call, 3651 calls/sec

eff 14 85.86 seconds 998.382 microsecs/call, 1002 calls/sec

eff 15 1768.34 seconds 20562.035 microsecs/call, 49 calls/sec

eff 16 0.12 seconds 1.338 microsecs/call, 747213 calls/sec

eff 17 0.15 seconds 1.729 microsecs/call, 578502 calls/sec

eff 18 56.95 seconds 662.179 microsecs/call, 1510 calls/sec

eff 19 3.01 seconds 35.002 microsecs/call, 28570 calls/sec

eff 20 19.90 seconds 231.391 microsecs/call, 4322 calls/sec

 

In this particular instance the new eff115 was more efficient, running for 1768 seconds over the course of the scenario (5 minutes:1 compression on autoplay) vs 1885 seconds for the old, much more simplified and not ideal response code. I was also shocked that either way this EMCON stuff is by far the most time consuming part of the game, who would have figured that deciding whether to turn on sensors is that CPU intensive.

 

for those at home not having a clue how to do this...

 

winharp32.exe -e -l perf

the -e uses the old eff115, omitting the -e uses the new eff115, the -l perf tells the game to log performance information. Find the results in ge.log unless you specified a particular log file with the -f command line option.

Well nobody besides me has posted any performance info or impact but the other release notes make it clear the new behavior is in place and working.