Wiggle update in developer preview

Started by Seth Cooper

Seth Cooper Staff Lv 1

We've made a small adjustment to wiggle in the developer preview. It should be similar to the wiggle in the main release, but iterations shouldn't run as long so it should not cause scripts to run too long.

gramps Lv 1

This is a 3-year-old generic clone PC, about a middle-aged desktop win7 PC

win7 workstation bought around August 2010

Windows 7 Professional
Processor: Intel(R) Core(TM)2 Duo CPU E8400 @ 3.00GHz 3.00 GHz
Installed memory (RAM): 8.00 GB (7.90 GB usable)
System type: 64-bit Operating System

20130722-61d9318740-win_x86-devprev

Extract from my puzzle 749 notes, remembering that this is an uncomplicated puzzle of just 111 segments.

Before latest S/W update (two clients running here) …
7/22 C24Z in the game …
from default new 749t4a, loaded the 10223.544 from rb18 of t4 above via autosave, C28Z starting GAB2.5 loss – C40Z end gen 1 10264.789 – I19Z end gen 72(28) 10331.440 – I21Z canceled NFP, fastsaved, I23Z tDRW2.1.1 min2 max4 – DR1.2.1 NFP – J08Z DR1.3.4 10342.968 – J14Z DR1.4.1 10349.952 – J35Z DR2.2.1 10352.627 – K43Z DR3.2.2(2.4.1) 10381.894 – L01Z DR3.2.5 10382.406 – L15Z DR3.3.2 10383.952 – L41Z DR3.4.2 10398.000 – M07Z DR4.2.4 10408.753(14/145 - in progress) – N34Z DR5.2.4(5.2.3) 10420.556(12/146 AD4) – O19Z DR5.3.5 10420.808(12) – P15Z DR6.2.2 10421.123(12) – R11Z canceled DR6.4.8(6.2.2) NFP(13/152 AD5), fastsaved, trying LS Quake – 10421.720(13/154), lay this try aside

after the S/W update (mostly only one client running) …
logged out/in, new updated client, back in, U46Z tDRW2.1.1 min2 max4 – DR1.2.1 – DR1.2.1 10430.531(16) – W54Z DR1.4.3 10453.739(14/180) – …

7/23 Z56Z DR2.2.2(2.2.1) 10453.830(14) – A34Z DR2.2.4 10459.414(14) – J44Z DR4.4.7(4.3.7) 10478.365(12) –

summarizing:
before the new wiggle using TvdL enhanced DRW 2.1.1
started July 22nd 09:23 UTC
ended July 22nd 18:11 UTC mostly finished cycle 6
cycles 6
time 8h 48m
***about 1 1/2 hour per cycle

after new wiggle
started July 22nd 21:46 UTC
to deep rebuild 4.4.7 July 23rd 10:44 UTC
cycles 4
time 12h 58m
***about 3 1/4 hour per cycle

Bottom line:
before the new wiggle I could do two clients and it was tolerable
after the new wiggle one client seems to be about it and it runs half as fast (at least with tDRW)

the new wiggle may prove to be a show stopper for mid-powered middle aged computers*

Timo van der Laan Lv 1

My DRW 2.3.0 1 1 mini cycle (so only 1 part is rebuild)
Main: 12 minutes
Devprev: 13 minutes (but 1 qstab more which is 1 wiggle all(2) extra)
Devprev in background: 7 minutes

Another script where wiggle all(2) is called for every segment:
Main: 13 minutes
Devprev: 11:45 minutes

Conclusion: the playing field is now level.
Scripts are becoming too slow for confort.

And this was on puzzle 747, only 71 segments. Bigger puzzles will be a lot worse.

Timo van der Laan Lv 1

When you take into account that every cycle will try 1 more part to rebuild (starting at 4 in the first), assuming you just left the defaults there, the old one did 114 rebuild try s in 9 hours and the new one 66 in 13 hours. So some 12 an hour in the old and 5 an hour in the new.

Deleted user

Timo's last two posts - in #6 he shows that the new wiggle in devprev is about the same speed as main.

20 minutes later in post #7, he posts that it's a lot worse than half the speed.

How can this be ?

Timo van der Laan Lv 1

Gramps compared the previous devprev with the current one in his measurements. I compared the current devprev to the current main.
I posted my second post as an answer to gramps post but it does not show that clearly.