TAS (tool-assisted) demos: part 2

[GarrettChan]

12 minutes ago, Grazza said:

Going for the hoarding world record, eh?

Obviously 4shock ran too fast and he crossed the wormhole, lol.

[4shockblast]

6 hours ago, Grazza said:

Going for the hoarding world record, eh?

Oops, I didn't fix that from copy-pasting. :P Fixed.

[almostmatt1]

Scythe Map 03 UV-Pacifist with Turbo in 0.80 seconds - sc03x-080.zip

[almostmatt1]

Ultimate Doom E1M1 UV-Max in 0:18.97 - e1m1xm18.zip

Edit a few days later: Just realized, it's reality too. :)

 

Edited July 23, 2021 by almostmatt1

[almostmatt1]

Alien Vendetta Map 25 NoMo in 0:36.91 - av25xo36.zip

UV-Speed would be a pretty cool demo but that hallway section would be pretty annoying.

YouTube link:

Spoiler

 

 

[RockyGaming4725]

Ultimate Doom E1M8 Nightmare! in 17.40
e1m8nx-1740.zip

Video link - 
https://youtu.be/W7xBR78UljA

 

[RockyGaming4725]

Ultimate Doom E1M8 UV Speed in 17.29
e1m8x-1729.zip

Video link - 
https://youtu.be/bwCpG81Cd2A

 

[Topflex]

Ultimate Doom E1M8 NoMo in 15.94

 

I saw @RockyGaming4725 use this glide, so I decided to put together this TAS of it.

Not optimized at all, and it's also viable in RTA I think. It's really a great find!

 

e1m8xo1594.zip

[Dimon12321]

@Topflex I don't think 14 secs is possible. You need to win the whole second in order to cut one period of floor damage.
However, everything could be. I didn't know a glide could be done in that direction.

Edited July 28, 2021 by Dimon12321

[RockyGaming4725]

@Topflex Damn! That's awesome to see. Also, by adding a wallrun at the final wall (and the one before it) I was able to lower it substantially to 15.37. Not taking credit for this of course, I just edited some tics. Feel free to submit this one if you like, just helping out! :)

e1m8xo1537.zip

Edit: The glide is super hard rta but I did manage to get a 26 with it. Had super high hp, though. Maybe someone else will grind for a better run.

Edited July 29, 2021 by RockyGaming4725

[almostmatt1]

Ultimate Doom E1M8 No Monsters in 0:14.77 - e1m8xo1477.zip

YouTube:

Spoiler

 

I finished building this and realized this trick would make for a faster UV-Speed demo than is possible via the pinky route, too. So I turned on monsters, but the barrel damage changed which caused an instant desync, because of course life couldn't be that easy. :) Props to RockyGaming4725 for finding the glide and Topflex for their above demo which inspired me to try this. 

[almostmatt1]

Doom 2 Map 07 Pacifist with turbo in 2.31 seconds - lv07xt-231.zip

[JCD]

Ultimate Doom e3m7 Coop Uv-Speed in 0:26.97
c3s7x026.zip

Low effort TAS. Show a coop void glide and a Zero press to open red doors early.

 

[MikeTheMerc]

Boomedit.wad MAP01 UV Speed in 19.17 seconds.

 

Considering this is my first TAS and I barely put any time into making it, I think I did an OK job.

boomedx1917.zip

Edited August 1, 2021 by MikeTheMerc

[almostmatt1]

TLDR: XDRE 2.06 brute force is approximately 7 times faster than XDRE 2.20 brute force.

 

@alps mentioned to me that XDRE 2.06 brute force felt faster to him than 2.20, so I ran a test. 2.06 starts a demo in CL3 and skill 3, whereas 2.20 starts a demo in CL17 skill 4, so I used the same demo and map for testing: E1M1, CL3, skill 3, all other header information the same, no demo inputs. I looked over an input range of 50:-49 for run, strafe and turn, meaning there was 100*100*100=1,000,000 million input combinations (not that this number matters - just note that it was the same range between versions). I searched for Y=0, which here is an unachievable result, meaning XDRE would need to check the whole input range. All other programs were closed while the brute force ran. For each XDRE version, I ran three of these tests and the result for that version was the same each time, to the second. I measured the time it took for the whole range to be examined.

 

XDRE 2.20 took 16:26 to run the test, and version 2.06 took only 2:17, meaning that in my testing version 2.20 took 7.19 times longer to perform the brute force process. Please keep in mind that the exact times you get really relies on your own computer - I'd imagine the comparative difference will be roughly similar but the exact times you get may well be faster or slower.

 

Version 2.20 allows tracking of Z position, thing information, sector information, and brute forcing for a Z position, which 2.06 does not (there are other smaller changes but these are the biggest and most obvious). It also allows you to change the header information within the program, which 2.06 does not seem to do. These things can be very useful and so 2.20 still of course holds some merit if you need them, but if like me your TASing involves a lot of brute forcing, it would appear that using version 2.06 is going to save you a lot of time.

 

Less interestingly to anyone else - I've recently started overclocking, and before running these tests I was comparing version 2.20 brute forcing on an overclocked vs. stock clock CPU, which resulted in stock taking about 22% more time. I ran the 2.06 test with a CPU overclock as high as I could go and it took 1:43... meaning that if this improvement stays consistent in most circumstances, stock CPU in version 2.20 vs OC CPU in version 2.06, I'm now getting brute forces 9.57x faster. This is a good day. :)

[Dimon12321]

@almostmatt1 I'm wondering if it has something to do with brute forcing algorithm. It seems to be downgraded, as I don't see any other reasons why its performance got so low.

My only guess is that XDRE 2.20 keeps track of all game's parameters regardless of conditions we set in brute force.
Also, we always set 1-3 conditions, not all of them. So maybe it stresses itself with checking empty required conditions?

@vita What's up, man?

Edited August 2, 2021 by Dimon12321

[almostmatt1]

I wish I knew enough about programming to check out the code and properly understand the differences! However while the source code for 2.20 is available I couldn't find the 2.06 source unfortunately. I would absolutely love to understand what's going on. 

I've done some brute forcing in other places and that 7x speed in the E1M1 test does seem to be an outlier but results in other places are still very positive. I've tried brute force tests in 3 other situations and version 2.06 has been roughly 2.8, 3.9 and 5.2 times faster than 2.20.

I should also mention that at one point I was using version 2.22 on Linux and it also felt faster than using 2.20 on Windows. But I didn't test it so that is totally anecdotal, and because the entire OS is different it certainly stops becoming a valid comparison of XDRE versions.

[pjotr_]

zone 300 map 1 TAS in 21.66 

tas2166.zip

[almostmatt1]

Noye map 21 Nightmare in 1.40 - ny21nx140.zip

Noye map 21 Nightmare with turbo in 0.34 - ny21ntx034.zip 

[Topflex]

Doom 2 MAP18 UV-Speed in 6.91

 

lv18x691.zip

[Dimon12321]

Resonance

Map 4 NM-Speed in 0:04

rs04nx004.zip

 

Map 7 NM-Speed in 0:01

rs07nx001.zip

 

Both runs can be improved. Especially Map 4.
 

Videos:

Spoiler

 

 

 

Edited August 15, 2021 by Dimon12321

[almostmatt1]

Eviternity Map 16 No Monsters in 30.77 - evit16ox30.zip

 

YouTube:

Spoiler

 

 

[almostmatt1]

Eviternity Map 01 No Monsters in 18.34 - evit01ox18.zip

 

YouTube:

Spoiler

 

 

[Dimon12321]

Icarus: Alien Vanguard

Map 1 UV-Max in 1: 45

ic01x145.zip

 

 

[RockyGaming4725]

E1M8 UV Speed in 15.11
Uses the new void glide!
e1m8x1511.zip
Video link - https://youtu.be/ZP-L-iblsAg

[MatrixCL]

First time that I commit a TAS run, so do tell me if this is not the way it's supposed to be done. I mostly copied the routes of the UV max records on IL's, adapted where the extra ammo allowed shortcuts, played at 25% game speed, and redid whatever part didn't go right (i.e. -recordfromto...a lot).

 

Valiant episode 1 UV Max in 16:46. (actual WR is 24:57)

Valiant episode 2 UV Max in 24:58. (actual WR is 39:28)

 

All level times are faster than the records on the levels individually.

vae1-1646.zip

vae2-2458.zip

 

Edit: the correct zip file for episode 2 is on the first post of the next page.

Edited September 22, 2021 by MatrixCL

[RockyGaming4725]

TNT Map 4 NoMo in 5.49
ev4xo549.zip
Video link - https://youtu.be/co3yIXNyzfw

 

[almostmatt1]

Kinda feel like a bit of a jerk for posting this right after RockyGamings, but it looked like a fun TAS. :)

TNT Map 4 NoMo in 4.77 - ev04ox477.zip

YT:

Spoiler

 

 

Edited September 23, 2021 by almostmatt1

[RockyGaming4725]

15 hours ago, almostmatt1 said:

Kinda feel like a bit of a jerk for posting this right after RockyGamings, but it looked like a fun TAS. :)

TNT Map 4 NoMo in 4.77 - ev04ox477.zip

And a jerk you are.

Just kidding of course, once again an excellent tas! Didn’t know you could wallrun on those walls as well :p

[Dimon12321]

On 9/21/2021 at 12:00 AM, MatrixCL said:

Valiant episode 2 UV Max in 24:58. (actual WR is 39:28)

 

vae2-2458.zip

Did you double-check what replay you were going to send? It doesn't look like a TAS and the final time is wrong - 42:13.