API tools faq
paste
Advertisement
entrpntr

buffered vs unbuffered saves

entrpntr
Feb 5th, 2018
46
0
Never
Add comment
Not a member of Pastebin yet? Sign Up, it unlocks many cool features!
text 4.59 KB | None | 0 0
raw download clone embed print report
  1. [11:26 PM] Exarion: [unknown] pokemonred40.dem 5.55 KB
  2. [11:26 PM] Exarion: ^failure from 00:05:07:52 with u-->U npc
  3. [11:28 PM] Exarion: and I just played the demo that generated a successful manip earlier, and it also got that failure from the same save file
  4. [11:31 PM] Exarion: and I redownloaded the successful save file that I posted on monday, then ran pokemonred3.dem through that, and it did not fail
  5. [11:31 PM] Exarion: so unless I am blanking on something, it has to be a save file difference, not printletterdelay
  6. [11:43 PM] entrpntr: agree with it being a save file difference
  7. [11:43 PM] entrpntr: exciting!
  8. [11:44 PM] entrpntr: everything is exactly the same on the first joypad hit in overworld after Nido catch
  9. [11:44 PM] entrpntr: it is diverging before entering Viridian gate
  10. [11:45 PM] entrpntr: C4 54 vs C7 53 for hra+hrs, internal divider is the same
  11. [12:00 AM] entrpntr: it is diverging when reloading first Viridian NPC
  12. [12:00 AM] entrpntr: lollolokl
  13. [12:00 AM] entrpntr: @Dabomstew
  14. [12:01 AM] Dabomstew: where could that possibly be stored
  15. [12:05 AM] Dabomstew: some old data must somehow be getting used
  16. [12:32 AM] entrpntr: ya, I didn't see anything that would be different, but I may very easily have overlooked something
  17. [12:36 AM] entrpntr: I just broke on 0:0F4D (joypadOverworld) until I found the divergence
  18. [12:36 AM] entrpntr: and it was a tile after NPC gets loaded (first tile his sprite is visible, at least with that breakpoint)
  19. [12:37 AM] entrpntr: 57 06 vs 56 06
  20. [12:38 AM] entrpntr: being hra+hrs at the first joypadOverworld with divergence
  21. [12:51 AM] entrpntr: sprite data at that point is also not the same
  22. [12:52 AM] Dabomstew: what fields differ
  23. [12:55 AM] entrpntr: C119; C102, C107, C108, C302, C306, C30A, C30E
  24. [12:55 AM] entrpntr: so, mostly players, and what I think is NPC's step counter (which wouldn't be surprising)
  25. [12:55 AM] entrpntr: first thought was that it might be buffered vs unbuffered save but idk
  26. [12:57 AM] entrpntr:
  27.  
  28. C10x (S) -- 01 00 0E 00 3C 01 40 03 02 0C 40 40 00 00 00 00
  29. C30x (S) -- 4C 50 08 20 4C 48 09 20 54 50 0A 22 54 48 0B 23
  30.  
  31. C10x (F) -- 01 00 0F 00 3C 01 40 00 03 0C 40 40 00 00 00 00
  32. C30x (F) -- 4C 50 88 20 4C 48 89 20 54 50 8A 22 54 48 8B 23
  33.  
  34. [12:58 AM] entrpntr: post-divergence, maybe I didn't pay close attention at what was going on pre-
  35. [1:04 AM] entrpntr: yeah I think it's buffered vs unbuffered
  36. [1:05 AM] entrpntr: and that must only matter when you load other sprites
  37. [1:06 AM] entrpntr: failure is on the buffered save
  38. [1:09 AM] Dabomstew: it's the very first random call after the last matching joypadoverworld
  39. [1:09 AM] Dabomstew: being called 22 cycles later
  40. [1:22 AM] Dabomstew: unbuffered arrives at UpdateNPCSprite later
  41. [1:22 AM] Dabomstew: checking more stuff
  42. [1:26 AM] Dabomstew: confirmed that it is a divergence in UpdatePlayerSprite (okay, that wasn't exactly rocket science)
  43. [1:33 AM] Dabomstew: oh it's actually 44 cycles not 22
  44. [1:33 AM] Dabomstew: derped on the internal divider stuff
  45. [1:33 AM] Dabomstew: in any case
  46. [1:34 AM] Dabomstew: the divergence for this specific part is caused by hitting or missing this https://github.com/pret/pokered/blob/46a94c63fc287e7290502776d02648476bc44171/engine/overworld/movement.asm#L71
  47. [1:34 AM] Dabomstew: failure (buffered) takes the jump, success (unbuffered) doesn't and spends 44 additional cycles
  48. [1:34 AM] Dabomstew: what it actually means is quite another question
  49. [1:37 AM] Dabomstew: but it should imply a ~44/256 chance of an RNG divergence between buffered and unbuffered for each NPC timer reset
  50. [1:39 AM] Dabomstew: the crucial bit of state being c107
  51. [1:45 AM] Dabomstew: just means that yes: buffered vs non-buffered matters for everything
  52. [1:46 AM] Dabomstew: unless you are doing a manip in an npcless environment
  53. [1:46 AM] Dabomstew: but it also means that it is 100% reasonable that a manip does not show any signs of buffered vs non-buffered mattering even if it does load a few NPCs
  54. [1:58 AM] entrpntr: yeah and for Moon it probably only matters until the first ladder
  55. -
  56. [11:44 PM] gifvex: the difference between un/buffered saves is d528 wPlayerMovingDirection being zero (unbuffered) vs nonzero (buffered)
  57. [11:45 PM] gifvex: it causes c107 PlayerIntraAnimFrameCounter to start at a different value and take the jump @Dabomstew noted previously at a different time https://github.com/pret/pokered/blob/46a94c63fc287e7290502776d02648476bc44171/engine/overworld/movement.asm#L71
  58. [11:46 PM] gifvex: in fact since it's an anim counter, the difference could have been seen this whole time (these are same frame of a demo playback from un/buffered saves) http://imgur.com/a/ZG4Xd
Advertisement
Add Comment
Please, Sign In to add comment
Advertisement
Public Pastes
Advertisement
We use cookies for various purposes including analytics. By continuing to use Pastebin, you agree to our use of cookies as described in the Cookies Policy.  OK, I Understand
Not a member of Pastebin yet?
Sign Up, it unlocks many cool features!