Post subject: VBA v19 vs. v20. How to compare runs?
Player (87)
Joined: 1/15/2006
Posts: 333
Location: Bangkok, Thailand
So, VBA v20 had a GB core overhaul, which in general, seems to be an improvement... ...except it's not, at least in terms of frame count. Upon a new glitch discovery, I was intending to redo my Pokémon Blue run, but I find that with v20, I'm losing frames absolutely everywhere. At first, I thought it was just going to be at the intro screens, due to a longer startup time, but it's not just there. Every dialog seems to lose a frame, entering or exiting a building seems to lose 4 frames, etc. A little over one minute into the run, I'm already 40 frames behind, and the only difference is that I've theoretically made 7 frames worth of improvement. My fear is that even though the run should be a minute or so faster, the timing difference in v20 will eat most, if not all of that. The question is, if a slower v20 run which theoretically should be faster than an existing v19 run were to be submitted, would it be accepted? Alternatively, would a run made with v19 still be accepted?
print reduce(lambda x,p:p/2*x/p+2*10**1000,range(6643,1,-2))
Banned User, Former player
Joined: 3/10/2004
Posts: 7698
Location: Finland
One relevant question: Which one, v19 or v20, emulates more correctly in this regard? Are those additional frames correctly emulated in v20 (ie. they do happen in the original console), or is it some kind of bug in v20?
Joined: 3/7/2006
Posts: 720
Location: UK
It's not a bug. v19 runs the GB slightly too fast (probably related to why certain bugs occurred on there - lazy processing)
Voted NO for NO reason
Banned User, Former player
Joined: 3/10/2004
Posts: 7698
Location: Finland
Well, if it was an emulation bug in v19 which allowed a faster run, and this bug has been fixed in v20, then I suppose there shouldn't be any problem. Just remember to clearly explain this in the submission. If you can more or less accurately estimate the amount of "additional" frames introduced by the more correct v20 in your submission text, the better.
Joined: 8/27/2006
Posts: 883
Is it possible to know how faster it could emulate ? It could be an approximation, but just to be sure that a run don't get rejected by being faster but have a greater frame count.
Active player (410)
Joined: 3/16/2004
Posts: 2623
Location: America, Québec
Since we know that v19 is inaccurate and runs might not be as quick in v20, personally, I think you should go with v20. I'm currently doing SCV4 with Angerfist and encounters the same problem. However, the extra losing frames, from now on, seems to only happen during fadeout screen transition. To solve the problem, I am using, thanks to moozooh, an in-game timer memory address that stops during screen transition. So far it seems 100% reliable. There might be a similar trick for your game. Or you could do like JXQ and Cardboard did with their Bionic Commando run. They made a table that compares the U with J version.