Editor, Experienced Forum User, Published Author, Expert player
(3572)
Joined: 11/30/2014
Posts: 2744
Location: US
Sour wrote:
Alyosha wrote:
NESHawk happens to set to FF
Yea, I just learned this from trying to add support for BK2 movies to Mesen.
natt wrote:
The vrc6 tests at the end are based on my best understanding of a translation of some Japanese documentation, but to my knowledge they've never been verified on real hardware.
Editor, Experienced Forum User, Published Author, Expert player
(3572)
Joined: 11/30/2014
Posts: 2744
Location: US
The values at the end are read from un-initialized RAM, which NESHawk happens to set to FF, so at least that much I'm not worried about. I'll look into the $06's a bit more though.
Some console testing was done on this and the results were highly variable.
Editor, Experienced Forum User, Published Author, Expert player
(3572)
Joined: 11/30/2014
Posts: 2744
Location: US
Played through the first race in TAS studio and seeked to various points in the run without errors. I'll need specific reproduction steps to do anything here.
Editor, Experienced Forum User, Published Author, Expert player
(3572)
Joined: 11/30/2014
Posts: 2744
Location: US
PikachuMan wrote:
I just received a error like this:
System.IndexOutOfRangeException: Index was outside the bounds of the array.
at BizHawk.Emulation.Cores.Nintendo.NES.PPU.FrameAdvance()
at BizHawk.Emulation.Cores.Nintendo.NES.NES.FrameAdvance(Boolean render, Boolean rendersound)
at BizHawk.Client.EmuHawk.MainForm.StepRunLoop_Core(Boolean force)
at BizHawk.Client.EmuHawk.MainForm.ProgramRunLoop()
at BizHawk.Client.EmuHawk.Program.SubMain(String[] args)
Seems like a NESHawk error but I'll need a lot more information then that in order to debug it.
Game?
Movie file?
Repro steps?
Editor, Experienced Forum User, Published Author, Expert player
(3572)
Joined: 11/30/2014
Posts: 2744
Location: US
Finally had a chance to get Ninja Golf working and fix A7800 save states.
Here is a bk2 that runs in the current dev build of BizHawk It should be usable to make a new encode for Ninja Golf that has correct color palette.
http://tasvideos.org/userfiles/info/38458302684398268
It should probably also replace the published movie file since it works in current builds and won't cause crashes but I guess that's up to you.
Editor, Experienced Forum User, Published Author, Expert player
(3572)
Joined: 11/30/2014
Posts: 2744
Location: US
Yeah having an updated test rom set hosted in one place would be really valuable. Stumbling across test roms by chance in the forums is frustrating.
Also, we really need to move beyond these fundamental test roms , and the first step in that is just accounting for all of them. True's console testing shows that we still don't have sufficient information to TAS games that require ppu tick level accuracy but there is nothing new to test to point out what is wrong.
Editor, Experienced Forum User, Published Author, Expert player
(3572)
Joined: 11/30/2014
Posts: 2744
Location: US
I didn't notice this until now. Almost certainly that RTA run was captured from a reset. Reset can change these internal timers slightly, you can see a similar effect in dragster.
Editor, Experienced Forum User, Published Author, Expert player
(3572)
Joined: 11/30/2014
Posts: 2744
Location: US
Sour wrote:
On a north american NTSC NES, count_errors should technically only have a maximum of 4 different results (there are only 4 possible alignments CPU/PPU alignments) - but I could be wrong on this.
Maybe, but since it's not characterized yet no one really knows. Based on some console testing with read2004.nes there might be more.
Editor, Experienced Forum User, Published Author, Expert player
(3572)
Joined: 11/30/2014
Posts: 2744
Location: US
I think is is still a bit misunderstood, but having '*' in the count error tests is passing, not failing. FCEUX is failing the test by having no '*' . The problem is that the test is inconsistent on things like powerpack since it DMC counter is always running.
I still suspect that if run from power on, and on a cart, there will be consistent (or only a small number) of actual test results.
If anyone is able to do this I think it will really clear things up and help understand DMC start up state.
Editor, Experienced Forum User, Published Author, Expert player
(3572)
Joined: 11/30/2014
Posts: 2744
Location: US
Also, if you don't mind the very dated gameplay experience, Atari 2600 has many possibilities for TASes where you can learn a lot in a pretty simple environment.
Editor, Experienced Forum User, Published Author, Expert player
(3572)
Joined: 11/30/2014
Posts: 2744
Location: US
Oh that's cool, thanks for linking that spikestuff, a lot of new test cases there.
Looks like there is more apu work to do in NESHawk then I thought.
Also @Sour nice work with Mesen! I look forward to being able to do some TAS type testing with it in the future!
EDIT: what is your criteria for passing count_errors, count_errors_fast, and read2004? It seems like they all have variable results from hardware testing, and I'm pretty sure NESHawk is at least emulating the related mechanics correctly.
Editor, Experienced Forum User, Published Author, Expert player
(3572)
Joined: 11/30/2014
Posts: 2744
Location: US
This reminds me of TheAxeMan's Final Fantasy console verification strategy:
http://tasvideos.org/4450S.html
Maybe a way forward for disc based runs is to develop TASing techniques that are robust to unknown RNG like TheAxeMan did.
The real world is messy and in my mind it seems better to have a robust run rather then doing extensive console modifications to make something work.
Although, the technical difficulty of this is probably much higher.
Editor, Experienced Forum User, Published Author, Expert player
(3572)
Joined: 11/30/2014
Posts: 2744
Location: US
I just played back this run: http://tasvideos.org/1962M.html in bizhawk 1.12.1 without issue.
Make sure you are using NESHawk and not QuickNES. I can support NESHawk pretty extensively if you need help but there isnt much that can be done with QuickNES.
Editor, Experienced Forum User, Published Author, Expert player
(3572)
Joined: 11/30/2014
Posts: 2744
Location: US
natt wrote:
I've played around a bit with the sound emulation. Does it sound any better now (on an interim build)?
That is a definite improvement, great work natt!
The mice in Advanced D&D now sound much better, and the exploding tanks in Armor Battle sound much better too.
Enterring a dungeon in Advanced D&D still sounds a bit too high pitched for some reason, but it is better.
Editor, Experienced Forum User, Published Author, Expert player
(3572)
Joined: 11/30/2014
Posts: 2744
Location: US
TheWinslinator wrote:
Hey, I was trying to record an AVI of some of my intellivision TASing work thus far now that Bizhawk 1.12.1 has been released. Only one problem, the AVI files all play back in that squished 176x208 frame. Do you know what settings I would need to apply in order to get the normal 277x208 frame? Thanks!
File -> Configure and record AVI -> check 'resize video' and input the video size you want. There you go!
Editor, Experienced Forum User, Published Author, Expert player
(3572)
Joined: 11/30/2014
Posts: 2744
Location: US
What is the ACE set up for Battletoads? I was only aware of the game end glitch of loading incorrect objects.
And yeah as True says DPCM will not always lead to ACE. Some games don't even use the vulnerable read routine (i.e. Ninja Gaiden) and those that do it is only happenstance of coding that would lead to ACE, it's more likely you'd just get a crash due to the stack corruption.
Editor, Experienced Forum User, Published Author, Expert player
(3572)
Joined: 11/30/2014
Posts: 2744
Location: US
I get a lot of enjoyment from tech demos, so for me this is just another example of why we really need a demo tier.
As long as it takes an existing game and dooes nothing but provide inputs through the console, for me that's close enough to TAS to warrant a place on the site.
For me the biggest issue then is that it wouldn't be a console demo, but an emulator demo (still a really cool one) and at least for me that makes a lot of difference.
Editor, Experienced Forum User, Published Author, Expert player
(3572)
Joined: 11/30/2014
Posts: 2744
Location: US
Beefster wrote:
Oh... That's good to know.
Sorta leaves me with less of a project though. Or I guess it just calls for a change of direction.
I still wonder what kinds of other patterns can be gleaned from throwing data science at the system RAM. It might turn out to be fruitful for something unexpected.
There are a few things I can think of where having an automated appraoch with data science concepts could help out in analzying RAM. Here are couple I thought of ordered in what difficulty I expect them to be.
Easy: tracing RNG usage: as feos mentioned, finding RNG is not so difficult. Finding all the things it impacts though can be more time consuming. If you could automatically make some lists of what values RNG impacts (and even indirectly impacts) that might be really useful.
Medium: Finding object tables: Usually enemy/sprite/player data used by a game is clumped into tables in RAM, if we could automatically get lists of object slots, hit boxes, attributes, and such that could save a lot of time.
Hard: Improving Emulation: Currently one of the challenges in NES emulation is how initial console state impacts the game as it plays out. The dfferences are subtle and might only emerge after several minutes of playing, and might go undtected until something like a desync happens. This could really use some data sceincy type of stuff to find exactly where things go awry.
Extra Hard: Stack Corruption: If you can heuristically analyze what actions effect the stack, you might be able to force situations where the stack becomes corrupted (overflow, underflow, gets overwritten etc) This would be like an automated form of glitch hunting which might be feasible using RAM. Might get some really cool results if it works.
