Editor, Experienced Forum User, Published Author, Expert player
(3536)
Joined: 11/30/2014
Posts: 2733
Location: US
You can select and run SubNESHawk in 2.3.2 and this run works without issue (I just tested.)
It will not work in later builds because of an MMC3 IRQ timing bug that I fixed.
The run is console verified, so the reason it doesn't work after a bug fix is that there are other factors that make NES runs verify on console. In this case, I can change the initial offset to 0 from 2 and the run will work again without changing any inputs.
2 is used because that is what a console verified Battletoads run uses which has been the standard for timing regression testing for that past few years, if I change it to zero, this run will break. This is a pretty annoying fact of life when working with NES, not everything will work out the same every time you turn it on.
I could change it to 0 to favor a published run, but I'm not sure what other runs it would break.
Editor, Experienced Forum User, Published Author, Expert player
(3536)
Joined: 11/30/2014
Posts: 2733
Location: US
Gambatte's 'equal length' frames option doesn't actually make every frame equal length. So, unfortunately there isn't a way to get a one-to-one comparison with GBHawk just by looking at frames, some adjustments have to be made.
Editor, Experienced Forum User, Published Author, Expert player
(3536)
Joined: 11/30/2014
Posts: 2733
Location: US
Thanks to lobsterzelda for finishing Battletoads, now only 6 runs remain (all SNES now.)
It's also interesting to note that only 90 of the first 1000 publications are still active, although i would suspect that of these there might actually be some that aren't improvable (Umihara Kawase and monopoly)
Editor, Experienced Forum User, Published Author, Expert player
(3536)
Joined: 11/30/2014
Posts: 2733
Location: US
I didn't mean to imply that I was actually going to do it, that's too tedious for me, just that it would be a necessary step. Feel free to give it a try, no need for me to be a co-author or anything.
Editor, Experienced Forum User, Published Author, Expert player
(3536)
Joined: 11/30/2014
Posts: 2733
Location: US
ThunderAxe31 wrote:
CasualPokePlayer wrote:
ThunderAxe31 wrote:
I'd also like to try if changing GBC color multiple times can allow more RNG possibilities, or delaying the GBC BIOS sequence by changing color later.
Yes and no, mostly has no effect, and when you actually get an effect, it's more because the bios is being delayed by some amount of frames (which eventually eats into the delays).
I didn't want to listen to that, so I've tried anyway. :) And, guess what? My bot was able to find one faster route: 364 CPU cycles saved over the current submission. Byte D35A is still 0xC2.
Alyosha, do you agree to replace the submission file and add me to the author list for this submission?
If we are going to go by cycle count then this is trivially sub-optimal since you can use SubGBHawk to reset exactly when needed. There are many, many ways to get the correct delay and I certainly didn't check cycle count for them, so I'd rather just cancel this for now.
Editor, Experienced Forum User, Published Author, Expert player
(3536)
Joined: 11/30/2014
Posts: 2733
Location: US
DJ Incendration wrote:
What do you mean by 0,0,5,5? Also, I can't seem to manipulate 0X64D0 or EA or anything like that in Yellow. I have tried many different frame delays, but I could only do the delays in Pokeguy's any% world record, with the soft resets.
I don't know how to do yellow.
There are 4 places to delay input in order to effect trainer ID. The 0,0,5,5 is just my shorthand for how many frames is delayed at each point. So no delay for the first two inputs, then 5 frame delay at third input and 5 frame delay at fourth input. Changing the palette is not included in this shorthand.
I'm not concerned by having an invalid save file, the game was completed successfully, which is the goal.
Editor, Experienced Forum User, Published Author, Expert player
(3536)
Joined: 11/30/2014
Posts: 2733
Location: US
ThunderAxe31 wrote:
I just noticed that the current movie manipulated the Trainer ID to be 0x64BD, while instead you got 0x64C3. Does the second byte make no difference, or did you have to adjust the ACE accordingly?
Editor, Experienced Forum User, Published Author, Expert player
(3536)
Joined: 11/30/2014
Posts: 2733
Location: US
ThunderAxe31 wrote:
Nice job, but did you take in consideration the fact that there is a faster SRAM glitch for the Japanese version? User movie #35593301208919904
Though I'm not sure if that movie should be considered for comparison, since it fails to trigger the final routine after showing "THE END" screen at the end of the credits roll, as pressing A doesn't send you back to the intro and it doesn't save the progress to the save data. More explanations in this post.
I'd also like to try if changing GBC color multiple times can allow more RNG possibilities, or delaying the GBC BIOS sequence by changing color later.
I did do some limited testing with delaying BIOS end, not that extensive since I just ended up with 10 total again (2 BIOS delay + 8 regular delay.)
Here is the script I used to test stuff:
http://tasvideos.org/userfiles/info/63900696766348833
You can fiddle with the palette however you want, let the game run to frame 500, make a state in slot 1, then run the script. It will brute force check the remaining 4 delay points from there. (If you are going to delay the BIOS you need to add that into the script though.) Even with Gambatte this takes a long time to run.
I'm not going to look too deeply into a movie that doesn't properly call the ending.
Editor, Experienced Forum User, Published Author, Expert player
(3536)
Joined: 11/30/2014
Posts: 2733
Location: US
natt wrote:
I worry that the tracelogger's notion of timing might be busted after a while; doesn't gambatte's cyclecounter variable occasionally reset backwards to account for 2^32 rollover?
I thought this was fixed, but I checked just now and it is not. Not sure how difficult it would be to make it 64 bit instead of 32.
Editor, Experienced Forum User, Published Author, Expert player
(3536)
Joined: 11/30/2014
Posts: 2733
Location: US
Trace Logger gives the same cycle counts for both cores. Looks like gambatte's count is being cut in half for the lua function somehow, I'll look into it.
EDIT: Yeah, Gambatte is using master clock rate (~4MHz) / 2 for TotalExecutedCycles. This comes from the audio code it seems like? Not sure yet how to get the count it uses for Trace Logger.
Editor, Experienced Forum User, Published Author, Expert player
(3536)
Joined: 11/30/2014
Posts: 2733
Location: US
Second pass through the game. Little over 10 seconds saved.
http://tasvideos.org/userfiles/info/63588782110175948
I think worlds 1 and 2 are pretty much done now, but the later worlds still have some work to do.
I thought about converting it to GBC mode, but when I do that there are a lot of sprite priority glitches, so not sure what the final product will be in yet.
Editor, Experienced Forum User, Published Author, Expert player
(3536)
Joined: 11/30/2014
Posts: 2733
Location: US
Thanks for the continued testing.
Looking at the schematic, even going through the headphone jack adds some more filtering, so there are probably lots of variations in what you can hear. I've always had a weak ear for audio issues, and all the variability just makes it more annoying.
Editor, Experienced Forum User, Published Author, Expert player
(3536)
Joined: 11/30/2014
Posts: 2733
Location: US
I made one more adjustment that 'fixes' ready to rumble and cannon fodder (sounds closer to videos I watch.) I don't think it changes anything here.
I tried adding a high pass filter, but it didn't sound any different and slowed things down a lot, so I didn't commit it.
So I think I'm done fiddling with audio for now unless something sounds very obviously wrong.
Editor, Experienced Forum User, Published Author, Expert player
(3536)
Joined: 11/30/2014
Posts: 2733
Location: US
Thanks for testing. Sounds pretty clear on GBA.
I looked into the documentation on GBDev, and apparently sound on GBA is produced differently then on GBC, but this particular case shouldn't be effected by it. The popping in Crystal here comes from turning off channel one in the mixer while it is still active. This should definitely produce a pop on all systems, due to DC bias in the DAC. On a real system, the pop is probably attenuated by filtering, but I don't have any filtering so it comes out pretty sharp.
I did tune the bias in master though so the popping is much less noticeable, hopefully it sounds better now.
(Side note, following the spec here broke sound on two games, ready to rumble and cannon fodder. This is because they turn the wave channel off to update the wave channel constantly. This produces a very noticeable humming. On Youtube I saw a video also mentioning it for ready to rumble, so maybe it's accurate? I believe on GBA these issues wouldn't show up at all because the DACs there are always treated as on. I have not updated audio for GBA mode though, so it will have to stay broken for now.)
Editor, Experienced Forum User, Published Author, Expert player
(3536)
Joined: 11/30/2014
Posts: 2733
Location: US
I compared the audio against Sameboy for this. The pops appear in all the same places, they are just louder in GBHawk. I'd be interested to hear what it sounds like on console. Also Gambatte definitely has overall less popping, not entirely sure why.
