libretro / mame2003-plus-libretro

Updated 2018 version of MAME (0.78) for libretro. with added game support plus many fixes and improvements
Other
189 stars 109 forks source link

Dragon Saber #1853

Closed Cibomatto2002 closed 1 month ago

Cibomatto2002 commented 1 month ago

On level 2 in the lava you can see something is not right I checked on the new MAME and I don't see this bug. here is a video.

https://www.youtube.com/shorts/1es69nZASJI

arcadez2003 commented 1 month ago

In order to fix this we'd need to update some of the Namco video code to MAME88 or so and alotta the games would take a rather large hit in the performance dept, which is a big no no for this core as per the lower spec devices MAME2003+ is aimed at.

0.87u3: Phil Stroffolino fixed older System2 ROZ system; in particular Dragon Blade glitches are no more.

I actually did this update for my Xbox MAME core maybe 10 years or so ago but had to roll back on it as the games dropped by 40fps down to 20fps, and im pretty sure the same would happen here.

Cibomatto2002 commented 1 month ago

The game needs to be set to imperfect graphics in the live compatibility table. Do I need to close this?

arcadez2003 commented 1 month ago

Yeah fair enough i can see the game has issues, as for closing wait and see if anyone else chimes in as to whether we should attempt this update or not.

grant2258 commented 1 month ago

Considering this core is aimed at low end in my perception a rpi3 is considered a baseline this update wouldnt be worth it if tanked so much on an xbox admittedly it would probably need tested to give a concrete answer but arcadez testing pretty much answers the question for me.

I seen some interesting debates on the system16 and some games for me are faster with the newer video code considering the way the old one draws the screen not entirely surprising but the old video code is still there as well as the new and only problematic games use the newer code when needed like the segaxbd we added lately. Perhaps we can limit the new video code at that one specific game only as a compromise.

I wouldn't recommend what was done with system32 updating it to the latest mame and take the hit. Both me and arcades in the past looked at this and fixing the issues lead to a tanking. Its nice to have systems working but there has to be a balance.

Whats done is done, I know today's low end runs mame mostly all 80s and 90s if you have a pi5 use mame current. If you have something older use plus. If your aiming at pi5 as low end you would probably be using a base something similar to advanced mame to get the most out the systems that require the extra beef. In short im with arcadez with this one.

I would like to hear mahoneyt944 thoughts on this update in a comparison with the system32 stuff pretty sure its less of a hit in the fps department on lower end. Maybe the bar is higher with lower end hardware improvements.

If that is the case it might be worth using a later mame base and keep this code for older hardware. This topic always muddles me at times,. I think its just sometimes people want a certain system to work. It probably worth adding the arm32 asm core for the v60 after these updates that in mame2000 in not really motivated to doing that myself though. It probably doesnt support the IO stall we need in its current state for some games.

arcadez2003 commented 1 month ago

Yeah it's not a straight forward update either as quite a few other namco drivers etc etc would be caught up in this so the workload is not on the lighter side however that has not always stopped us previously :)

However as i said already we'd be robbing peter to pay paul as to fix the gfx for Dragon Saber will in turn hit the performance for that game and quite a few others.

EDIT tagged the game as having imperfect gfx while we decide on whether the update is worth doin.

mahoneyt944 commented 1 month ago

You guys would have more knowledge on this system than I, so I don't have a strong opinion either way. I suppose if you think the hit isn't worth the reward, it would be be fine to leave it as imperfect or if it's wanted to try to add it to an isolated driver.

With system32 I felt the vast improvement was worth the hit and users had a higher demand for it. KMFDManic was able to squeeze some better performance out of the new code to boot and with the multi32 speedup hack I added we still try to balance accuracy and speed to some degree.

arcadez2003 commented 1 month ago

You guys would have more knowledge on this system than I, so I don't have a strong opinion either way. I suppose if you think the hit isn't worth the reward, it would be be fine to leave it as imperfect or if it's wanted to try to add it to an isolated driver.

With system32 I felt the vast improvement was worth the hit and users had a higher demand for it. KMFDManic was able to squeeze some better performance out of the new code to boot and with the multi32 speedup hack I added we still try to balance accuracy and speed to some degree.

Yeah i agree with ya with regards to Sega System 32 as for this update looking at the driver no other games barring Dragon Saber and Mach Breakers aka Numan Athletics 2 actually benefit from it so we'd be taking a performance hit for a fair few games to just fix two which is not a good ratio.

As for how big the hit would be we'd only know if we done the graft and i gotta admit im not bursting with enthusiasm to work on it as i'd need to split up all the mem maps as they merged em as part of the changes.

arcadez2003 commented 1 month ago

On a seperate note I noticed Final Lap 2 is tagged as not working maybe by adding the proper sprite roms from MAME79 would sort this.?? i've not much experience with these games as they played in slow motion on the xbox however i do remember Final Lap 2 booting up and it looked like it would be playable in MAME84

Hence looking at the commits between this core and that the most likely fix i thought was the sprite roms

grant2258 commented 1 month ago

Ill update the core and give it a shot and let you know how it goes.

grant2258 commented 1 month ago

The sound has a drone a little drone and the gfx are quite right but is shows signs of life. At least its booting now. Wasnt able to get it to accelerate in a straight line either probably need a calibration havent tested a lot just a boot check for some feedback.

Screenshot from 2024-09-05 21-05-28

arcadez2003 commented 1 month ago

Yeah looks ok apart from the road what about Final Lap and or Final Lap 3 does the road draw correctly for those ones.??

As for the update you'd likely have to start from the MAME79 changes and work forward from there but hey man just a warning the workload looks really on the highside.

grant2258 commented 1 month ago

finallap controls work but turning is weird. Finalap3 has road markings at least but has the same control issues as finalap2. I might look up the controls touching this video code would require too much testing with the amount of games involved for me.

arcadez2003 commented 1 month ago

With alotta the Namco Sytem 2 games you had to enter a service type mode to init the controls but our ole driver doesn't firstly have the code to access said input service modes nor the ways and means to save the settings afterwards im sure this was done via an eeprom but i'd have to check.

Yeah i agree it's too much work to update the video code as for Final Lap 2 missing the road i wonder if we need these...

0.97u3: Brian Troha added Final Lap road clut PROM to all Final Lap sets. 0.97u1: Phil Stroffolino fixed road colors in the Suzuka games. Added user3 PROM ($0 - 256 bytes).

Then again Final Lap 1 and 3 have the road drawing correct with our current code so maybe the reason why it doesn't display correctly for Final Lap 2 lies elsewhere.

arcadez2003 commented 1 month ago

Alright i hooked up the road colur clut proms for the Final Lap and Suzuka 8H games which hopefully will sort the road gfx niggles also added one for Four Trax, i'll also backport a fix for bad oversized sprites in the above from MAME234.

arcadez2003 commented 1 month ago

Apparently we should be using the Japan set of Marvel Land as it turns out the US version is a prototype with missing game elements and it ends on level 3-4 instead of 4-5, they swapped the sets around for current MAME i'll likely do the same for here sometime....

More info if interested :) https://tcrf.net/Marvel_Land_(Arcade)#Gameplay

grant2258 commented 1 month ago

The road looks as good as getting yer gnashers shined up at the dentist. I think ill look at the controls but there is another bad rom in finalap2 at least Ill need to check what cpu4 is actually doing and if its input related. Ive never even played marvel land to be honest. holding down 9 and pressing f2 lets you calibrate the game for finalap2 assume its the same for all games

arcadez2003 commented 1 month ago

Good to hear the road is sorted now as for that bad rom maybe there is a redump i'll need to check the WIP i know they switched to using another CPU type to handle the IO's as per CPU4 but they also changed the input style from dial to paddles which might improve the handling.

arcadez2003 commented 1 month ago

@grant2258 aye Marvel Land is not bad, but game diffs between versions have always interested me especially if extra levels or gore can be had, take Pacmania i like the music on the opening levels so the Japan version is better for me as it has an extra one not present in the world sets infact it has an extra level for each world so well worth keeping that set if you like that game.

Anyway i digress :)

arcadez2003 commented 1 month ago

Right then i think we can close this as the workload is too high for a large namco update to fix the gfx niggles in one game