NEW Reicast Cores in RA

[kelv]

Hello

Checking Retroarch page I saw that threre are new cores for Dreamcast https://www.libretro.com/index.php/introducing-reicast-oit-libretro-core-updated-reicast-regular-core/ It would be great if it can be in the next release.

THX!!!

[Gaetan]

The changelog is really promising !

[barbudreadmon]

@kelv said in NEW Reicast Cores in RA:

It would be great if it can be in the next release

Only usable on a 3Ghz x86_64 computer with proprietary nvidia/radeon GPU and drivers at the moment.
Just so you know, to get this running on a rpi we would need all of this :

• a more powerful rpi (i would say 2Ghz at the very least, and i don't expect the future rpi4 to have this kind of cpu)
• a new gpu on the raspberry compatible with opengl 4.3 (considering what the rpi fundation told in the past, i don't think it'll ever happen)
• working multi-thread support in libretro-reicast (or a 3Ghz cpu in the rpi, which is very unlikely)
• working arm dynarec in libretro-reicast (or a 4Ghz cpu in the rpi, which is very very unlikely)

Also, the last 2 issues have been known since the beginning of libretro-reicast (almost 3 years ?), it seems unlikely they'll get fixed any time soon.

[Gaetan]

@barbudreadmon Thanx for your enlightement. So, sadly nothing is coming for the Recalbox users before 2025 ^^

[Substring]

@barbudreadmon that's something i just don't understand with libretro ports ... Why do they diverge that much from upstream ? Why not work with mains devs for an "easy to keep up-to-date" libretro core as they did with the ppsspp core ?

Reicast does have armv7 dynarec ...

[barbudreadmon]

@substring said in NEW Reicast Cores in RA:

Why do they diverge that much from upstream ?

Well, that's definitely not on purpose, from what i can tell, arm/x86/x86_64 dynarecs in standalone reicast are really different, i guess the x86_64 one was easy to get working in the libretro port, while the 2 others weren't. Actually, i looked at recent commits and it seems twinaphex is still trying to fix the arm one.

[barbudreadmon]

@substring You can find the "standalone version" of reicast-oit there : https://github.com/flyinghead/reicast-emulator
This one got the dynarec and multithread support, but you still need opengl 4.3 (perhaps it also works with gles3).

[Substring]

@barbudreadmon yup, found the repo while browding through TA's commits on reicast OIT. Will give it a try after summer unless someone else in the tea wants to have fun with it

[Flyinghead]

Hi @substring and others.
The master branch of my reicast fork should work fine on a Pi. I tested it and made sure to keep performance to an acceptable level. It adds a number of features compared to upstream reicast: shadow volumes, clipping, render to texture, multipass rendering support and a number of bug fixes.
The OIT branch (fh/deferred-shading) on the other hand is very experimental and require desktop-level cpu and gpu (and OpenGL 4.3). It also only works with NVIDIA gpu at the moment and has a couple of bugs, but I intend to improve this in the coming weeks.

[Substring]

@flyinghead Hey ! Welcome on board, glad to see you here

how does your reicast fork work compared to skmp's in terms of performance ? like ... does the date finally get saved ? We've also spotted a bug of improper exit of reicast with some dirty die here and there, not cleanly deinit-ing SDL2 which can cause some problems on some boards using GBM rather than FBDEV.

[Flyinghead]

@substring said in NEW Reicast Cores in RA:

@flyinghead Hey ! Welcome on board, glad to see you here

Thank you!

how does your reicast fork work compared to skmp's in terms of performance ? like ... does the date finally get saved ? We've also spotted a bug of improper exit of reicast with some dirty die here and there, not cleanly deinit-ing SDL2 which can cause some problems on some boards using GBM rather than FBDEV.

Performance should be similar to skmp reicast although most features I added do have a slight performance hit.
I also integrated fixes for clean exit, so that the date and VMU are properly saved on exit.

[Gaetan]

@flyinghead That's impressive work !
Does Shenmue 2 run on pi3 now ? It's stop after the end of the introduction.
It's great to see that there is a hope of seeing improvement with DC emulator on pi

[Substring]

@flyinghead I'll definitely give it a go (in august)

[Flyinghead]

@gaetan I believe the problem with Shenmie 2 is that it has self-modifying code, which breaks the recompiler. I haven’t made any change is this area unfortunately.

[Gaetan]

@flyinghead Ok. Thank you for your work !
DC is a great system and it's nice to see people involved in such a project

[Gaetan]

Et une nouvelle màj Reicast 8.1 :
https://github.com/reicast/reicast-emulator/releases/tag/r8.1

[Gaetan]

@flyinghead est-ce que tu sais si sur ton fork les jeux Worms fonctionnent ?
Avec la version actuelle il y a un retour à la gamelist systématique quelque soit le format de la rom.

[Flyinghead]

@gaetan Malheureusement, non. Les jeux Worms utilisent Windows CE, qui n'est pas supporté pour le moment.

[Gaetan]

@Flyinghead ok, j'avais beaucoup joué à cette version du jeu à l'époque

@substring said in NEW Reicast Cores in RA:

@flyinghead I'll definitely give it a go (in august)

Je sais que tu étais très occupé ces derniers temps, mais as-tu pu depuis y jeter un œil ?