PS3 RetroArch Corner (cores for free)

[Crystal]

Hi.
Is there a way to apply these shaders to PS2 games, running on jailbroken PS3( CECH 2004 A.CFW 4.85; non backwards compatible)?
I mean using Retroarch or default PS2 emulator inside PS3.

Only retroarch cores... so if will be developed a working PS2 core on PS3 (very unlikely)..... yes.

 

[Crystal]

About Gearboy core, i compiled https://github.com/libretro/Gearboy thinking it is official repository, but it isn't.
It's old, version 2.6. Official Gearboy repository is https://github.com/drhelius/Gearboy (version 3.1.1)
I discover that beetwen v3.0.4 and v3.1.0 was introduced an issue on PS3 that makes Gearboy anusable.
Woking in progress.... https://github.com/drhelius/Gearboy/issues/124

 

[kdash2k19]

What firmware are you using on your PS3? In which version of retroarch did you put those cores?
Could you setup a RPCS3 emulator with the same firmware of your ps3 and install Retroarch?
To see and copy & pase log is easy on RPCS3...


Easier to investigate problems....

PS3 Slim 4.87 HEN 3.0.2 and Retroarch 1.9.0

 

[Crystal]

My PS3 is a FAT with CFW. Not using HEN, with 4.87 version, they work.
If you try to use all core posted in the first post, they work? Just to understand if what i compile work for you.

 

[kdash2k19]

My PS3 is a FAT with CFW. Not using HEN, with 4.87 version, they work.
If you try to use all core posted in the first post, they work? Just to understand if what i compile work for you.

All the cores works fine, the mGBA core loads but when starting load a game the console just freezes and i have to reset.

The Gearboy core works fine too.

 

[tps]

I ask for help and advice from those who have more experience than me.
@aldostools @bucanero @Zar @tps and all others (my memory is short)
About multipass sharders, Retroarch on PS3 uses some PSGL functions like glFramebufferTexture2DOES, glGenFramebuffersOES, glBindFramebufferOES, etc etc
Exactly in the source file Retroarch/gfx/drivers/gl.c there is the issue:

static bool gl2_create_fbo_targets(gl_t *gl, gl2_renderchain_data_t *chain)
{
   unsigned i;

   glBindTexture(GL_TEXTURE_2D, 0);
   gl2_gen_fb(chain->fbo_pass, chain->fbo);

   for (i = 0; i < (unsigned)chain->fbo_pass; i++)
   {
      GLenum status;
      gl2_bind_fb(chain->fbo[i]);
      gl2_fb_texture_2d(RARCH_GL_FRAMEBUFFER,
            RARCH_GL_COLOR_ATTACHMENT0, GL_TEXTURE_2D, chain->fbo_texture[i], 0);

      status = gl2_check_fb_status(RARCH_GL_FRAMEBUFFER);
      if (status != RARCH_GL_FRAMEBUFFER_COMPLETE)
         goto error;
   
   }

gl2_fb_texture_2d is glFramebufferTexture2DOES and after calling it, the check of status fail because gl2_check_fb_status (that is glCheckFramebufferStatusOES) return GL_FRAMEBUFFER_UNSUPPORTED_OES
I suppose that graphics driver is GCM and not PSGL.... so.... are all PSGL functions full operational?

Edit: Rectification, video driver is PSGL.

from what I've tried so far, PSGL works fine, but didn't use the ES part. it may be that some initialization/lib is missing for this. I don't have a lot of experience with it due to time constraints and had to reprioritise.
try looking at some examples if you have any.

 

[Crystal]

from what I've tried so far, PSGL works fine, but didn't use the ES part. it may be that some initialization/lib is missing for this. I don't have a lot of experience with it due to time constraints and had to reprioritise.
try looking at some examples if you have any.

Already fixed ^_^

 

[sandungas]

Already fixed ^_^

You mean that it was fixed some weeks ago, or that you fixed it today ?

And btw, PSGL can be used to compile the "XMB clone" interface of retroarch under psl1ght ?

 

[Crystal]

You mean that it was fixed some weeks ago, or that you fixed it today ?

I fixed it and here is my message:

"LOL.... at end it was easy .... in gl2_create_fbo_texture just change GL_RGBA with GL_ARGB_SCE when glTexImage2D is called"

 

[Crystal]

And btw, PSGL can be used to compile the "XMB clone" interface of retroarch under psl1ght ?

No, PSGL is a library of old sdk sony.

 

[sandungas]

I fixed it and here is my message:

Now i remember, i was reading about it in your github comments the other day, the damned suffix "_SCE"
Nice finding btw

No, PSGL is a library of old sdk sony.

Ok, not sure what i was thinking when i wrote that, yeah the prefix "ps" means it have some additions from sony, so it belongs to the official sdk

But which graphic library is used in the retroarch "XMB clone" interface ?, the standard openGL ES ?
I remember when retroarch team decided to port the code to psl1ght one of the things missing was the "XMB clone" interface, and i guess is because there is no openGL in psl1ght, right ?

 

[Crystal]

But which graphic library is used in the retroarch "XMB clone" interface ?, the standard openGL ES ?
I remember when retroarch team decided to port the code to psl1ght one of the things missing was the "XMB clone" interface, and i guess is because there is no openGL in psl1ght, right ?

GCM .... a basic interface to RSX ... i think without shaders managment and others things.....
and yes... no official (or good) OpenGL for psl1ght

 

[ZeroOne]

Loading and saving input remaps are broken on 1.9 cores

PSL1GHT issue?

The issue I mentioned earlier in the thread is unrelated but something I noticed using previous official sdk releases 1.8.4 etc re: default directory structure:

Official: /retroarch or /SSNE10000/USRDIR contains retroarch.cfg and folders e.g /cores /system /saves /remaps etc

(PS3): /SSNE10000/USRDIR/cores contains everything from /USRDIR e.g retroarch.cfg cores and folders e.g /system /saves /remaps etc

The retroarch config folder has been changed to /cores in the source and default folders are recreated every boot even after changing in retroarch.cfg or GUI.

 

[tps]

Already fixed ^_^

indeed, I noticed the rest of the thread after I responded.
where can I see the code in Retroarch? I would need some inspiration.
cheers

nevermind, I saw your references to cores above and found it

 

[Crystal]

PSL1GHT issue?

The issue I mentioned earlier in the thread is unrelated but something I noticed using previous official sdk releases 1.8.4 etc re: default directory structure:

Official releases: main retroarch folder /retroarch containing retroarch.cfg and folders e.g /cores /saves /remaps etc or

PS3 release: main retroarch folder /SSNE10000/USRDIR containing retroarch.cfg and folders e.g /cores /saves /remaps etc

Recent releases: main retroarch folder /SSNE10000/USRDIR/cores containing cores and everything from /USRDIR e.g retroarch.cfg folders e.g /saves /remaps etc

The directories have been moved in the source and they're recreated every boot even if the defaults are changed in retroarch.cfg or GUI.

I did not know that issue and at the end if all works correctly, the folders structure could be egual to official release or not.

 

[Crystal]

Hi, your mGBA core is broken in my ps3 slim. Cannonball multipass is broken too.

I offer a barrel full of beer too! :P

EBOOT.BIN update could solve some problems...
LINK

 

[sandungas]

Now that you are talking about paths... im the only who thinks the path name "dev_hdd0/game/SSNE10000/USRDIR" is a bit... weird ?
I dont even know what means SSNE but is not intuitive (it makes me remember Super Nintendo Entertainment System)

Is specially weird because the word RETROARCH fits perfect with the PS3 directory name lenght requirements (4+5=9)
I dont get why they didnt used "dev_hdd0/game/RETROARCH/USRDIR"
In my oppinion it looks perfect, and they are lucky the lenght of the name fits so well

 

[Crystal]

Why not.... PSXPLRA01 or PSXRETARC etc etc... at the end we are making a custom RetroArch for PS3
Anyway i dont know if app folder names must have a number at the end for some reason.....

 

[Coro]

My PS3 is a FAT with CFW. Not using HEN, with 4.87 version, they work.
If you try to use all core posted in the first post, they work? Just to understand if what i compile work for you.

AFAIK anything for hen must be signed with 3.55 keys.

 

[sandungas]

I think "dev_hdd0/game/RETROARCH/USRDIR" is more straightforward, the problem i see with the actual SSNE10000 name is that is not intuitive, and my suggestion to rename it to RETROARCH is to solve the problem by using the most intuitive name posible
Also, is a bit like the default name... incase we could make a giant poll with all retroarch users i bet RETROARCH was going to be the winner, it was going to have some votes even from people that doesnt cares about it
Im not telling it only as a personal preference, i think most users are going to be fine with the name change of that directory

The official standard is to use 4 letters + 5 numbers
But there are many homebrews that doesnt follows that rule... in the practise, for picky people like me that would like to see the 100% of homebrews following the standards thats imposible to achieve, because we are not modifying all that names from all the homebrews we use, and eventually we are going to download some homebrew that does it, and well... never minds, we get used to break that official standard of 4+5
As far i know there is not any problem in using 9 letters