holy sh!t.

I said that a couple of times reading this in public transport. holy sh!t.



long live kas1e

Much appreciated great news !!!

Absolutely amazing.
Is there a sdl2/sdl3 context for it?

WOW!

Wow!! Well done kas1e !!

Good to see news like this popup.
Keep up the good work.

AHHHHHHHHHHHHH!! YOU ARE THE MAN! WELL DONE!

This is what AmigaKit should have done from the start instead of reinventing the wheel with a proprietary API, which spawned a 100 different wrappers and workarounds to have OpenGL working. Props for releasing it on OS4Depot instead of making the users jump through hoops.

Does it also support compute shaders?

Thanks for all your hard work in bringing quality software to OS4!

When I try to run the glmark2 test program, I get the following output:
Quote:

(I've got a Radeon RX Polaris12 board in my SAM460.)

Way to cool!

You are amazing! Kudos!!!!!! We all own you a Big Pizza and also a Fiorentina with Chianti (Do you plan to come and visit Italy??)

@Everyone
Thanks!

@BSZili
Probably Mattew were in wish to go uniq way.. but really, it was just Hans alone for almost everything, in compare with thousand of amd/mesa/etc developers.. Strange that it was at least in the form as it was..

@Capehill
Yes, all works. For now porting: https://github.com/pablode/flut , it works, through slow like hell. I need to dig it what wrong with, is it bad code, or something to improve in mesa port. But at least all compute shaders compiles, renders, etc.

@PEB
At least first bug report ! Thanks ! Can you plz say me that:

1). is it only glmark, or those 7 shader test cases fail too ?
2). what screenmode and depth you use
3). what version of OS, kernel and graphics.library
4). what version of radeonrx driver

Also if you can, run plz from c:radeon_gpu_info binary and post output there too.

ps. if you can write me at kas1e@yandex.ru, we can work faster then to fix it.

Thanks!

Of course, I'm not blaming Hans. It wasn't his call and he did his best as a sole developer. If AmigaKit wasn't suffering from the "Not invented here" syndrome he could have focused more of his energy on improving the drivers.

That's pretty cool; I hope I can use it to do some crazy stuff.

Thanks Roman !

@kas1e This is amazing news! Thank you!

What a great surprise, thank you Kas1e 👍

Amazing work. :-)

Can/should we use gl helpers like glad? If so, how would we use them? Thank you and great work!

kas1e figured out what the problem was. (I was running a 16bit screen, and mesa is currently hard-coded to run only in 32bit.)

He plans on adding 16bit support in the next release, but in the meantime, if you run into the same problem, that's the explanation.

thank you for your fantastic work....go on..

Amazing work! Thanks for that!!

I didn't understood the part about the non-GART fallback. Will the SAM460 benefit as well?

@skygecko
All public SAM460 AEON's drivers be it RadeonHD or RadeonRX do not have GART support at all. The same as older RadeonHD 3.7 drivers didn't GART support anywhere too. So for those situations when no GART available by any of reasons, i fallback to "no GART" ways, which slower. Usable, but still slower.

That for example reason why some 3d stuff done for gl4es / ogles2.library / warp3dnova.library slower than sam460, and there we can't do anything about while drivers for SAM460 will not have GART support.

I do have in TODO to patch current public RadeonRX drivers for same and add GART support there (as i know what to do, because Max from Acube share with me how to setup GART on sam460) , so can release a patch which users can run to patch their RadeonRX drivers for SAM, but that not today/tomorrow :)

Wow! Great work!

Wow, Roman, great job! This takes the 3D performance on our system to a whole new level.

@Goos
Not the latest motivation factor was your games which always were in needs to hard patch to avoid Nova's issues, like broken shader compiler , missing switches and co :) But mesa port for sure have where to growup and need stability fixes anyway. Hope you with Thomas can help to catch more bugs :)

@kas1e

Congrats on getting this fully working. Having full MESA is great.

@BSzili

To be fair to A-EON, back when I started the AMDGPU shader compiler was based on the monster that is LLVM. That's why I ended up writing my own compiler instead of using theirs. More recently they switched to a new compiler that's based on mesa3d's own compiler infrastructure. This makes porting it much more possible. Switching the RadeonHD/RX Warp3D Nova drivers from the DIY compiler to AMD's new mesa one was on my to-do list.

Not needed any more, though.

Hans

@kas1e

Great achievement :)

HunoPPC's ports (especially Doom3) will benefit from your MESA port ? Did you have any contact with him ?

Thanks.

AWESOME !

Amazing. kas1e for president! And congrats!

@Sinan

> HunoPPC's ports (especially Doom3) will benefit from your MESA port ? Did you have any contact with him ?

I think Huno no needs any contacts for as he can use mesa with no probs, and maybe if he will rebuild it over mesa and added his patchs on top of it, it may be a little bit better, but, with doom3 we can't expect miracles because limitation is our hardware. At least, if we see there:

https://forum.hyperion-entertainment.c ... topic.php?t=4645&start=10

25fps is maximum on origina mesa over PPC linux on x5000/040

But if we see old original PC benchmarks witch machines better than ours and which not big-endian ones (so no needs for conversion), then:

https://hardforum.com/threads/official ... -benchmark-thread.786540/

So results hardly reach 60 fps for them too..

Anyway, i ported doom3 to mesa too, and in UltaMaximum quality with everything enabled (shadows, lights, whatever), it gives just ~25fps in timedemo. Maybe not that bad as well, but not 60 :) See:

Amazing news, great work!

@kas1e:
...I do have in TODO to patch current public RadeonRX drivers for same and add GART support there (as i know what to do, because ...

Ohhh yesssss!!!! Sounds very very good man!!!!!