SDL2

@Capehill understood. But just doing a plain make -f Makefile.amigaos4 will build which version? Ie, is there a flag to set for a specific version?

@davec555

It tries to build MiniGL + Ogles2 + Mesa support by default.

Check build config and disable unwanted: https://github.com/AmigaPorts/SDL/blob ... DL_config_amigaos4.h#L357

@Capehill I see you have this in the SDL_os4mesa.c file but don't see that define being used anywhere (and where you include the mesa includes). Can you tell me where that is? Just trying to get my head around it.

Thanks
Dave

Quote:

@davec555

Basically following

#ifdef USE_MESA_HEADERS
#include <GL/Mesa/gl.h>
#else
#include <GL/MiniGL/gl.h>
#endif

is what I have in <GL/gl.h> and other GL headers.

Similarly for <GLES2/gl2.h>:

#ifdef USE_MESA_HEADERS
#include <GLES2/Mesa/gl2.h>
#else
#include <GLES2/Ogles2/gl2.h>
#endif

@Capehill that explains it. Didn't even occur to me to actually look in the specific referenced includes themselves.
So in order for a thirdparty to compile it, they would need to create these custom headers, otherwise it just creates the minigl version only?

Thankyou!

Good morning to all,

I've a question regarding SDL2 and the compositing renderer.

Why while using, for example, the sdl2benchmark application, while using the compositing renderer the SDL all the tests related to:
- FillRects
- RenderCopy
- RenderCopyEx
- Color Modulation

the window content is completely black?

The corresponding tests with software renderer works as expected.

Infact, if I select "compositing" as renderer in the SDL2prefs, all the SDL2 window application stays black.

@Kirirur

If you use an emulator which doesn't support compositing, then compositing renderer doesn't work. You need PatchCompositeTags for that.

@Capehill

Hello, thank you for your hint. I'm currently using QEMU both on Linux MINT e macOS.

I'll try the patch you posted, even if the readme talks about WinUAE.

EDIT: just tried the patch on QEMU+Linux and after I launched it, not only the workbench crashed but also the X Server, I've to close it with Ctrl+Alt+Backspace.

I'm using https://github.com/derfsss/VirtIOGPU as my graphics card since it support 32 bit instead of 16 bit like sm501.

EDIT II: tested also with sm501: it does not crash, but the windows of the Workbench, for example, are missing the top border.

@Kirirur
Use the ATI driver for QEMU by BALATON Zoltan; “PatchCompositeTags” should work in 32-bit mode.
https://zero.eik.bme.hu/~balaton/qemu/ ... ti-vga-driver-latest.html
Don’t expect miracles.
The ATI driver by BALATON Zoltan alone will definitely work better for you than @derfsss’s VirtIOGPU.
VirtIOGPU uses a running flush process (this virtio-gpu implementation requires it—it’s not a bug by @derfsss) that consumes a lot of CPU on the guest side.
It would be possible to implement full compositing, but that’s completely pointless when the flush process itself on the AmigaOS4 side would consume 30–40% of the CPU.
Take a look here—you’ll find a benchmark of empty functions at https://hdrlab.org.nz/benchmark/gfxbench2d/os/amigaos/Result/3056, in the “Composite Score” section.
I don’t know how @Hans’s (A-EON) driver works. It might work with IGFX_COMPOSITETAGLIST_OFFSET set to 804 or 806, but if it operates on the same principle—with the flush process consuming CPU resources—then it’s definitely better to use BALATON Zoltan’s ATI driver for QEMU.

Hello @smarkusg thank you very much for your reply, I wasn't aware of another potential "chip" to be used in conjunction with AmigaOS 4.1 FE and QEMU 11.

I've downloaded it and modified my Kickstart.zip and QEMU startup batch file to test it.

The first impressions is that it doesn't works as well as @derfsss’s VirtIOGPU:
- its default resolution is at 24 bit and this result in solid black block at the edges of any window and a solid black block over the windows's title (both these artifacts are solved switching to a 32 bit resolution).
- moving the mouse around creates glitches on the underlying areas (the entire horizontal line from left to right) emphasized with the option: "guest_hwcursor=on" suggested in the README file, and still present but reduced without that option.

The test with PatchCompositeTags almost crashed my mac, everything was frozen for a couple of minutes, then I regain control and killed QEMU.

I think that for my purposes @derfsss’s VirtIOGPU and SDL2 "software" renderer are the best solutions, at least at the moment.

@Capehill
Btw, checking you readme update in repo, and just in case, Mesa also support OpenGL ES up to and include till 3.2. I even supply it with casual libGLESv2.a, despite a name it support all ES standards till latest today (same on other oses and linuxes, they have it named like that, while it up to 3.2).

@Kirirur
With the ati-vga Radeon.chip driver have you tried to go to go to ScreenMode prefs and select a 32bit mode? If not, why? Compositing generally needs 32bit modes as the support for it in 24bit is buggy in AmigaOS4 as you've found. This isn't a problem in the driver but in graphics.library or whatever should handle it on AmigaOS.

Hello @balaton of course I changed the display to 32 bit, I wrote it.

However if it may helps: I founded that the mouse artifacts on Linux does not appears, only on macOS.

I don't know if it may depends on QEMU version/implementation.

I've 11.1.1 on macOS and 11.0.2 on Linux Mint.

I stopped testing PatchCompositingTags and sticked to software renderer.

@kas1e

SDL2/3 currently does this:

1.3: Open MiniGL
ES 2.0: Open Ogles2
Else: Open Mesa

@Capehill
Nice, so if one want ogles2.x or more via mesa, he can just ask for ogles2.1 to have mesa's ogles ?