GroovyArcade live-CD 2020

[galaxian77]

I've put together a video tutorial, on how to download, write to flash drive, and install GroovyArcade, for your own arcade machine !

[Substring]

Been digging in the past for xf86-video-* patches, looks like ati and nouveau are concerned. Fine. But shouldn't amdgpu be patched too ?

[Doozer]

Been digging in the past for xf86-video-* patches, looks like ati and nouveau are concerned. Fine. But shouldn't amdgpu be patched too ?

Which kind of patches are you talking about?

No need to patch them to enable 15kHz output.

[Substring]

As the googlecode repo is empty, i'm havibg a hard time to find what's been patched. Found https://github.com/tob1k/groovyarcade.diff?files=1 for example. One of the ATI patchs allows resolutions down to 160x100 instead of 320x200. Looks like there is some dpms stuff too.

The nouveau drivers looks unpatched

[Calamity]

There was also a patch for sdl that allowed new video modes to be available once the sdl instance is initialized.

[Substring]

for SDL1, which is totally outdated now

[Doozer]

SDL 2.x does not need to be patched, unless someone figured out how to implement frequency refresh parameter with resolution switching function. This would eliminate the need of the DRM code from GM side. 

Xf86 does not need to be patched anymore. Stock version is fine as such. Just about the resolution limit, which game would be suitable to test 160x100 resolution? I want to test this to confirm resolution limit.

[Substring]

SDL 2.x does not need to be patched, unless someone figured out how to implement frequency refresh parameter with resolution switching function. This would eliminate the need of the DRM code from GM side. 

do you mean this ? Not sure if it's for fullscreen or windowed mode though

Just about the resolution limit, which game would be suitable to test 160x100 resolution?

A few examples from mame0212.xml :

aim65 160x200
cybikov1 160x100
cybikov2 160x100
fp200 160x64
gamate 160x150

Just browed the first 25% of the .dat looking for width="160". Would need XML parsing for better results. Don't know either if they are part of arcade or not.

[Calamity]

SDL 2.x does not need to be patched, unless someone figured out how to implement frequency refresh parameter with resolution switching function. This would eliminate the need of the DRM code from GM side. 

Xf86 does not need to be patched anymore. Stock version is fine as such. Just about the resolution limit, which game would be suitable to test 160x100 resolution? I want to test this to confirm resolution limit.

As far as I know the only DRM code we have in GM is the drmWaitVBlank call.

Yes, I forgot the SDL patch was for 1.x. Back then we could dynamically switch modes after a fullscreen window was created, the issue was SDL just polled available modes on startup, so adding new modes didn't work, we needed to force SDL refresh its internal mode list.

IIRC, in SDL 2.0, the fullscreen mode needs to be defined when creating the SDL instance, and can only be changed afterwards by resetting the video subsystem. And even so, it's usually problematic.

[Doozer]

In SDL 1.X the function was named SDL_SetVideoMode. Unfortunately, the function does not have a refresh parameter. The function was picking the best available resolution but not honoring the refresh rate. On top, when the system was exposing different refresh rates for the same resolution via DRM. SDL was not switching at all if already in the same Width X Height. But 1.x is not used anymore.

The v2 API removed the prototype and introduced SDL_CreateWindow with the same drawback. To go full screen, flag SDL_WINDOW_FULLSCREEN/SDL_WINDOW_BORDERLESS can be used. I do not know a method to force a refresh rate in SDL2.X.

[Doozer]

SDL 2.x does not need to be patched, unless someone figured out how to implement frequency refresh parameter with resolution switching function. This would eliminate the need of the DRM code from GM side. 

Xf86 does not need to be patched anymore. Stock version is fine as such. Just about the resolution limit, which game would be suitable to test 160x100 resolution? I want to test this to confirm resolution limit.

As far as I know the only DRM code we have in GM is the drmWaitVBlank call.

Yes, I forgot the SDL patch was for 1.x. Back then we could dynamically switch modes after a fullscreen window was created, the issue was SDL just polled available modes on startup, so adding new modes didn't work, we needed to force SDL refresh its internal mode list.

IIRC, in SDL 2.0, the fullscreen mode needs to be defined when creating the SDL instance, and can only be changed afterwards by resetting the video subsystem. And even so, it's usually problematic.

That's right. We use drawolg to get access to DRM to push a new resolution with correct refresh and do the switching. Things might change in case MAME decide to switch from OGL to Vulkan API. But not so much.

[Calamity]

In SDL 1.X the function was named SDL_SetVideoMode. Unfortunately, the function does not have a refresh parameter. The function was picking the best available resolution but not honoring the refresh rate. On top, when the system was exposing different refresh rates for the same resolution via DRM. SDL was not switching at all if already in the same Width X Height. But 1.x is not used anymore.

The v2 API removed the prototype and introduced SDL_CreateWindow with the same drawback. To go full screen, flag SDL_WINDOW_FULLSCREEN/SDL_WINDOW_BORDERLESS can be used. I do not know a method to force a refresh rate in SDL2.X.

Aaah, yes, I forgot that sh*it 

And the strategy we used, IIRC is to only make available one resolution with the desired refresh rate so SDL can't fail.

Same crap with all libraries these days. Coders don't understand what refresh rates are, they heard it's something about screen flicker. They trim that redundant parameter from their APIs making everything a mess.

I believe it's the same issue with Attract Mode switching to odd modes, they use a video library that only takes width x height.

You can bet anything that's multiplatform is broken in this regard.

Other emulators changing to random modes suffer from this design flaw (but it's always CRT Emudriver fault).

[Substring]

Considering wayland seems to forbid users to change resolution (no official xrandr alike tool and probably lib), future looks dark

[Doozer]

Considering wayland seems to forbid users to change resolution (no official xrandr alike tool and probably lib), future looks dark

No big worries on that side. In the GM code, setting resolution/refresh rate is done by using libdrm calls. To summarize, the manipulation is performed at kernel level. I am confident I can sort out the quirks when Wayland will be deployed widely.

[Calamity]

Considering wayland seems to forbid users to change resolution (no official xrandr alike tool and probably lib), future looks dark

No big worries on that side. In the GM code, setting resolution/refresh rate is done by using libdrm calls. To summarize, the manipulation is performed at kernel level. I am confident I can sort out the quirks when Wayland will be deployed widely.

In my view, the most important trouble we're facing is buffering caused by desktop compositers. In Linux, "fullscreen" is equivalent to Windows' borderless mode, so even if you try to handle vsync manually, you can't bypass desktop compositing and its buffering.

I've measured GM with the amdgpu driver and it lags 1 frame. I'm quite sure it's due to this.

[Calamity]

Also, I'm seeing we're using XRRSetCrtcConfig in order to switch modes, which indeed is from the libxrandr library.

[Doozer]

A few examples from mame0212.xml :

aim65 160x200
cybikov1 160x100
cybikov2 160x100
fp200 160x64
gamate 160x150

About the lower resolution, you are right on the 320x200 limit. But I don't think it is an issue thanks to Calamity integer scaling implementation (only talking about MAME here)

Integer scaling of 5 will fit perfectly with 800 x 500 resolution for those games and a x2 factor makes if fine at 320x200. I did manage to go lower the height limit (below 200 lines) but not to the exact wanted resolution.

One more test would be to force EDID 160x100 at boot (which should work) then X will use this resolution as default for the session. But I think this is getting tricky and involve unnecessary tweaking. I can imagine impracticability (look and feel) using a launcher/front-end at this resolution.

Personally, with the help of integer scaling, I do not think patching the xf86 driver is a necessity.

[Calamity]

The height is the only critical value. As long as we can get 240p we're fine. Lower (e.g. 224) can be done by adding borders.

[Doozer]

Considering wayland seems to forbid users to change resolution (no official xrandr alike tool and probably lib), future looks dark

No big worries on that side. In the GM code, setting resolution/refresh rate is done by using libdrm calls. To summarize, the manipulation is performed at kernel level. I am confident I can sort out the quirks when Wayland will be deployed widely.

In my view, the most important trouble we're facing is buffering caused by desktop compositers. In Linux, "fullscreen" is equivalent to Windows' borderless mode, so even if you try to handle vsync manually, you can't bypass desktop compositing and its buffering.

I've measured GM with the amdgpu driver and it lags 1 frame. I'm quite sure it's due to this.

The allocation of the frame buffer is performed by the RADEON driver (X side) and have only a single frame buffer allocated. Do not forget the GL layer and possibly another buffer allocated by the library.

[Doozer]

It would be nice to test frame lag with the following in the xorg.conf

Code: [Select]

Section "Extensions"
    Option         "Composite" "Disable"
EndSection


[Calamity]

The allocation of the frame buffer is performed by the RADEON driver (X side) and have only a single frame buffer allocated. Do not forget the GL layer and possibly another buffer allocated by the library.

I'm talking about the amdgpu driver now, not the (old) radeon driver. I've been testing this one lately for the Ryzen 2400G I got.

My previous tests showed the radeon driver finally had no lag, but now the lag is back with amdgpu 

[Calamity]

It would be nice to test frame lag with the following in the xorg.conf

Code: [Select]

Section "Extensions"
    Option         "Composite" "Disable"
EndSection


Is it enabled by default? Sounds interesting, I need to test it when I'm back home.

[Doozer]

It would be nice to test frame lag with the following in the xorg.conf

Code: [Select]

Section "Extensions"
    Option         "Composite" "Disable"
EndSection


Is it enabled by default? Sounds interesting, I need to test it when I'm back home.

It is at least taken into account, check inside the Xorg.0.log file:

Code: [Select]

[    12.536] (**) Extension "Composite" is disabled

Yes, it is enabled by default. Here is the log with vanilla xorg.conf.

Code: [Select]

[    15.043] (II) Initializing extension COMPOSITE


[Substring]

So :

- should we patch xorg video drivers to allow resolutions lower than 320x200 ? 160x100 was not a good case at all as 320x200 is a multiple of it. Should we simply rely on the fact that GM can upscale anything lower than 320x200 and fit it in interlaced 640x480 with borders eventually ?
- regarding the Composite xorg.conf value, can be disabled I guess. The only case I can see where GA would need it is when booting to LXDE, which seem not to handle composition by default. Worth the try !

[Rolyretro]

Hi there!

I’ve just completed a new GroovyArcade install on a PC that I created from scratch using and ASUS motherboard and a HD4350 graphics card.  For testing I am using a HDMI cable from the HD4350 to a modern LCD monitor.

I installed from a USB onto a hard drive, with the partitions etc and then booted from the hard drive, so far so good.

But now when I try to run the software I’m getting an error:

Using config file: “/etc/X11/.org.conf”
Using system config directory “/user/share/X11/.org.conf.d”
Server terminated with error (1).  Closing log file.log” for additional information.

Any advice on next steps?  Could it be a graphics card driver issue? 

Thanks

[MonkeyJug]

i'm getting the exact same issue after a fresh install.  i'm using the same card, but with DVI output direct to a WG D9200 CRT monitor.

here is my log:

[   757.288] (WW) Failed to open protocol names file lib/xorg/protocol.txt
[   757.289]
X.Org X Server 1.20.5
X Protocol Version 11, Revision 0
[   757.289] Build Operating System: Linux Arch Linux
[   757.290] Current Operating System: Linux GroovyArcade 5.2.4-arch1-1-ARCH #2 SMP PREEMPT Fri Aug 2 07:06:57 CEST 2019 x86_64
[   757.290] Kernel command line: BOOT_IMAGE=../vmlinuz-linux root=/dev/disk/by-label/GA  rw quiet rd.udev.log-priority=3 splash vga=0x311 video=DVI-I-1:640x480ec initrd=../initramfs-linux.img
[   757.291] Build Date: 30 June 2019  09:52:01AM
[   757.292] 
[   757.292] Current version of pixman: 0.38.4
[   757.293]    Before reporting problems, check http://wiki.x.org
   to make sure that you have the latest version.
[   757.293] Markers: (--) probed, (**) from config file, (==) default setting,
   (++) from command line, (!!) notice, (II) informational,
   (WW) warning, (EE) error, (NI) not implemented, (??) unknown.
[   757.295] (==) Log file: "/var/log/Xorg.0.log", Time: Sat Jul 20 20:07:57 2019
[   757.295] (==) Using config file: "/etc/X11/xorg.conf"
[   757.295] (==) Using system config directory "/usr/share/X11/xorg.conf.d"
[   757.296] (==) ServerLayout "General"
[   757.296] (**) |-->Screen "Screen0" (0)
[   757.296] (**) |   |-->Monitor "DVI-0"
[   757.296] (**) |   |-->Device "Card0"
[   757.296] (**) |-->Input Device "WiiMote0"
[   757.296] (**) |-->Input Device "WiiMote1"
[   757.296] (==) Automatically adding devices
[   757.296] (==) Automatically enabling devices
[   757.296] (==) Automatically adding GPU devices
[   757.296] (==) Automatically binding GPU devices
[   757.296] (==) Max clients allowed: 256, resource mask: 0x1fffff
[   757.296] (WW) `fonts.dir' not found (or not valid) in "/usr/share/fonts/misc".
[   757.296]    Entry deleted from font path.
[   757.296]    (Run 'mkfontdir' on "/usr/share/fonts/misc").
[   757.296] (WW) The directory "/usr/share/fonts/TTF" does not exist.
[   757.296]    Entry deleted from font path.
[   757.296] (WW) The directory "/usr/share/fonts/OTF" does not exist.
[   757.296]    Entry deleted from font path.
[   757.296] (WW) The directory "/usr/share/fonts/Type1" does not exist.
[   757.296]    Entry deleted from font path.
[   757.296] (==) FontPath set to:
   /usr/share/fonts/100dpi,
   /usr/share/fonts/75dpi
[   757.296] (==) ModulePath set to "/usr/lib/xorg/modules"
[   757.296] (II) The server relies on udev to provide the list of input devices.
   If no devices become available, reconfigure udev or disable AutoAddDevices.
[   757.296] (II) Module ABI versions:
[   757.296]    X.Org ANSI C Emulation: 0.4
[   757.296]    X.Org Video Driver: 24.0
[   757.296]    X.Org XInput driver : 24.1
[   757.296]    X.Org Server Extension : 10.0
[   757.296] (++) using VT number 1

[   757.297] (II) systemd-logind: took control of session /org/freedesktop/login1/session/_31
[   757.297] (II) xfree86: Adding drm device (/dev/dri/card0)
[   757.298] (II) systemd-logind: got fd for /dev/dri/card0 226:0 fd 10 paused 0
[   757.299] (--) PCI:*(1@0:0:0) 1002:954f:174b:174b rev 0, Mem @ 0xe0000000/268435456, 0xf7e20000/65536, I/O @ 0x0000e000/256, BIOS @ 0x??/131072
[   757.299] (WW) Open ACPI failed (/var/run/acpid.socket) (No such file or directory)
[   757.299] (II) LoadModule: "glx"
[   757.299] (II) Loading /usr/lib/xorg/modules/extensions/libglx.so
[   757.300] (II) Module glx: vendor="X.Org Foundation"
[   757.300]    compiled for 1.20.5, module version = 1.0.0
[   757.300]    ABI class: X.Org Server Extension, version 10.0
[   757.300] (II) LoadModule: "admgpu"
[   757.300] (WW) Warning, couldn't open module admgpu
[   757.300] (EE) Failed to load module "admgpu" (module does not exist, 0)
[   757.300] (II) LoadModule: "evdev"
[   757.300] (WW) Warning, couldn't open module evdev
[   757.300] (EE) Failed to load module "evdev" (module does not exist, 0)
[   757.300] (EE) No drivers available.
[   757.300] (EE)
Fatal server error:
[   757.300] (EE) no screens found(EE)
[   757.300] (EE)
Please consult the The X.Org Foundation support
    at http://wiki.x.org
 for help.
[   757.300] (EE) Please also check the log file at "/var/log/Xorg.0.log" for additional information.
[   757.300] (EE)
[   757.301] (EE) Server terminated with error (1). Closing log file.

[MonkeyJug]

i have spotted an error in the log:

[   757.300] (II) LoadModule: "admgpu"
[   757.300] (WW) Warning, couldn't open module admgpu
[   757.300] (EE) Failed to load module "admgpu" (module does not exist,

It is looking for admgpu and it should be AMDgpu.  i have changed it in the xorg.conf, and it now loads the 'amdgpu' module, but it still doesn't run the front-end...

[Calamity]

Hi MonkeyJug,

Yes, the iso from August was just a test build for me, I've used it for testing Ryzen 5, it uses amdgpu instead of radeon, it has some bugs as you've seen that need to be fixed manually. You can't use it for older cards.

Sorry for the annoyance.

I've renamed the iso as "_test".

[Substring]

Anyway, X driver detection should be rewritten in GA, you can't just rely on PCI vendors. X has an autoconfigure feature that should be taken advantage of.

[Calamity]

Anyway, X driver detection should be rewritten in GA, you can't just rely on PCI vendors. X has an autoconfigure feature that should be taken advantage of.

I agree.

[baldgeek]

So which iso should I use if plan on using Ryzen 3400G with AMDGPU?

[b4nd1t0]

I have a strange problem, recently if I change a game 3 or 4 times, at the next change attract (or mame?) no longer loads any rom, I have a black screen and then it returns to the frontend, as if the rom was not available. Some idea? Has it happened to someone else?
Today i compiled mame 215 and attract 2.6.0 to see if updating solves the problem, but it's very strange, never happened before.

[Doozer]

I have a strange problem, recently if I change a game 3 or 4 times, at the next change attract (or mame?) no longer loads any rom, I have a black screen and then it returns to the frontend, as if the rom was not available. Some idea? Has it happened to someone else?
Today i compiled mame 215 and attract 2.6.0 to see if updating solves the problem, but it's very strange, never happened before.

Do you have a mame log or other?

[b4nd1t0]

no, no log, now i have updated the groovymame with 0.215 (before it had been updated with a 206) and the problem seems to have disappeared, i've tried changing game at least 20 times and everything seems to work again, very strange, as if after some reboot mame could no longer find the roms folder.

[Doozer]

no, no log, now i have updated the groovymame with 0.215 (before it had been updated with a 206) and the problem seems to have disappeared, i've tried changing game at least 20 times and everything seems to work again, very strange, as if after some reboot mame could no longer find the roms folder.

happy ending then

[b4nd1t0]

happy ending then

Unfortunately no  , the problem is back, how can I create a log? if I start the commandline mame the problem does not occur, I should create an attract log maybe, do you have advice on how to proceed?

[Doozer]

I do not know if their is an easy way to do it with GA from the front-end.

Generally, you have to run GM from the shell and add the "-verbose" parameter and redirect it to a file "> filename.ext". Here is an example which redirect the output and print it on the screen at the same time.

Code: [Select]

groovymame -verbose <name_of_the_rom> 2>&1 | tee /tmp/mame.debug.txt


[Substring]

Read Calamity's signature 

_{Steps to create a log:
 - From command line, run: groovymame.exe -v romname >romname.txt
 - Attach resulting romname.txt file to your post, instead or pasting it.}

[b4nd1t0]

I know, but in this case is different, if I launch mame from bash the game start normally, the problem is with attract, I need to verify if the lastrun log is created as in windows.

Sent from the sixth floor

[Substring]

Ask keilmillerjr how you could log everything from Attract