UAE4ALL1 : Full speed Amiga 500 emulator with DispmanX

[Vanfanel]

I have the same problem. Was working a few months ago after spending some time realizing the need to add -lpthread in LDFLAGS line in the Makefile. Compiled with no errors after making larger swapfile as described in earlier post.

Before I was able to switch in/out of GUI with LCTRL, but now it just hangs after pressing "Run" in GUI. As described frozen black screen with the colored boxes. No kickstart image and doesnt respons to kill command via ssh. Not using any CPU according to top.

Very annoying as this seems like the best amiga emulator speedwise at the moment. Using latest firmware, might be related to earlier update.

Is this working for anyone else at the moment?

I have exactly this same problem.
I have backtraced the problem to the vsync-related semaphore, or gdb says... It's waiting to be signaled but it neved happens it seems.
I'm using the latest kernel & sdk versions (rpi-update). Other dispmanx apps work here: the problem seems to be related to the threaded mechanism this program uses to avoid waiting on vsync.
Chips: can you confirm vc_dispmanx_update_submit() callback works as expected with newer firmwares?

[Chips]

Hello,

I just tried and i have the same problem as you: the submit callback is no more called with the latest Raspberry firmware.
I tried to find what they changed and found a possible candidate:

https://github.com/raspberrypi/userland ... c8d0334a8b

They add a new Vsync callback service which could have broke something in the way i set Submit callback.

If i set my callback as both submit and the new vsync one, it solve this problem...
Which means that the submit callback is not taken in account but the Vsync is...

If i look at line 1293/1204 of the above RPI FW file, now there is a check at an handle before calling submit callback... this handle could be 0 but i have no idea where does this handle come from...

I will continue to see if i understand what changed or what is missing in my code...
EDIT:
The vsync callback seems quite buggy :
https://github.com/raspberrypi/userland/issues/218

Bye,

[RetroFredde]

Chips, others might have asked it already but have you had any time looking at Raspberry Pi 2 yet? and do you plan to bring your Amiga emulator to Pi 2?
I just got my Pi 2 and RetroPie already seems to run quite fine on it, RetroArch emulator with SNES/NES etc runs great but it would be even greater to have amiga games at full speed starting from EmulationStation GUI too with game art etc, not sure how difficult it would be to implement though.

[Chips]

If I understand well, It should work without any change on RPI 2. Of course compiling for Cortex instead of Arm11 should give a slight boost... but since it is already near full speed on RPI1 it should be full speed on RP2.
I think RetroPie used another version than my own Uae4all ... but i never try it.

[RetroFredde]

Oh I didn't know it already worked, I need to try it out then.
I tried the emulator included with buzz's RetroPie image for Pi2 but didn't get any game to run properly, I saw the iconic Turrican intro welding logo but never got to the main screen, it's like it never started loading again from the "disk" after the first stop.

[RetroFredde]

I tested the emulator on my RPi2 just now, the GUI works fine and runs smoothly but the emulator freezes after pressing Run, it shows the green and red boxes at the bottom right corner but nothing else happens and I have to pull the power cord. Don't know if it's because of the vsync issue you're talking about above or if I've messed something up, I've changed a bunch of config files to get sound and the arcade joystick working with MAME, I guess it could interfere somehow, I'll test some more to see if I get it to work.

[misterjbam]

For Information : In Raspberry PI "V2"

- uae4all-src-rc3.chips.0.2 : Works
- uae4all-src-rc3.chips.0.3 : Freeze
- uae4all-src-rc3.chips.0.4 : Freeze
- uae4all-src-rc3.chips.0.5 : Freeze

Thanks Chips for your great works .
Roll on V0.6 with HDF support

[Chips]

I tested the emulator on my RPi2 just now, the GUI works fine and runs smoothly but the emulator freezes after pressing Run, it shows the green and red boxes at the bottom right corner but nothing else happens and I have to pull the power cord. Don't know if it's because of the vsync issue you're talking about above or if I've messed something up, I've changed a bunch of config files to get sound and the arcade joystick working with MAME, I guess it could interfere somehow, I'll test some more to see if I get it to work.

This is the submit callback problem...
I made a new version which use sync API for submit, performance seems slightly lower but i should anyway release it as a temporary workaround until issue is understand...

[RetroFredde]

I tested the emulator on my RPi2 just now, the GUI works fine and runs smoothly but the emulator freezes after pressing Run, it shows the green and red boxes at the bottom right corner but nothing else happens and I have to pull the power cord. Don't know if it's because of the vsync issue you're talking about above or if I've messed something up, I've changed a bunch of config files to get sound and the arcade joystick working with MAME, I guess it could interfere somehow, I'll test some more to see if I get it to work.

This is the submit callback problem...
I made a new version which use sync API for submit, performance seems slightly lower but i should anyway release it as a temporary workaround until issue is understand...

Slightly lower performance you say. Is this on RPi2 and have you tried running the emulator with the same corrections on RPi1? I'm just asking because I tried the 0.2 version which misterjbam above said should work, and it did, but the performance was not good at all, like very very far from full speed, and this is on RPi2. Cyclone and fame gave the same results. I don't remember having any problem at all with speed on my RPi1. Could RPi2 actually be slower than RPi1 in some situations or when trying to run older code?

[Chips]

This version is using submit_sync API:
http://fdarcel.free.fr/uae4all-src-rc3. ... .6.tar.bz2
Don't forget to set your RPI to 50Hz mode...
In /boot/config.txt following do the trick (1080p 50Hz):

hdmi_group=1
hdmi_mode=31

It should work on RPI2 too, but i didn't tried since I don't have one . However there is no reason it would be slower on RPI2...
FYI, It looks like correction is ongoing on RPI firmware side... we will see.

[RetroFredde]

Wonderful, 0.6 works and runs perfectly on RPi2! Great work Chips!

Now I'm off to figure out how to get my joystick to work. I have two alternatives, a 360 fighting stick and a usb-joyport adapter which I can connect a real Amiga joystick to (would be awesome if I could get that to work).

Edit: Oh wow, I just plugged in the Retronic Design joyport usb adapter and didn't even have to map one single button, I'm playing Amiga games on Pi 2 with a real Amiga joystick right now! Woooo

[Vanfanel]

chips: correction on the firmware bug is now complete!

https://github.com/raspberrypi/firmware/issues/355

You can remove the blocking function call, go back to non-blocking, and it will work as fast as before. 0.5 works well here after latest firmware upgrade!
Also, could you please add the possibility of passing ADFs as parameters so we can bypass the menu?
Frameskip and OCS/ECS commandline options would be great, too! That would allow us to launch Amiga games from the commandline directly. Can you imagine? Simply type "lemmings", and there you go!

[dicko72]

Works great now thanks chips.

Although I get weird looking graphics. I'm running pal 50hz.

Any ideas?

What GPU mem should I use?

Thanks guys!

[Chips]

What do you mean by weird graphics ? Corrupted graphics or jitter in animations ? In every adf of for some adf only ? Was they working previously ?

I just checked the 0.6 delivery i made it looks like i uploaded the wrong one: the 0.6 i uploaded is using the vsync callback instead of using submit_sync API as the workaround... this solution wasn't so great so i would have prefered to upload the _sync version. I should be more carefull of what i upload.

Anyway the best solution would be to update your RPI Firmware and use the old 0.5.

Also, could you please add the possibility of passing ADFs as parameters so we can bypass the menu?
Frameskip and OCS/ECS commandline options would be great, too! That would allow us to launch Amiga games from the commandline directly. Can you imagine? Simply type "lemmings", and there you go!

I could look at merging the parameter management of Uae4all2...

[dicko72]

Corrupted graphics in 5 games that I have tried.
Some graphics appear OK.

Eg. Battle squadron the ship flys above the clouds at the start of the game normally. But now the ship disappears when the clouds come!

I can't say whether this is a new issue as I've never had it working before due to pi firmware issue.

[dicko72]

Does anybody else have this issue? Any things I should be checking?

Thanks

[Chips]

Try the other core... i think you are using the cyclone core, the fame core is better in term of compatibility but slighthly slower.

[dicko72]

I did try both. From a graphics perspective very similar results.

Any recommendations on overclock or gpu?

[Mamejay]

Hey everyone,
I am new to this so be kind. Just got a RP2 Model B and I am learning as I go.
I have installed Raspian and have it all working correctly.
i am trying to compile UAE4ALL but I seem to be getting an error everytime

/usr/bin/ld: /tmp/cc3NVm03.ltrans1.ltrans.o: undefined reference to symbol 'sem_post@@GLIBC_2.4'
//lib/arm-linux-gnueabihf/libpthread.so.0: error adding symbols: DSO missing from command line
collect2: ld returned 1 exit status
Makefile:200: recipe for target 'uae4all' failed
make: *** [uae4all] Error 1

I have read some threads about putting a -lpthread switch when compiling but I still get the same error.
Can any of you experts help on this one?

[Chips]

Since correction is available in Raspberry Firmware, it's better to go back to double buffering since it improve fluidity quite a lot on Raspberry Pi 1.
First update your firmware:
sudo apt-get update
sudo apt-get upgrade

Then use 0.7 version which use double buffering and added parameters support:

http://fdarcel.free.fr/uae4all-src-rc3. ... .7.tar.bz2

Following parameters are supported:
-df0 <name of the adf>
-df1 <name of the adf>
-kick <name of the kickstart>
-frameskip <nb of frameskip>

Mamejay : Use this new version, I think you added -lpthread in a wrong place inside the Makefile.

[Mamejay]

Thanks Chips. i did give it another go but now get a new error

src/m68k/fame/famec_opcodes.h: In function ‘void OP_0x57E7()’:
src/m68k/fame/famec_opcodes.h:18529:2: error: ‘l’ was not declared in this scope
src/m68k/fame/famec_opcodes.h:18529:4: error: ‘top’ was not declared in this scope
src/m68k/fame/famec_opcodes.h:18530:1: error: expected ‘;’ before ‘dd’
src/m68k/fame/famec.cpp:6049:1: error: expected ‘}’ at end of input
<builtin>: recipe for target 'src/m68k/fame/famec.o' failed
make: *** [src/m68k/fame/famec.o] Error 1
pi@raspberrypi ~/Downloads/uae4all-rpi $

What could I be doing wrong? I am going to try and find a pre-compiled version somewhere but its not looking good.

[Chips]

I re-uploaded it since it looks like some .h files were corrupted.
Between it's already compiled, you have two binary for each core: look for uae4all_fame and uae4all_cyclone files inside the archive.
And don't forget that compiling fame core fails quite easily on Rpi 1 due to low RAM quantity.

[Mamejay]

Perfect! Thanks for that. Will give it a try tonight when I get home. I am compile on a RPI2B so should be OK in terms of memory. Pardon my ignorance but which core version should I use?

[Chips]

They are both two different type of 68k core emulator.
cyclone core is less accurate but faster, fame core is slower but more accurate.
There is not so much differences between them anyway, fame is not so bad in term of speed.

[lubomyr]

I hope soon we will can see new ("old") cpu-engine for uae4all.
Which will be same fast as cyclone and famec, but more amiga compatible. But I can't write more info before its not released for public testing.