Emulator to test ComLynx ?

[KanedaFr]

Is there any emulator with ComLynx support ?

 

I was thinking of an emulator you can launch several instances (with same rom) and able to talk to each other for exemple...

 

I'm working on a puzzle game and I'd like to add multi players support if I can test it.

[obschan]

You can give a try to https://github.com/LLeny/VkHandybug
It's not perfect by it does the basics for comlynx. It helped me enough to make some progress on my project.

[KanedaFr]

Thanks, the Vulkan part made me run away when I found it some months ago but it seems I had to give it a go

Let's see if I'm able to build it for OSX !

 

Edit: it's a fail and, as usual with CMake, it's such a black box that I don't know why
Edit2: I finally found why I wasn't able to build but while everything seems to work (tools update as expected), only display tab hangs

Edited February 22 by KanedaFr
semi-fail

[+karri]

On 2/21/2024 at 11:01 AM, KanedaFr said:

the Vulkan part

Really! I used to play with Vulkan on the RaspberryPi 400. I wonder if this compiles on the Pi 400?

The serial port of Raspberry Pi's can support weird baudrates  

[42bs]

On 2/21/2024 at 10:01 AM, KanedaFr said:

Thanks, the Vulkan part made me run away when I found it some months ago but it seems I had to give it a go

Let's see if I'm able to build it for OSX !

 

Edit: it's a fail and, as usual with CMake, it's such a black box that I don't know why
Edit2: I finally found why I wasn't able to build but while everything seems to work (tools update as expected), only display tab hangs

After applying your hints, I could build on MacOS (M1) and it runs.

[KanedaFr]

On 2/25/2024 at 10:02 AM, 42bs said:

After applying your hints, I could build on MacOS (M1) and it runs.

 

So lucky!

Which version of Vulkan did you use ?

I suspect the last version wasn't the one to use....

I'm still on Intel but I doubt it would be the reason

 

I'll retry later today, because I really like what I see for now : I'm hungry for more!

 

[42bs]

I use

Vulkan Instance Version: 1.3.275

 

which is the latest AFAIK. Be sure to install the vkmemalloc which is optional.

 

Also I had to do "cmake -G Xcode" , simply "cmake" chooses Unix/clang

[KanedaFr]

CMake Error:
  Xcode 1.5 not supported.
CMake Error: Could not create named generator Xcode

 

not my day...

 

======

 

Ok, it doesn't help but for some reason, Vulkan wrongly detects my device properties

 

imageExtent (93888,1816), which is bigger than max extent (16384,16384)returned by vkGetPhysicalDeviceImageFormatProperties()

 

and if i select the other one, it's ever worst (20000 pixels width!!)

 

unfortunately, I'm unable to fix anything, since I learnt about Vulkan 5 days ago

Edited February 26 by KanedaFr

[42bs]

Wow, you are having 10 5K screens around you? Pretty awsome.

 

But, yes, I am also lost with these things. If they do not work out of the box.

 

But Xcode 1.5? I have 15.2. How old is your machine?

 

[42bs]

I could try building for an older MacOS and x86, but I never worked with Xcode, this tool is so non-intuitive to use

Which MacOS version do you have?

[42bs]

Ok, it seems _I_ cannot build VkHandybug from within Xcode. At least I did not find how. I hope I never _have_ to write iOS apps

[KanedaFr]

I use a XCode switcher because mobile apps is my main job.
I forced XCode 14.2 to be used (since I'm locked to Monterey)

 

cmake -G Xcode

build in progress !! 🤞

[KanedaFr]

So I successfully created an XCode project

I found how to build the vkHandyDebug subproject from within XCode

It runs and no longer thinks I have a 100000 pixels wide screen (my record)

...but display window is still dead (black or still garbage)

 

so, same issue, unfortunately

Edited February 27 by KanedaFr

[KanedaFr]

now that I have a XCode project, I'm able to debug and, from what I see, the data is read from the right location(s) on Lynx RAM

so it's not a source data issue

 

I can also confirm the view.allocationInfo.pMappedData->buffer is correct (ie memcpy didn't bug)

so it's not a destination data issue

 

So, the problem seems located on the renderer...
I hope it's not on the shader, it's my worst nightmare!

 

(hopefully I debug my own project, so I KNOW what should be drawn on screen  )

Edited February 27 by KanedaFr

[obschan]

2 hours ago, KanedaFr said:

now that I have a XCode project, I'm able to debug and, from what I see, the data is read from the right location(s) on Lynx RAM

so it's not a source data issue

 

I can also confirm the view.allocationInfo.pMappedData->buffer is correct (ie memcpy didn't bug)

so it's not a destination data issue

 

So, the problem seems located on the renderer...
I hope it's not on the shader, it's my worst nightmare!

 

(hopefully I debug my own project, so I KNOW what should be drawn on screen  )

What does the screen look like?

[KanedaFr]

41 minutes ago, obschan said:

What does the screen look like?

 

it's never the same...

most of the time full black, mid black or "pink corrupted" or total noise

If I connect 2 screens and switch for one to another, I can corrupt the display non stop....(something different every time, pure "art" like you can see on the zip file)

 

If that help, I found something strange:

I have a lot of validation errors after the call to VulkanRenderer::prepare_view_compute

 

For exemple:

VUID-VkShaderModuleCreateInfo-pCode-08740(ERROR / SPEC): msgNum: 115483881 - Validation Error: [ VUID-VkShaderModuleCreateInfo-pCode-08740 ] | MessageID = 0x6e224e9 | vkCreateComputePipelines(): pCreateInfos[0].stage SPIR-V Capability Int16 was declared, but one of the following requirements is required (VkPhysicalDeviceFeatures::shaderInt16). The Vulkan spec states: If pCode is a pointer to SPIR-V code, and pCode declares any of the capabilities listed in the SPIR-V Environment appendix, one of the corresponding requirements must be satisfied (https://vulkan.lunarg.com/doc/view/1.3.268.1/mac/1.3-extensions/vkspec.html#VUID-VkShaderModuleCreateInfo-pCode-08740)
    Objects: 0

 

while using Vulkan Capability viewer I can confirm shaderInt16 is available.

the same apply for many others "missing" capabilities

 

 

corrupted_display.zip

[KanedaFr]

if it could help

 

and i tested several of the samples (particles, wave, triangle....) with success

 

be free to contact me by MP if you want me to make some tests, even updating code since XCode project is now live

[KanedaFr]

For anyone interested, with the help of @obschan, I finally made it works.

What I did:

- pass all the required features to physical device selector (remove 2/3 of errors)

- force required API version to 1.2 on instance creation (remove the last 1/3 of errors)

- move font and shaders folder on Debug folder (wrongly put on bin folder)

- force the device to my Intel Graphic card using vkConfig (while every demo worked like a charm, vkHandybug dislikes my Radeon Pro 455 but I wasn't able to force the intel card w/o removing the errors first)

 

I'm now ready for multi players development!  

Edited February 29 by KanedaFr

[42bs]

Can you post a diff because, it will make it easier to make these changes.

 

For example I get these errors though it runs:

VUID-RuntimeSpirv-uniformAndStorageBuffer16BitAccess-06332(ERROR / SPEC): msgNum: 231017936 - Validation Error: [ VUID-RuntimeSpirv-uniformAndStorageBuffer16BitAccess-06332 ] Object 0: handle = 0xf56c9b0000000004, type = VK_OBJECT_TYPE_SHADER_MODULE; | MessageID = 0xdc50dd0 | vkCreateComputePipelines(): pCreateInfos[0].stage SPIR-V contains an 16-bit OpVariable with Uniform Storage Class, but uniformAndStorageBuffer16BitAccess was not enabled

 

No idea how to fix this.

Edited February 29 by 42bs

[KanedaFr]

@42bs I tried to do something easier than diff but it took me so long with the Github authent' stuff I'm not used to

 

https://github.com/LLeny/VkHandybug/pulls waiting for review and merge