vulkano-rs/Lobby

Please provide a minimal reproduction of your issue in a repo when asking for help on a specific problem.

Faule Socke

@faulesocke

btw, who is currently maintaining vulkano? looks like @rukai gave up

Antonino Siena

@nifker

Probably @tomaka

Faule Socke

@faulesocke

@nifker I've looked some more into this now. rspirv looks promising, however vulkano-shaders already has some parsing code dedicated for this purpose that also seems to be a bit more suited to what I want to do. For example it is capable of detecting all the group and member related instructions I will need in order to generate shader input/output specifications.

Oh and I found that someone else is also working on this topic :) vulkano-rs/vulkano#945

Antonino Siena

@nifker

@faulesocke That seems nice so far but Ill primary use SPIR-V shaders for my code

I'd like to know if CpuAccessibleBuffers are allocated on the GPU's RAM or the DIMM's RAM?

Faule Socke

@faulesocke

Why is this an unsafe trait? https://docs.rs/vulkano/0.13.0/vulkano/pipeline/shader/trait.ShaderInterfaceDef.html

It looks like something that could be easily derived

@nifker cpuaccessible is always host memory

Antonino Siena

@nifker

Then it wonders me that its not possible to increase the buffer size

Faule Socke

@faulesocke

that's a vulkan limitation. a buffers size is determined upon creation and can never change, no matter where it's memory is allocated.

AFAIR in modern opengl (i.e. with bindless functions) this limitation exists as well

so if your buffers size changes on a regular basis just create a new one each time and throw the old one away (which you can do easily in vulkano since it keeps track of the buffer as long as it's in use)

Antonino Siena

@nifker

Ah thats good to know

AvengeMe

@AvengeMe

hey everyone! i'm trying to break into game dev via rust, and vulkano looked like a solid option. i've got several questions, one of which might actually be appropriate for this forum lol. sorry if this is the wrong place for any of em.

1 - how hard would it be to set up a pipeline that uses many attributes and buffers for the graphics pipeline? or am i just using the impl_vertex! macro wrong?? it seems like all of the examples really only use one, and all of the implementations of standard set ups (ie vertex_input_single_buffer, etc) seem pretty darn complicated to implement. i feel like using many attributes is pretty standard - is there something i'm missing here?
like, i could probably do it if i had an example that actually used more attributes. the deferred rendering seems to fake normal vectors with constants. i had vulkan up and running with one, no problem, but couldn't figure out how to set it up with more. i think what makes it so hard is that several places all have to agree as to what the pipeline looks like, so my normal strategy is sussing the answer out by poking it won't really work. the stars have to align lol.

2 - lucas said vulkano is "pretty much unmaintained at this point". is vulkano dead? should we abandon ship? :( :( :( is it close enough to finished to make it viable anyway?

3 - are there other forums more appropriate for these kinds of questions? like actually using vulkano questions, not developing it or pointing out bugs and such.

Lucas Kent

@rukai

I mean if you can get value of it by all means use it. But even if you fix things there will be no one to merge the fixes.

AvengeMe

@AvengeMe

sooooo abandon ship? XD

bummer - vulkano looked to be exactly what i was hoping for!!

any suggestions for alternatives?

Lucas Kent

@rukai

https://github.com/gfx-rs/wgpu-rs

Its in heavy development so use the latest git version and it lacks in documentation but its api is somewhat similar to vulkano

AvengeMe

@AvengeMe

i overlooked this one because of the WEB part - wgpu doesnt lose performance for being web friendly does it? im trying to develop games with large render distances and too many lights and such, so i seriously need to not pay too much overhead if possible. is this still the right tool for me?

Lucas Kent

@rukai

It does have a lot more runtime checks than vulkano. But I found vulkano's compile time checks to be unwieldy. If you really truly need as much performance as possible gfx-hal + rendy might be what your after.

AvengeMe

@AvengeMe

hey i see youre a dev for it! thats really cool man. id love to hear about your work some time.

Lucas Kent

@rukai

hehe I'm just learning

AvengeMe

@AvengeMe

i was looking at rendy but i couldnt figure out how to even initialize it lmfao. based on your suggestion of using both, i take it i must initialize gfx hal and then use rendy.

im just learning too. ive been coding here and there for a long while but im just getting my toes wet with game dev at this high level. rust is /really/ exciting.

Lucas Kent

@rukai

yeah rendy is a collection of helpers built on top of gfx hal, so you will be using both together

right, well if your just learning, i'd recommend trying wgpu first. Then you can try something lower level if your really hitting its limits.

AvengeMe

@AvengeMe

ill have to give them both a close look. pretty sure you just saved me countless hours - thanks a ton!

Lucas Kent

@rukai

btw you might want to check out the rust community discord game-and-graphics-dev channel, lots of cool people there. https://bit.ly/rust-community

AvengeMe

@AvengeMe

yeah thats what ive been thinking - need to get a prototype going asap not mess around learning overly hard bits. i'll poke the perf AFTER i get funding XD - premature optimization and whatnot

didnt know it existed - thanks!

i was looking pretty close to that amethyst engine - seems like it might save me time to learn that instead of trying to roll my own solution. and rendy is part of that. but i might go ahead and use wgpu to save time.

if anything all those runtime checks will help me debug and save time, so i should have saved up enough time to hot swap it later on~

Lucas Kent

@rukai

to be clear wgpu is still missing a lot of the runtime checks that would make it safe. But vulkano's compile time checks werent perfect either...

AvengeMe

@AvengeMe

well i'm not necessarily looking for safe so much as fully functional. i don't want to not be able to do something i want to. that's my main concern with several of these other libs. one of the tools i considered had the ability to write to the gpu but not read from it. that seems like something that might be a problem for me! ;) anyway, by the time i'm ready to ship i'm sure y'all will be further along. :) thanks again for everything.

Antonino Siena

@nifker

@rukai Why do you keep recommending WGPU if it was only developed for the Web?
Most people that land here look for implemented stuff on top of Vulkan, and not on three totally different graphics API.

Lucas Kent

@rukai

It's not only developed for the web

In fact you can't currently run it on the web

AvengeMe

@AvengeMe

yeah i thought it was for the web at first too but its basically just a big runs-everywhere project from what i can tell, which i have no problem with

Antonino Siena

@nifker

When I execute a commandbuffer and it sometimes returns this error https://paste.gg/p/anonymous/684d886e7a854723bac87187c321a33c
When it returns this error three times, the next time acquire_next_image() is called it will block forever even when a timeout is. Any ideas on this?

Antonino Siena

@nifker

Validation layers dont reveal issues too

Ruslan Feizerahmanov

@Russoul

Hi all ! Is it possible to give OpenGL assess to an image buffer created in vulkano

CreepySkeleton

@CreepySkeleton

When I try to create an instance by Instance::new(None, &InstanceExtensions::none(), None) I get Err(OomError(OutOfHostMemory)). What would that mean? While I do admit my laptop is not hi-end tech at all it still has 512 Mb of video memory and I don't do any memory consuming stuff at this point. In fact, this line is the only line in main function :)

Antonino Siena

@nifker

@rukai are you gonna accept the PR? rukai/vulkano-text#4

Antonino Siena

@nifker

Seems to be AWOL

@CreepySkeleton Could be an issue in your Vulkan implementation - You might better ask in Mesa/AMD/Nvidia forum

Faule Socke

@faulesocke

Whats the difference between AutoCommandBuffer and SyncCommandBuffer?

Antonino Siena

@nifker

@faulesocke It seems to have a lock an is thus accessible from multiple threads

Where communities thrive

People

Repo info

Activity