Well, I heard you can do some impressive stuff with modern GL in terms of performance
Well, I heard you can do some impressive stuff with modern GL in terms of performance
Lyea performance from GL wont matter for 99% of cases
Lwhat makes D3D11 a decent option is the arguably better API than GL
ZHmmm well I still have time to choose as I don't have as much time right now to code so
Lif you've already started it doesnt matter, stick with GL
Lhopping APIs isnt good for a beginner
ZHaven't started yet but thanks for the tip
ZBtw has anyone ever implemented ECS (doesn't matter which specific implementation) with Silk?
LECS implementations are generally independent of windowing and even rendering, so any ECS implementation can generally be made to work
ZYeah, just thinking about possible implementation and handling input and etc.
ZBut I think it's still manageable
ZFrom what I've seen Silk seems nice and flexible
ZWanted to try it out with Arch
ZSeems pretty user-friendly while also retaining one of the best performance results
ZProbs gonna take a while to get used to ECS
FThanks, I will take tutorial classes as base. They look logical, even though I don't need multiple shaders to draw simple unlit object.
ZGood luck!
FWatched “water rendering” tutorial on YT, they used image of landscape for reflections on water. In this case two shaders (landscape and water) are merged in code in one wall of code.
But this aproach looks not very manageble. What if we add some new effect for our “Landscape”, will need to redo water. Must be some “Manager” class which will glue all together
But this aproach looks not very manageble. What if we add some new effect for our “Landscape”, will need to redo water. Must be some “Manager” class which will glue all together
RAbsolutely! That being said, GL also makes it harder to do anything performantly and the moment you want to try and do any form of multi-threaded rendering or multi-threaded resource creation. You're fucked
RSo if you're looking at writing a multi-threaded engine, you end up either passing a context around a lot (which kills all the perf advantages), or doing a lot of deferred object creation (which might end up being faster but more complicated and requires quite a bit more synchronization).
RSo it really depends on what you're trying to do. DirectX 11 has no bindless texturing, or more accurately only 11.1 and newer has it, Don't bother targeting 11.1 exclusively, especially not with something as intrusive as that, there's literally no reason. Everything DirectX 11.1 compatible is probably fully DirectX 12 compliant as well, so it's kinda pointless. (Like nvidia 9xx and older IIRC don't have support for it)
RBut bindless texturing does allow you to do some neat optimizations, especially in 2d and billboard geometry for instance particle systems
FSingle game loop when we exactly know timings is God blessing 
Multithreaded async code is nightmare, in web we handle it all the time, but all "async" in reality is single threaded process on V8 engine which just switching tasks.
Multithreaded async code is nightmare, in web we handle it all the time, but all "async" in reality is single threaded process on V8 engine which just switching tasks.
Rnever said single game loop
Rcan absolutely be a dedicated renderthread
Rnot to mention especially if you are working on a 3d game, it can ABSOLUTLEY be useful to create and load meshes on a seperate or maybe a few seperate threads
Rfor speed and improved loading times
Respecially if working on an open world game were you are constantly streaming in and out game assets
Z@redhacker2 thanks for clarifying things!
FYou load assets on game start and forget. Loading at runtime is not good idea. You always limited by HDD or SDD, no matter how many threads will be created, it will not speed up IO process.
ZYou load assets on game start and forget.The entirety of them?
ZWouldn't it like, take quite a lot of RAM?
ZThan loading them runtime
FYou need only current level which is limited by size of GPU RAM.
Loading something bigger than 4GB on user device with 8Gb / 4 Gb RAM not really productive. It will swap as hell.
Loading something bigger than 4GB on user device with 8Gb / 4 Gb RAM not really productive. It will swap as hell.
ZSo the best strategy is to load resources required at the current level before it starts?
FUntil ALL user devices have fastest drives which really could load files without impact on overall system. Maybe in future we will have nice dirves and yes, load data at runtime without waiting.
FSoftware solutions in 1st order depend on user devices, and many users stil have old hardware, they use Windows 7 and install games on HDD.
LNo time to explain someone ping jvyden to patch me back in I know this is the wrong channel
LI am on my watch
Z@jvyden4
Jgot it
Jthey lost internet and have to talk through their watch, we have to communicate via them replying to discord messages
J
Z:0
NRThat is not always a valid strategy
RAgain, open world games
ZHmm
PIf you can’t find it in Silk.NET.OpenAL.Extensions.Soft then the answer is probably no, but thankfully OpenAL extensions are super easy to add if you wanted to give it a go!
Ntime to become a contributor
Nhow would you guys implement openal buffers and sources? since the preferred method is streaming audio data I guess its to make a pool of buffers and sources?
