Announcing rlsw, the new raylib software renderer backend. No GPU required. The Future is Now. #raylib

I still prefer RLSW off. You clearly see the vision and direction, which is brutally destroyed by RLSW!!!

@tsoding would love this.

DLSS 5 ON

Huh this means raylib could theorically be ported to something like the GBA maybe...

not really, performance is way lower than GPU-accelerated graphics, it need some decent CPU but for example, it can run on an ESP32 S3

CPU-only backend is legit useful. Containerized game dev just got way easier.

Headless without virtual screen buffers etc?

It renders to a frame buffer, the final output depends on the platform selected, PLATFORM_MEMORY just gives you the buffer that can be saved to a file if you want, PLATFORM_DESKTOP_WIN32 preconfigures a window with GDI and outputs that framebuffer

raylib can now run doom

Building my finance app, I've noticed that sometimes the most powerful solutions come from removing dependencies, not adding them. Same thing happened when I had to optimize for low-end devices, had to get creative with rendering. No GPU required is a game changer.

funny how everyone's chasing GPU acceleration when embedded devs have been doing CPU-bound rendering since the 90s. the real win here isn't 'no GPU' - it's predictable performance on $5 microcontrollers where every cycle counts. ESP32-S3 at 60fps for simple UIs beats a $2000 GPU that can't boot without drivers.

This is beautiful haha

? we could always since the dawn of time render stuff using the cpu it's just that it's so inefficient being able to render stuff using a cpu is not the problem heck, even right now I'm replying to you and seeing your vid in a vm without gpu acceleration!

software rendering.. holy shot

that's cool but CPU computed is gonna destroy my laptop

My CPU watching the GPU doing nothing be like:

i'll assume it doesnt support shader code for obvious reasons, unless you straight up decided to make GLSL - > software rendering

you assume correctly, no shaders support, at least for now...

Doesn't the information still eventually need to go through the GPU to be actually placed on the screen?

yes, software renderer just outputs to a framebuffer that needs some way to be displayed on screen, raylib supports that through different platform backends, like Win32 GDI on Windows BUT it also supports PLATFORM_MEMORY backend allowing to save buffer directly to a file image.

Ok, thanks, interesting.

Nice! More for the Raylib SDK

This is hilarious lmao.

so cool!

Nvidia in shambles

this demo has a very cool Y2K feel to it

I can finally make games on the toaster

That is an amazing achievement.

Thank you so much! This makes Raylib unironically a really good library for UI rendering.

but why would we do this through CPU? Is this for smaller MCU platforms??

yes, there are multiple microcontrollers that can benefit from this, also automated graphics processing on servers in headless mode, there are multiple applications

Whats the difference between this and how raylib normalny works? Sorry, i dont know much about computer graphics

the timing on this is crazy lol

How is it different from other cpu based renderers?

On a threadripper.... must diminish window size and add viewport to play decently with old CPU.

If I’m developing a game for regular PCs, how it can benefit me?

correct me if I'm wrong, but the point is that it looks and performs almost the same even when it's on, in other words, even when GPU acceleration isn't available.

software renderer is +10 times slower than GPU-accelerated mode but small and relatively simple examples can still run at 60 fps without problem

oh I was addressing people in the comments who were confused about it! (thinking it is some kind of graphics enhancement) thanks for clarifying!

Would it be easy to switch between software and hardware renderer backend? If so, this would be perfect to make desktop software and add hardware rendering afterwards.

At this moment, software renderer must be selected at compile time, not runtime, the final output to draw generated frames depends of how the window is configured, for example in Windows using GDI