Axmol RHI Takes a Major Step Forward with Cross-Backend GPU Compute
Published on September 9, 2026 by halx99
Axmol's Render Hardware Interface has taken another major step forward.
We have introduced first-class cross-backend GPU compute support, expanding the RHI from a graphics-oriented abstraction into a broader foundation for modern GPU workloads.
The new infrastructure includes compute programs, pipelines and dispatch, storage buffers, 3D textures, compute shader reflection, runtime sampler overrides, device capability and limit queries, compute-capable OpenGL / OpenGL ES context selection, and per-shader target profiles in the shader toolchain.
The implementation is available across Direct3D 11, Direct3D 12, Metal, Vulkan, OpenGL, and OpenGL ES, and has been successfully tested across the supported RHI backends.
The first major real-world feature built on this infrastructure is full Effekseer GPU particle integration. Rather than introducing a backend-specific implementation, Effekseer GPU particles now run through Axmol's common RHI architecture.
GPU compute remains an optional capability. It does not raise the engine's minimum requirements for normal rendering: older OpenGL 3.3 / OpenGL ES 3.0 devices can continue using the existing graphics path, while compute features require OpenGL 4.3+, OpenGL ES 3.1+, or an equivalent compute-capable backend. WebGL is not supported by the current compute path.
This work lays the foundation for more GPU-driven features in the future, including GPU culling, procedural data generation, GPU animation preprocessing, instance-data generation, compute-based post-processing, and other simulation workloads.
Read the full technical article on GitHub:
From Graphics to GPU Compute: A Major Step Forward for the Axmol RHI
We have introduced first-class cross-backend GPU compute support, expanding the RHI from a graphics-oriented abstraction into a broader foundation for modern GPU workloads.
The new infrastructure includes compute programs, pipelines and dispatch, storage buffers, 3D textures, compute shader reflection, runtime sampler overrides, device capability and limit queries, compute-capable OpenGL / OpenGL ES context selection, and per-shader target profiles in the shader toolchain.
The implementation is available across Direct3D 11, Direct3D 12, Metal, Vulkan, OpenGL, and OpenGL ES, and has been successfully tested across the supported RHI backends.
The first major real-world feature built on this infrastructure is full Effekseer GPU particle integration. Rather than introducing a backend-specific implementation, Effekseer GPU particles now run through Axmol's common RHI architecture.
GPU compute remains an optional capability. It does not raise the engine's minimum requirements for normal rendering: older OpenGL 3.3 / OpenGL ES 3.0 devices can continue using the existing graphics path, while compute features require OpenGL 4.3+, OpenGL ES 3.1+, or an equivalent compute-capable backend. WebGL is not supported by the current compute path.
This work lays the foundation for more GPU-driven features in the future, including GPU culling, procedural data generation, GPU animation preprocessing, instance-data generation, compute-based post-processing, and other simulation workloads.
Read the full technical article on GitHub:
From Graphics to GPU Compute: A Major Step Forward for the Axmol RHI