The graphics programming landscape is witnessing an interesting evolution in SPIR-V tooling, with discussions centering around LLVM's implementation and the emergence of new alternatives like the Hero C Compiler (HCC). This development comes at a time when GPU programming frameworks are becoming increasingly crucial for modern graphics applications.
LLVM's Current SPIR-V Status
Recent community discussions reveal that LLVM's support for SPIR-V is in a state of transition. While LLVM does support compute mode SPIR-V, its graphics-mode SPIR-V implementation is still in early stages, being tested through a work-in-progress HLSL front-end in Clang. The Microsoft DirectX Shader Compiler, though based on LLVM and supporting SPIR-V output, is reportedly constrained by technical debt and version limitations.
The Microsoft HLSL compiler for dx12+ is based on LLVM and supports spir-v output... Though my understanding is that it's a complete hackjob stuck on an old version of LLVM, and not in a remotely upstream-able state
Key LLVM SPIR-V Implementation Status:
- Compute mode: Fully supported
- Graphics mode: Early implementation stage
- HLSL front-end: Work in progress
- DirectX Shader Compiler: Based on older LLVM version
Hero C Compiler Features:
- Vulkan 1.3+ support
- C11 targeting
- Direct SPIR-V compilation
- Vertex, Pixel & Compute shader support
- No pointer support (outside compiler intrinsics)
- No function pointers
- Windows & Linux platform support
Hero C Compiler as an Alternative
The Hero C Compiler has emerged as a promising alternative, offering direct C-to-SPIR-V compilation for Vulkan graphics API. Unlike the current LLVM implementation, HCC takes a more straightforward approach by implementing shader functions with specific specifiers like __hcc_vertex rather than using complex attribute annotations. This design choice reflects a focus on simplicity and directness in shader programming.
Performance and Optimization Considerations
The community has raised interesting points about optimization potential, with some developers suggesting the exploration of QBE-like optimizations for GPU targets. However, technical limitations, particularly around SIMD support, present challenges for direct adaptation of CPU-oriented optimization techniques to GPU shader compilation.
Future Outlook
The landscape appears to be evolving rapidly, with Microsoft's recent announcement about DirectX adopting SPIR-V indicating a broader industry shift toward standardization. This movement, combined with the emergence of alternative tools like HCC, suggests a future where GPU programming becomes more accessible while maintaining high performance capabilities.
In conclusion, while LLVM continues to develop its SPIR-V support, new tools like HCC are providing alternative paths for developers looking to work with modern graphics APIs, particularly in the Vulkan ecosystem. This diversity in tooling options may prove beneficial for the graphics programming community as a whole.
Source Citations: Hero C Compiler