Skip to content

Add comprehensive Vulkan GPU support to IREE backend#6

Merged
athrva98 merged 1 commit into
masterfrom
private/athrva/add-comprehensive-vulkan-gpu-support
Dec 26, 2025
Merged

Add comprehensive Vulkan GPU support to IREE backend#6
athrva98 merged 1 commit into
masterfrom
private/athrva/add-comprehensive-vulkan-gpu-support

Conversation

@athrva98
Copy link
Copy Markdown
Owner

Features:

  • Add VulkanTarget dataclass and VulkanGPUVendor enum
  • Add VULKAN_TARGETS dict with 40+ GPU presets:
    • AMD: RDNA3/RDNA2 (RX 7000/6000), gfx1100/gfx1030
    • NVIDIA: Ada/Ampere/Turing (RTX 40/30/20), sm_89/sm_86/sm_75
    • Intel: Arc A-series (A770, A750)
    • ARM: Mali Valhall (G715, G710)
    • Qualcomm: Adreno
  • Add IREECompileOptions dataclass with configurable:
    • opt_level (0-3), data_tiling, vulkan_target
    • opset_version for ONNX upgrade, strip_debug, const_eval
    • extra_flags for custom compiler options
  • Add IREECompilationError with actionable suggestions
  • Add opset_version param to emit_mlir() for ONNX compatibility

Code quality:

  • Fix ruff linting (SIM102, B904, SIM115)
  • Add proper exception chaining with 'from e'
  • Use context manager for tempfile

Tests:

  • Add TestIREECompileOptions (8 tests)
  • Add TestVulkanTargets (8 tests)
  • Add TestIREEErrorHandling (4 tests)
  • Add TestMLIRExportWithOptions (3 tests)
  • Fix test_intel_devices.py backend name assertions

Documentation:

  • Update README with Vulkan GPU targets table
  • Add IREE backend configuration examples
  • Add troubleshooting for common IREE errors

@athrva98 athrva98 self-assigned this Dec 26, 2025
@athrva98 athrva98 merged commit 2a1b1fd into master Dec 26, 2025
13 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant