Skip to content

Latest commit

 

History

History
75 lines (51 loc) · 2.35 KB

RELEASE_NOTES.md

File metadata and controls

75 lines (51 loc) · 2.35 KB

Scenario Runner — Release Notes

Version 0.8.0 – Format Coverage & Tooling Upgrades

Execution & Scenario Control

  • Enabled storagePushConstant16 when the device advertises it so shader workloads can consume 16-bit push constants without custom tweaks.
  • Added --pause-on-exit to the runner CLI for easier debugging on interactive sessions.
  • Experimental:
    • Expanded Windows® parity by supporting nearly the entire DXGI format catalog, ensuring shaders and tensor IO behave consistently across platforms.

Build, Packaging & Developer Experience

  • Modernized the pip package: switched to pyproject.toml, added the missing metadata, and fixed package naming/installation ordering issues that affected --install.
  • Defaulted the build system to Ninja, refined the CMake packaging flow.
  • Introduced clang-tidy configuration and streamlined cppcheck invocation/CLI integration (including build-script driven execution).

Platform & Compliance

  • Added Darwin targets for AArch64 to the pip packaging matrix.
  • Refreshed SBOM data and adopted usage of REUSE.toml.

Supported Platforms

The following platform combinations are supported:

  • Linux - AArch64 and x86-64
  • Windows® - x86-64
  • Darwin - AArch64 via MoltenVK (experimental)
  • Android™ - AArch64 (experimental)

Version 0.7.0 – Initial Public Release

Purpose

Executes ML workloads for functional validation and performance exploration.

Features

Execution

  • Workload Execution: Runs shader and neural network graph workloads through both Vulkan® core compute and Vulkan® ML extensions for comprehensive ML workload testing
  • Complete Workflow: Loads input stimuli, executes computational graphs, and writes output artifacts - providing end-to-end scenario execution
  • Validation & Benchmarking: Serves as a validation vehicle for drivers and devices, plus performance exploration and benchmarking capabilities

Test description

  • Declarative Scenarios: Test cases and scenarios defined in JSON format for easy configuration, repeatability, and programmatic test generation
  • Several types of resources: Supports tensors, images, buffers, etc. so scenarios can exercise varied input/output.

Platform Support

The following platform combinations are supported:

  • Linux - X86-64
  • Windows® - X86-64