Windows/Vulkan plugin: compute on iGPU, render on dGPU. Streams to RenderTextures with fallbacks + metrics. C# wrapper & samples for faster post-FX, upscaling (Built-in/URP/HDRP). Source included.The plugin is render-pipeline agnostic. It writes to a Unity RenderTexture via the native Vulkan interface, so it works with Built-in, URP, and HDRP as long as Windows 64-bit + Vulkan is enabled. Not compatible with DirectX/Metal/OpenGL.Not required for use, but if you want to build the native plugin from source: Visual Studio 2019/2022, CMake, and the Vulkan SDK.DescriptionMulti-GPU Direct Transfer for Unity is a native Windows/Vulkan plugin that splits workloads across GPUs: run compute on your iGPU, render on your dGPU, and stream results straight into Unity RenderTextures. It auto-selects the fastest transfer path (shared memory → staging → CPU), exposes metrics/logging, and ships with a clean C# wrapper, SPIR-V samples, and an example scene.Customizable? Yes.Source included—extend or rebuild the native plugin.Swap in your own SPIR-V compute shaders.Force/choose transfer methods programmatically.Works with Built-in, URP, HDRP (Vulkan) because it targets textures, not a specific render pipeline.Suitable for a particular genre?It’s genre-agnostic. Ideal wherever GPU compute meets rendering: post-FX, upscaling/super-res, stylized filters, simulation/AI preprocessing, VFX, imaging tools, and editor utilities.RequirementsWindows 64-bit, Vulkan graphics API enabled, ideally a system with both an iGPU (compute) and a dGPU (render). Falls back gracefully when direct sharing isn’t available.Technical Details — Key FeaturesHeterogeneous GPU pipeline: run compute on iGPU, render on dGPU.Direct RenderTexture streaming (Vulkan): native plugin writes straight to Unity textures.Automatic fallbacks: Shared Memory → Staging Buffer → CPU copy, selected at runtime.SPIR-V loader: load your own compute shaders (.spv) and swap kernels freely.C# wrapper & example scene: P/Invoke API + ready-to-run demo for quick integration.Metrics & logging: per-stage timings, throughput, method stats, error/log queue.Buffer API: create/reuse shared GPU buffers sized for your workloads.Pipeline-agnostic: works with Built-in, URP, HDRP (when Vulkan is enabled).Thread-safe lifecycle: clean init/shutdown; per-frame Update pump for jobs.No Asset Store dependencies: plug-and-play on Windows x86_64.Source code included: extend, debug, or rebuild the native plugin.Requirements: Windows 64-bit, Vulkan graphics API, ideally both iGPU + dGPU available.AI Usage StatementI used AI (ChatGPT) as a productivity assistant during development. Specifically:Drafted and refined documentation (summary, description, release notes, tech details), later edited by me.Reviewed C++/C# structure and naming; suggested cleaner APIs and logging/error patterns which I validated.Helped scaffold a minimal C# wrapper and example usage; I implemented, tested, and adjusted to fit the plugin.Suggested refactors for initialization/shutdown flow and thread-safety checks; final code decisions are mine.Generated boilerplate (e.g., copyright header) and checklists for Asset Store submission.What’s NOT in the package:No AI runtime dependencies; the plugin contains only native code and C#.No AI-generated art/audio/content included.No data collection/telemetry and no model training on my code.All AI outputs were reviewed, edited, and tested by me; I take responsibility for the final implementation.
