minigpu_platform_interface 1.9.0

The platform interface for minigpu. A library that brings multiplatform GPU compute to Dart and Flutter using WebGPU and Google Dawn.

minigpu_platform_interface CHANGELOG

1.9.0

• Add PlatformSharedOutputTexture.copyFromBufferF32Async. The default runs the blocking copy.
• setLogCallback defaults to warnings and errors (level 2), not info.
• Document that the blocking shared-output copies wait for the GPU, and how toRGBA fails.
• Add YuvMatrix and ExternalVideoBuffer.yuvMatrix / yuvFullRange (defaults: BT.709, full range).
• Add PlatformVideoTexture.toRGBAAsync(). The default calls toRGBA().
• Add PlatformRenderPipeline.settled, which completes when the pipeline stops compiling. The default is null, and callers poll status.
• Add PlatformComputeShader.buildPipeline(), which compiles without running the kernel. The default dispatches one workgroup.

1.8.1

• Add webBufferHandle to PlatformBuffer (default 0), so implementations can expose the underlying WebGPU buffer handle on web.

1.7.0

• Add the shader cache API: configureShaderCache(...), shaderCacheStats, shaderCacheDirectory, clearShaderCache() and the ShaderCacheStats value type. Every member has a default (no-op / null / 0), so existing implementations need no change; web keeps the defaults.
• Add MinigpuPlatform.drainWorkQueue(): blocks until every GPU task already queued on the native worker thread has run. Defaults to a no-op.

1.6.1

• released 08/13/26 - MR

1.6.0

• Version bump for the minigpu 1.6.0 family. No functional change in this package.

1.5.9

• No interface change. For backend implementors: since minigpu 1.5.9 the facade sends every buffer bind through PlatformComputeShader.setBufferFire, never setBuffer. A backend whose setBufferFire is not ordered with its dispatches will corrupt fire-and-forget callers. The default delegates to setBuffer, which is only correct for backends that run everything in order (such as web).

1.5.8

• Every addition has a default, so a backend that extends these classes keeps compiling. A class that implements them must add the new members.
• New GpuAdapterInfo (name, total and used dedicated VRAM, freeVramBytes) and MinigpuPlatform.listAdapters(), default const [].
• New MinigpuPlatform.createSecondaryPlatform(String adapterFilter): an independent context on the matching adapter, with its own device and queue. Never mix two contexts' resources in one dispatch. Default null (no multi-adapter support).
• New MinigpuPlatform.drainSpinBudgetMs: the GPU-wait spin budget of the loaded native library, null where it cannot be asked. A library without that fix costs ~15.6 ms per GPU wait on Windows, so latency-sensitive callers can assert > 0.
• New PlatformComputeShader.dispatchFire(x, y, z): dispatch without waiting. Order with reads and writes is kept, so a later awaited PlatformBuffer.read waits for every fired dispatch. Default delegates to dispatch. Bindings are read when the dispatch runs, not when it is fired.
• New PlatformComputeShader.setBufferFire(tag, buffer): a bind ordered with dispatches, reads and writes, so bind → fire → rebind → fire is safe on one shader. Default delegates to setBuffer, which is correct for backends that run in order.
• New PlatformBuffer.writeRawBytes(bytes, {dstByteOffset = 0}) for large transfers; implementations should stream in bounded chunks. Offset and length must be multiples of 4. The default uses write at offset 0 and throws UnsupportedError for any other offset.

1.5.7

• Version lockstep with minigpu 1.5.7; no interface change.

1.5.6

• New MinigpuPlatform.preferDisplayAdapter([enable]) and MinigpuPlatform.selectedAdapterName, with defaults (false / null) so existing implementations keep compiling.

1.5.5

1.5.4

• fix release version pins

1.5.3

• Add PlatformSharedOutputTexture.copyFromBufferAsync and PlatformVideoTexture.bgraToRgbaSharedOutputAsync (the defaults call the synchronous versions), so backends can run the shared-output copy and present sync off the calling isolate.

1.5.2

• Add buffer copy

1.5.1

• fix frame wait and timeouts

1.5.0

• Fix handle issue, release 1.5.0

1.4.15

• fix fallback paths

1.4.14

• fix Tier C

1.4.12

• fix texture path

1.4.11

• fixing build hook

1.4.9

• tryfix central dawn location

1.4.8

• fixes texture path on windows, fixes texture view on web

1.4.7

• Fixes logger characters

1.4.6

• Fix garbled adapter names in logs.

1.4.5

• Fix Dawn being built inside the pub cache instead of the shared Dawn directory.
• Add Minigpu.setLogCallback / setLogLevel: native Dawn/GPU log lines are routed to a Dart callback.

1.4.4

• Fix a FormatException in setLogCallback when a log line is not valid UTF-8.

1.4.3

• Prefer the discrete GPU when selecting an adapter; fixes the wrong adapter being picked on Optimus laptops.

1.4.2

• fixes dawn library not being found

1.4.1

• added bindings observer

1.4.0

• adds minigpu_view, gpu_pipeline libraries

1.3.0

• Adds Texture Sharing

1.2.4-WIP

• working on texture imports

1.2.3

• adds VRAM API
• fixes memleaks and broken tests

1.2.2

• fixes memleaks and broken tests

1.2.1

• Breaking: Changed setData and references to .write
• Change: 1.2.0 migrates minigpu to direct webgpu usage
• Fix: fresh builds need dawn find off
• Fix: broken compute shader and buffer finalizers

1.2.0

1.1.9

• fix pubspec version issue

1.1.8

• fixed concurrent buffer op crash

1.1.6

• fixed problem with audio input capture providing raw data

1.1.5

• sync with other versions