Repository navigation
Switch to the shared Sdcb.SimdOnnx engine - #4
Merged
Merged
Conversation
sdcb
force-pushed
the
feature/simdonnx
branch
from
October 14, 2026 11:12
844f8b1 to
bcc0436
Compare
…xtensions (2.0.0) Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
…nfig); README engine note
…e NOFUSE / shared-device ConvDot toggling (fixes flaky Vulkan byte-exact tests), opt-in device-loss test
sdcb
force-pushed
the
feature/simdonnx
branch
from
October 14, 2026 11:55
bcc0436 to
a2a0bc9
Compare
…tile-read slack included); fused pipeline opts into the staged SG16 tensor-core convs (CompileOptions.StagedCoopConv); CI builds against sdcb/SimdOnnx main from source Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…extensions, PixelOps, fused pipelines) and points to Sdcb.SimdOnnx for the engine; migration note for the engine types that moved to Sdcb.SimdOnnx.*; package description; rk3399 doc links the NEON kernels in SimdOnnx Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…s and packs against the pinned package (what ships: the nupkg now depends on exactly [0.1.0-preview.4]) and keeps one job on sdcb/SimdOnnx main from source; NU5104 accepted while the engine is a preview Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…org; wasm doc links the PackedSimd kernels in SimdOnnx Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
sdcb
marked this pull request as ready for review
October 16, 2026 11:20
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Switches the inference engine to the shared Sdcb.SimdOnnx (pinned
0.1.0-preview.3, ProGet test feed). The engine grew out of this repo, so the in-repo interpreter, kernels and Vulkan/Metal backends move there. The GCDR-specific glue stays here: pixel shaders (bgr2nhwc/div6/nhwc2bgr), the fused GPU pipelines andPixelOps.Performance (interleaved A/B, Ryzen 9 5950X + Arc B580,
Sdcb.SimdGcdrNet.Benchmarke2e)Tests: 74 passed and 30 skipped (device-specific: AVX-512, NEON, Metal); they also pass against the published package. Script:
SimdOnnx/tools/ab-gcdr.ps1.Before merge (TODO)
simdonnx-previewsource fromnuget.config, and bump the pin.NoFuse_MatchesFusedsetGCDRNET_NOFUSEfor the whole process, and the ConvDot test flippedConvDoton the shared device. Both are fixed: the test now usesCompileOptions.Fuse, and the ConvDot test runs on a private device. Result: 0 failures in 35 parallel full-suite runs.Enhanceretries the image once on a fresh device. The test is opt-in:SIMDONNX_TEST_DEVICELOSS=1.testworkflow.