Close the stable ABI
Independent review and a practical command-delivery decision remain before freeze.
M5 is complete and M6 is active: accelerator nodes 0–217 and the explicit host tail pass tested inputs. Independent review and a command-delivery decision remain before a stable ABI; final corpus release follows review.
Independent review and a practical command-delivery decision remain before freeze.
Validate independent fixture derivations and tie the final corpus to the frozen ABI.
Build a faster command-equivalent runner, recheck fixtures and six inputs, then evaluate all COCO val2017 images.
Working software →M6 proceeds against the explicitly pinned M3/M4 candidate. RTL starts after independent review and stable contract freeze; a complete v0 additionally requires the full workload execution path.
Structure, build entry points, documentation, CI skeleton, and licensing are tracked.
FP8/INT8 semantics and numerical, layout, activation, and operation tests pass.
Versioned 128-byte command encoding, memory behavior, errors, and simulator tests exist.
ABI 0.1 is a candidate. Stable freeze awaits independent review and a practical command-delivery decision.
46 independent binary fixtures, schema, registry, integrity checks, runner, and CI are implemented.
Artifact, graph, preprocessing, host boundary, FLOAT baseline, INT8 calibration, proxy accuracy, and diagnostics are recorded. Exact target execution belongs to M6.
Accelerator nodes 0–217 and the explicit Python host tail pass synthetic and multi-input comparisons. Next: a faster equivalence-checked runner and COCO-wide command-path accuracy.
Decode one command path through INT8 MAC, accumulation, requantization, completion, errors, and differential tests.
Add DMA, convolution, and utility operations one conformance-gated step at a time.
Select from measured probes and publish reproducible timing, resources, accuracy, transfers, and latency.
Run a separately pinned perception graph on the same measured FPGA bitstream.
Add only features justified by measured v0 and reprogramming evidence.
Resolve process, SRAM, IO, clock, power, DFT, verification, packaging, and fabrication collateral.
These controls apply across software, compiler, RTL, FPGA, and eventual silicon.
Opcodes, fields, layouts, error codes, and ABI versions stay aligned across every consumer.
The golden expectation remains separate from the implementation being tested.
Unsupported operations, host partitions, errors, saturation, and approximations are declared.
Pin revisions, tools, commands, seeds, artifacts, models, and measurement conditions.
Compare tensors layer by layer instead of trusting only a plausible final detection.
Negative cases and illegal states receive the same care as successful execution.