Skip to content

chore(bcv): regenerate the lang-core API dump missed by #1171 - #1174

Merged
michalharakal merged 3 commits into
developfrom
chore/apidump-bitnet-planes
Aug 26, 2026
Merged

michalharakal merged 3 commits into
developfrom
chore/apidump-bitnet-planes

Conversation

@michalharakal

Copy link
Copy Markdown
Contributor

develop currently fails apiCheck: #1171 merged the BitNet-planes public surface without regenerating skainet-lang-core.api. CI doesn't run apiCheck on PRs, but every local pr-gate.sh and the golden-parity job do — this is also what turned #1172's golden-parity leg red (that branch additionally needs its own dump for its lm_head additions). Purely additive dump regeneration, no code changes.

🤖 Generated with Claude Code

michalharakal and others added 2 commits August 26, 2026 17:24
#1171 merged the BitNet-planes public surface (TernaryCodec
decode/encode/planesRowScale additions, BitNetPlanesTensorData) without
regenerating skainet-lang-core.api, so apiCheck fails on develop —
which CI does not run on PRs, but every local pr-gate and the
golden-parity job do (it is what turned #1172 red). Purely additive
dump regeneration, no code.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
…rands

The JVM vector paths guarded on 'data as? FloatArrayTensorData', so a
slab-backed operand — what a ForwardScope produces since #1145/#1146 —
silently fell back to the common scalar loops: correct, but the JVM
lost SIMD exactly where the scope machinery is in use.

The vector entry points now read dense-FP32 windows (array + base
offset) matching both plain array data and StorageFloatTensorData:
vectorFloatBinary in all four broadcast shapes, vectorFloatUnary
(relu &c.), silu, the reduce-all sum (same left-to-right accumulation
order, offset base), and the FP32 matmul FloatArray path — whose kernel
SPI already took offsets; only BLAS keeps its offset-0 guard, since the
JNI shim takes whole arrays. JvmVectorKernels.binaryFloat/unaryFloat
gain default-0 offset parameters.

SlabOperandVectorPathTest drives every vector-eligible shape with both
operands sliced from a slab at nonzero offsets (length 67 — no species
multiple) and requires bit-identical results vs Ambient: the tightest
guard against off-by-offset reads, which produce plausible garbage
rather than crashes.

Closes #1173.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
…offset-operands

JVM Panama vector kernels accept slab-backed (offset) FP32 operands
@michalharakal
michalharakal merged commit 00996ea into develop Aug 26, 2026
1 of 2 checks passed
@michalharakal
michalharakal deleted the chore/apidump-bitnet-planes branch August 26, 2026 15:46
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant