[gilby]

What this changes

Prompt chunks on Metal and the exact decoder suffix: decoder layer L computes only the rows the prompt's end depends on (kept + (39 - L) * 127), with the same bits as running every row; ced gains the cone's rules and the snapshot replay.

neutral.zig keeps main's engram_manifest export beside ced, and shared_test.zig keeps main's engram_manifest_test.zig and dsq_storage_test.zig imports, so no existing test file stops running.

Part 13 of 18 of the stack for #1676; stacks on dsv41-metal-pr/12-decode-forward (part 12, #1716), on main with parts 1 to 5 (#1705 to #1709) merged.

Receipt

• Environment, with the TensorFold commit, MLX or PyTorch version, machine and GPU, checkpoint and revision: TensorFold 5273393 on main ad98b60 (native Zig 0.17.0; no MLX or PyTorch); Apple M3 Ultra (60-core GPU), 96 GB, macOS, for the kernel and device tests; checkpoint deepseek-ai/DeepSeek-V4.1-Flash-1017.
• Exactness, from tools/bench_concurrent.py --alone --serial or token_sha against "draft": false: n/a for served output; chunk sizes 1, 7, 64 and whole give the same logits, taps and slot; the suffix cone equals every decoder layer on every row; a resumed prompt equals a fresh one.
• Decode speed before and after, from tools/bench_openai.py: measured on the last PR of the stack.
• Prompt speed before and after: the cold first request's ttft_s from tools/bench_cache.py --require-cold on a long prompt, or "n/a" with the reason when the change cannot reach prompt processing: measured on the last PR of the stack.
• Tests run, with pass, fail and skip counts: zig build test tf-kernel-contract -Dcpu=apple_m1: exit 0, 625/625 steps succeeded; kernel contract, tf-kernel-contract --only dsv41 (Metal): SUMMARY 89 kernels: 89 pass gates 1-4, 0 fail, 0 known findings; 186 controls: 186 red as planted, 0 missed; 0 limit problems; gates 119124 ms, limits 162 ms; zig build test-dsv41-metal -Dcpu=apple_m1 (the device suite, its own step): exit 0, 7/7 steps succeeded; 22/22 tests passed; python3 tools/zig/lean_check.py: 5124 files, 424 problems, the same count as main (5027 files, 424). Zig's summary counts only the test binaries it ran in that invocation, so the step count is the comparable figure.
• Not run, and why: CUDA: no NVIDIA device here; Metal on M1, M2, M4 and M5: one M3 Ultra only; the release-layer checks (test-dsv41-metal-real, test-dsv41-metal-vision-real) were run at the stack's top and are in dsv41 metal: image prompts #1722's receipt.

Checklist

• Every token still goes through the lane rounds. No serial path, and nothing that needs drafts off.
• Drafted output equals "draft": false, a resumed prompt equals a fresh one, and concurrent equals solo.
• No precision traded for speed. If the bits change, the description says which sums change and why.
• A new or changed GPU kernel is registered with the kernel contract (docs/kernel-contract.md); each backend's SUMMARY line is in the receipt.
• Prompt processing is no slower than the last release.
• The description says which platforms I ran, Metal M1 to M5 and CUDA, and which I could not.
• New tests fail before the change, pass after it, and skip cleanly without their dependency.
• Comments and docstrings are one line. No measurements or history in the source.
• No personal data, machine names, internal hosts or local paths. No AI attribution lines.

[ashhart]

Thanks @gilby. One change, a rebase artifact we confirmed against main: the diff replaces main's engram_manifest export in deepseek_v41/neutral.zig with ced, and drops the engram_manifest_test.zig and dsq_storage_test.zig imports from shared_test.zig. As diffed, two existing test files would stop running. Please rebase onto current main, keeping the export (add ced beside it) and both imports. Everything else is sound: decoder layer L computes kept + (layers-1-L)*(window-1) rows, matching the sliding-window dependence; kept = 127 keeps every layer's next-step KV rows; and the host cone oracle (one row fewer fails) is a good planted control.

[gilby]

Thanks @ashhart, you were right. Rebased onto current main with neutral.zig keeping engram_manifest beside ced (the export now lands in this part instead of part 16, which loses the line it was carrying) and shared_test.zig keeping both engram_manifest_test.zig and dsq_storage_test.zig. Against main, the diff of those two files is additions only, so no existing test file stops running. The receipt in the body is at this head (5273393e, on main ad98b602).

[ashhart]

Thanks @gilby. This part landed on main with #1719's merge (cfcbecb), which carries #1712-#1719 as your own commits in order; closing it as included. The same Metal contract, full test and DeepSeek device suite cleared the whole stack.