Intel's compiler turns rotate(x, (0u - n) & 31u) into a rotate LEFT by n: lane 0's register trace on the B580 diverged
at instruction 6 of iteration 0 (rotr) and nowhere before, in both exchange modes, with every other family and the
dataset kernels bit-exact. proto-opencl/intel_rotr.h rewrites the one helper line when the device's vendor or
platform string holds Intel (host.c's buildProgram and the prepare path), no other vendor sees a change, no pack or
consensus text moves. proto-opencl/test_intel_rotr.c (the pre-push gate runs it) feeds the line through the rewrite
under Intel, AMD and NVIDIA strings and asserts the outputs.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Every instruction family of the live hash is bit-exact on Intel's compiler (the family probe, runs a and b) and the hash
is wrong on every lane in both exchange modes: the trace of lane 0's registers after every instruction, GPU against
the CPU interpreter, names the first diverging instruction.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
After restart b the three workers sat faulted on "no status line from the miner for 90 s" (the node still syncing).
The chain: a read-only wait for node.synced and the exec tip (api/state and igneum_getExecStatus, 10 s, cap 10 min),
the signed restart --what miners kind (the app's own path for a faulted worker; a script never posts to api/cards),
and a read-only 3-minute watch of state, hash_now and message per card with the log tail of any card still faulted.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Run b (10:46 UK) exited 1014 = the inner installer's 14 wrapped: Level Zero 1.32.0, the Support Assistant and Intel
Graphics Software went on, the 9034 display driver did not bind while the device sat in Code 12 on Windows' inbox
6733. The restart job (the project lead's word) ran at 11:01 UK, a plain run job (shutdown /r needs no elevation); the read-back
after it: the Arc OK, Code 0, the Intel OpenCL 3.0 platform listing it with 11,928 MiB, 160 compute units, 2,850 MHz,
local 128 KiB and cl_khr_subgroup_shuffle; the app's row other:Intel(R) Arc(TM) B580 Graphics, 8 identities. The
read-back's list line used $_ inside a foreach statement (empty) and its first-light filter matched the job's own echo
in the app log: both fixed. The plan's section 1 carries the measured line.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
docs/plans/intel-arc.md: the OpenCL worker runs the B580 first (Intel's OpenCL 3.0 runtime, exchange auto with the
local-memory fallback the worker already carries), the expected ceiling from 456 GB/s (13 to 20 MH/s, approximate until
the memprobe), 12 GB = mine only (the prover is CUDA only), the Arc OpenCL gotchas (sub-group 16 vs 32, SLM, spills,
64-bit multiply, IGC's build options and compile time, two Intel platforms, no power API through OpenCL: IGCL next),
the fallback ladder with hours (SPIR-V into the same runtime, a Level Zero host, SYCL), the consequences per tier.
App: vendor "intel" from the OpenCL vendor string and the Windows name; Arc cards discrete, the Arc-branded processor
graphics integrated; the INTEL badge in the Intel blue token, light and dark; the row's no-reading sentence, the
measure-only tune words and the details panel's no-cap note; sweep.rs names IGCL as the source to come; detect.rs test
on a B580 line beside PC 1's list; view.test.mjs row test; ui-mock scenario intel.
PC 1 jobs (tools/intel-arc): the elevated driver job (ran twice on 7 October: run 1 exit 1007 = the bootstrapper's wrap
of the inner installer's 7 on my -report flag; run b with -s alone exit 1014 = inner 14, installed, restart required),
the restart job (the project lead's word), the read-back job, the packs kit and the bench job with the memprobe, class v4 and v3
benches, the local-exchange row, the bench-log row, the card-picker entry and the owner's consequences line.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
remote-run.sh checkout_tree ran `git clean -fd` on the box mirror before every build from any agent, so the attack rows
lost attack-f3/, attack-f1-venv/ and tools/attack/*/target to each other's builds (7 October 2026, 09:2x UK). The clean now
also spares the fixed prefixes attack-*, scratch-*, target-attack-*, .build-remote.log and every glob in the mirror-local
.igneum-scratch-spare (one per line, # comments, the file itself spared), keeps the target and stamp excludes and still runs
without -x. The clean-tree test asks `git clean -nd` with the same excludes instead of filtering the status list, so a spared
dir is not "not clean". --self-test: a fixed-prefix dir at the root and nested, a declared dir and the spare file survive, an
undeclared dir is removed. tools/ci/scratch-spare-check.sh in the pre-push gate fails when the clean line loses the spare
arguments, spare_args stops reading the file, a fixed prefix goes, or -x appears. docs/plans/build-server.md R4a says how a
lane declares its prefix.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Light theme shows light screenshots on the miner and app pages (the
0.3.16 renders from the RTX 5090 Windows rig's live state, the 390 px
pair included). The X9 never shipped: pre-orders 26 December 2025,
withdrawn mid-May 2026 before any unit shipped, no benchmark; every
public sentence that had it shipping now says so, with k about 0.33
labelled as the claimed, unmeasured core; ledger X36, notes on X34, X35,
M34 and C2. Every public Discord link is https://discord.gg/igneum.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Nav sitewide: Litepaper, Miner, App, Wallet, Live devnet, Ledger and the
Download button; the rest in the sheet and the footer. /miner is the
product page for the installers: the shipped app in its frame, the card
picker with the live time to a block, the downloads with sizes, HiveOS,
the fee in full view, three real shots. /app is new: the Igneum Ember app
page by page (Overview, Cards with Ember, the chain drawn live by the
module the app ships, Earnings, Prove, Settings, phone width, what keeps
it mining), light and dark, every shot a real capture or a render from
the RTX 5090 Windows rig's live state with the hostname painted out.
/live is redrawn to show the system working: the step scene, six live
cells, finality in plain words, miner lanes, provers paid, recent blocks
landing, the light client in one line, the DAG below. site.css carries
the shared product components (frames, download list, lines, your-card
line, phone download bar); yourcard.js reads the network rate. The
ledger and wallet pages no longer overflow at 390 px.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
The owner's word this morning: the 56 px header, the eleven-word
statement, two buttons, the step scene full bleed, three facts, the
downloads and one line to the ledger. Kept from the nights since: the
X31 testnet sentence, the X35 chip range with k stated, the Discord
line. The RandomX correction (X34), G4, C2 and X8 stand on the
litepaper; the ledger notes and the text check say so. The app showcase
moves to the miner page in the next step.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Three classes from the 0.3.17 night. (1) bash reads a script incrementally: tools/build-remote.sh was edited while a four-minute remote
build ran, the running copy continued at shifted bytes and died with a syntax error after the build had succeeded on the box; the
four long-running tools (build-remote, cross-remote, workers-remote, move-hand) now keep their body in one brace block ending in exit,
parsed whole before a line runs; tools/ci/whole-body-check.sh (in the gate, self-test with a block-less copy) holds the shape.
(2) A fork build pairs with the igneum-pow of the igneum worktree it sits in: a fork at 12153428 under a master worktree failed in
kaspa-pow four minutes in (no chain_program_shadow; master's igneum-pow predates release-0.3.17's); build-remote.sh says the
pairing on its first line ('pairs with igneum 6f8d7a7e (detached): igneum-pow 0.2.0') and the JSONL line carries pairs_with.
The first version of that line used '[ -n ... ] && echo' inside an assignment's $( ) and set -e ended the script on the false
status; fixed. (3) move-hand.sh restart <hand> [--digest <hex>] [--go]: after binary installed a release, restart ONE unit and read
it back (first exec line, commit string in the running binary, digest against the wanted one, igneum_getNodeInfo powEngine over the
node's loopback EVM RPC); the digest readers tolerate a missing line.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Main's order of 7 October 2026 after RunPod's Ubuntu 22.04 canaries (glibc 2.35) refused the box's GLIBC_2.38 binaries and HiveOS
turned out Ubuntu 20.04 based (2.31). The table lives once, in tools/ci/glibc-ceiling-check.sh (--class, --ceiling-of; self-test
covers the table, an unknown class and a 2.34 need against hive); tools/build-remote.sh --ship hive|rig|seed|linux (default seed)
and tools/workers-remote.sh --class (default rig) build with zig at the class's glibc and check against it. provision.sh installs
docker.io (user build in the docker group) for the proof. Proof: fork 3bfe346f at class hive, igneumd and igneum-miner need
GLIBC_2.30; the workers at class rig need GLIBC_2.17; all four run in ubuntu:20.04 (ldd 2.31) and ubuntu:22.04 (ldd 2.35) and
print their version or usage lines (the OpenCL worker its own no-libOpenCL message, the binary running in a GPU-less container).
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
The previous commit's 'the workers at 2.36' was not yet true: the first zig build died on __isoc23_strtol because -I /usr/include
for CL/cl.h put Ubuntu's glibc 2.39 stdlib.h in front of zig's bundled 2.36 headers. Fixed; proof on the box (6 s):
igneum-worker-cuda 6,759,272 B sha256 690c8e91..., igneum-worker-opencl 307,264 B sha256 329fb6a9..., each needing GLIBC_2.34
and libc only, the ceiling check passing. The two static flags zig ignores are dropped on the zig path (zig links its libc++ in).
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Main's order of 7 October 2026 after a seed took 14 restarts and three minutes down on a glibc 2.39 binary. provision.sh:
step_zig (zig 0.17.0 from ziglang.org, sha256 from the official download index) and cargo-zigbuild 0.23.4 in the cargo tools.
tools/build-remote.sh --ship [--glibc 2.36]: cargo zigbuild --target x86_64-unknown-linux-gnu.2.36 with zig as the C/C++
toolchain (the Mac's infra/cross/build-linux.sh recipe), artefacts from the target-triple dir, each checked by
tools/ci/glibc-ceiling-check.sh (need at most the ceiling; self-test fires on 2.38 against 2.36, passes 2.34 and 2.36;
--symbols reads a saved objdump -T text so CI needs no ELF tools). tools/workers-remote.sh builds the two GPU workers with
zig at 2.36 by default (GLIBC=native for clang). Proof on the box: fork 3bfe346f igneumd needs GLIBC_2.34 (47,023,120 B,
sha256 345dfb95...), igneum-miner GLIBC_2.34 (9,248,808 B, d09dc27b...), 3 min 17 s cold through zig; the workers at 2.36.
Rule: anything that ships to a seed or a HiveOS rig is built with --ship; a plain build (glibc 2.39) is for the box and
Ubuntu 24.04 hosts only.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
