release: tag-triggered binaries and a README that states the measured rates #28

grenade · 2026-09-13T22:14:04Z

grenade commented

2026-09-13 22:14:04 +00:00

For independent miners (quantus/miner#27). The repository is public but had no on-ramp: no releases, and a README that still opened with upstream's build instructions.

release.yaml: pushing a v<version>-lair.<n> tag builds quantus-miner and quantus-bench on the cuda-13.0 runner (Fedora 43, nvcc, so the binary carries sm_86/sm_89/sm_120 cubins plus PTX) with the tag's commit in --version, and publishes a Gitea release with the tarball, its sha256 and a BUILD.txt (commit, toolchain, glibc). Excluded from deploy triggers; a release changes nothing on the fleet.
README: a "This fork" section up front with the measured rates per card at the power caps they were measured at, against upstream's --cuda-gpu engine and the two closed pool binaries; where the binaries are and what they need; the exact quantus-bench commands for the hashrate and the 2000-job parity check; the interleaved-rounds method; and the statement that there is no dev fee, licence server or telemetry.
Carries the LAIR_MINBLOCKS knob commit (c1cac91, no default change) so the redeploy this merge triggers is not for nothing.

First tag after merge: v4.0.2-lair.1.

🤖 Generated with Claude Code

https://claude.ai/code/session_01Ue5ZZm1Hiv5zPnucykKKuF

For independent miners (quantus/miner#27). The repository is public but had no on-ramp: no releases, and a README that still opened with upstream's build instructions. - **`release.yaml`**: pushing a `v<version>-lair.<n>` tag builds `quantus-miner` and `quantus-bench` on the cuda-13.0 runner (Fedora 43, nvcc, so the binary carries sm_86/sm_89/sm_120 cubins plus PTX) with the tag's commit in `--version`, and publishes a Gitea release with the tarball, its sha256 and a BUILD.txt (commit, toolchain, glibc). Excluded from deploy triggers; a release changes nothing on the fleet. - **README**: a "This fork" section up front with the measured rates per card at the power caps they were measured at, against upstream's `--cuda-gpu` engine and the two closed pool binaries; where the binaries are and what they need; the exact `quantus-bench` commands for the hashrate and the 2000-job parity check; the interleaved-rounds method; and the statement that there is no dev fee, licence server or telemetry. - Carries the `LAIR_MINBLOCKS` knob commit (c1cac91, no default change) so the redeploy this merge triggers is not for nothing. First tag after merge: `v4.0.2-lair.1`. 🤖 Generated with [Claude Code](https://claude.com/claude-code) https://claude.ai/code/session_01Ue5ZZm1Hiv5zPnucykKKuF

grenade added 2 commits 2026-09-13 22:14:06 +00:00

engine-cuda: LAIR_MINBLOCKS launch-bounds knob, with the occupancy results c1cac91419

No default change. Records on the knob what forcing the register budget
measured on beast against the quanpool kernel (quantus/miner#27): 64
registers by any launch shape spills and loses 5%, 80 is neutral.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Ue5ZZm1Hiv5zPnucykKKuF

release: tag-triggered binaries, README section with measured rates

ci / fmt (pull_request) Successful in 20s

Details

bench / build (pull_request) Successful in 1m15s

Details

ci / clippy (pull_request) Successful in 1m43s

Details

ci / doc (pull_request) Successful in 2m0s

Details

bench / measure (pull_request) Successful in 2m52s

Details

ci / test (pull_request) Successful in 6m55s

Details

7831323362

For independent miners (quantus/miner#27): a `v<version>-lair.<n>` tag
builds quantus-miner and quantus-bench on the cuda-13.0 runner with the
tag's commit in --version, and publishes a Gitea release with the
tarball, its sha256 and a build report. The README now leads with what
this fork is, the measured rates per card against upstream and the two
closed pool binaries, where the binaries are, how to reproduce the
numbers with quantus-bench, and that there is no fee, licence server or
telemetry. release.yaml is excluded from deploy triggers.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Ue5ZZm1Hiv5zPnucykKKuF

gitea-actionsbot commented

2026-09-13 22:18:53 +00:00

quantus-bench `7831323362` on benjy.hanzalova.internal

worker	median MH/s	spread	windows
0	708.54	0.2%	707.1, 708.9, 708.6, 707.9, 708.5

total median: 708.54 MH/s (batch 1000000, 5 x 30s, engine gpu-cuda)

gpu 0 NVIDIA GeForce RTX 4090: driver 580.173.02, limit 250 W, draw 250 W, sm 1920 MHz, mem 810 MHz, 60 C
parity: OK (25/25 jobs found solutions, all verified against CPU)
--- resume miner ---
active

## quantus-bench 783132336259d54ea0706be0aa4171fd29d693f3 on benjy.hanzalova.internal | worker | median MH/s | spread | windows | | --- | --- | --- | --- | | 0 | 708.54 | 0.2% | 707.1, 708.9, 708.6, 707.9, 708.5 | total median: **708.54 MH/s** (batch 1000000, 5 x 30s, engine gpu-cuda) - gpu 0 NVIDIA GeForce RTX 4090: driver 580.173.02, limit 250 W, draw 250 W, sm 1920 MHz, mem 810 MHz, 60 C - parity: OK (25/25 jobs found solutions, all verified against CPU) --- resume miner --- active

grenade merged commit 8b01387e97 into main

2026-09-13 22:24:07 +00:00

grenade referenced this issue from a commit

2026-09-13 22:24:08 +00:00

Merge pull request 'release: tag-triggered binaries and a README that states the measured rates' (#28) from release/binaries-and-readme into main

grenade deleted branch release/binaries-and-readme

2026-09-13 22:24:09 +00:00

grenade referenced this pull request

2026-09-14 12:28:28 +00:00

CI: replace inherited workflows with a Gitea fat-binary build and deploy-on-main #8

grenade referenced this pull request

2026-09-14 12:28:53 +00:00