Skip to content

Lindoze v0.2.0

Choose a tag to compare

@awn007-eng awn007-eng released this 08 Jun 14:58
· 22 commits to main since this release

What's new

Polishes the experimental Intel GPU support that landed after v0.1 and adds a debug flag so Intel users can file actionable bug reports without me needing access to their hardware.

Intel GPU

  • Frequency sysfs fallback — older Gen7 (Ivy Bridge HD 4000-era) places gt_cur_freq_mhz directly under cardN/, not cardN/device/. We try both now, so freq reads work across kernel generations.
  • --dump-gpu flag — prints raw PMU file descriptors, counter values, sysfs paths, and event configs to stderr. If your Intel GPU is detected but util/temp/freq look wrong, paste the output into a GitHub issue.
  • PMU parser unit tests — 15 tests with synthetic counter fixtures exercise the engine-busy math without needing real Intel hardware.

Known thin spots

  • Ivy Bridge (Gen7): detection + freq work; PMU util may still report 0. Awaiting a user report with --dump-gpu output.
  • xe driver: temp/freq only; Arc-on-xe testing pending.
  • Old AMD radeon driver: no gpu_busy_percent sysfs → not detected. Use amdgpu if available.

Unchanged from v0.1

Processes / Performance (CPU / Mem / NVIDIA+AMD / Disk / Network) / Startup tabs, plus the perf knobs (1500ms sampler, paused-when-hidden, /proc/loadavg thread count) all carry over.

Install

git clone https://github.com/awn007-eng/lindoze.git
cd lindoze && git checkout v0.2.0
python3 -m venv .venv && .venv/bin/pip install -e .
.venv/bin/lindoze

Vibe-coded with Claude Code. Design and decisions are mine; Claude handled most of the typing.