Idea: system performance readings alongside the frame counter, with optional file output #92

New issue

Open

opened 2026-11-13 02:27:28 +00:00 by mooki · 0 comments

mooki commented

2026-11-13 02:27:28 +00:00

The idea

link2929 asked for monitoring beyond FPS: GPU use and temperature, CPU and RAM. I took this on with a three-part plan: a system view, choices for testers and sampled output, without adding work to existing counter levels when off.

Longshot already has the Dev renderer-counter view. Keep it; this request is for the machine's side:

GPU use and temperature when the selected GPU's driver supplies them, and video-memory use against its allowance.
CPU use for the game and its busiest thread; the game's memory and what the PC has available.
Selectable readings and an optional local file at a chosen interval, so test results and problem reports can name the same sampled values.
Off by default; missing or unsupported readings stay unavailable, not zero. Measure enabled and disabled cost before claiming overhead.

A local implementation and earlier Linux NVIDIA/Intel observations exist and are being moved. That first change implements sampled logging, not the requested counter/UI view; the system view remains outstanding. Neither part is a merged Longshot feature or Windows/AMD validation. Earlier tests found and corrected a peripheral battery being mistaken for system power; GPU-utilization comparison outliers remain an accuracy limitation rather than a pass.

Pull request 65 separately proposes periodic renderer/runtime memory and cache counters. It does not provide the requested CPU/GPU-utilization/temperature system view. Coordinate sampling and formats rather than collecting the same counters twice. Issue 44 concerns local problem-report bundles, which could consume reviewed samples later but are not this monitor.

Anything else

The request names no hardware or OS, and the question asking which readings matter most has no reply in the checked thread. Linux on an RTX 4090 and Intel UHD 770 is the available test coverage. Windows, AMD, unsupported counters, sampling overhead and accuracy still need explicit evidence. Output stays local; this is not an upload or telemetry service.

Reported on GitHub as NotRexed/Hookshot#2915; taken on there by mooki.

### The idea [link2929 asked](https://github.com/NotRexed/Hookshot/issues/2915) for monitoring beyond FPS: GPU use and temperature, CPU and RAM. I [took this on with a three-part plan](https://github.com/NotRexed/Hookshot/issues/2915#issuecomment-6834139903): a system view, choices for testers and sampled output, without adding work to existing counter levels when off. Longshot already has [the Dev renderer-counter view](https://git.magicbean.dev/MagicBeanDev/Longshot/pulls/8). Keep it; this request is for the machine's side: - GPU use and temperature when the selected GPU's driver supplies them, and video-memory use against its allowance. - CPU use for the game and its busiest thread; the game's memory and what the PC has available. - Selectable readings and an optional local file at a chosen interval, so test results and problem reports can name the same sampled values. - Off by default; missing or unsupported readings stay unavailable, not zero. Measure enabled and disabled cost before claiming overhead. A local implementation and earlier Linux NVIDIA/Intel observations exist and are being moved. **That first change implements sampled logging, not the requested counter/UI view; the system view remains outstanding.** Neither part is a merged Longshot feature or Windows/AMD validation. Earlier tests found and corrected a peripheral battery being mistaken for system power; GPU-utilization comparison outliers remain an accuracy limitation rather than a pass. [Pull request 65](https://git.magicbean.dev/MagicBeanDev/Longshot/pulls/65) separately proposes periodic renderer/runtime memory and cache counters. It does not provide the requested CPU/GPU-utilization/temperature system view. Coordinate sampling and formats rather than collecting the same counters twice. [Issue 44](https://git.magicbean.dev/MagicBeanDev/Longshot/issues/44) concerns local problem-report bundles, which could consume reviewed samples later but are not this monitor. ### Anything else The request names no hardware or OS, and the question asking which readings matter most has no reply in the checked thread. Linux on an RTX 4090 and Intel UHD 770 is the available test coverage. Windows, AMD, unsupported counters, sampling overhead and accuracy still need explicit evidence. Output stays local; this is not an upload or telemetry service. Reported on GitHub as NotRexed/Hookshot#2915; taken on there by mooki.