WIP: docs(tested): RX 7700 XT on CachyOS, Hookshot v2.2.0-rc.1 and Longshot v3.0.0-rc.3 in Kokiri Forest #240

Draft
mynameismawky wants to merge 2 commits from mynameismawky/Longshot:docs/tested-config-rx7700xt into main
First-time contributor

Why

Linux on AMD has the fewest measured runs in tested configurations. This adds two runs on an RX 7700 XT (RADV), the
first on this card: Hookshot v2.2.0-rc.1 and Longshot v3.0.0-rc.3 in the same scene. Longshot ran about 16-18% slower
there, and FSR 3.1 Quality no longer reaches the 60 fps cap.

What changed

  • docs: two rows in docs/tested-configurations.md, one per build. The v2.2.0-rc.1 row was first opened as
    NotRexed/Hookshot#3014 before that repository went away.

Flags and settings

None.

Measurements

These are tested-configuration rows, not a code change: the two builds are compared to each other, not main against
this PR.

Baseline: Hookshot v2.2.0-rc.1 @ 4aee2365
Builds: both the release AppImages; Hookshot run on 2026-11-12, Longshot on 2026-11-13
Hardware: RX 7700 XT 12 GB, Mesa RADV 26.2.6, Core i7-14700KF, 64 GB, CachyOS (kernel 7.3.3), COSMIC Wayland
Scene: Kokiri Forest, standing still; 2560x1440 fullscreen output, the default 60 fps cap, mesh shaders Automatic
(the compute path); upscaler Off, then AMD FSR 3.1 at 100%, then FSR 3.1 Quality (1707x960), in one session
Method: about 30 s or more per setting, one run each; fps from longshot_log.txt / hookshot_log.txt "N presents, X fps"
lines, the slowest frame and VRAM from the "status: 600 frames" lines
How to reproduce: build each release's AppImage, set Output resolution to 2560x1440, stand still in Kokiri Forest and
switch the upscaler from the Insert menu

Kokiri Forest, 2560x1440 Hookshot v2.2.0-rc.1 @ 4aee2365 Longshot v3.0.0-rc.3 @ 0c74253d
Upscaler Off 55.6-56.3 fps 42.5-52.3 fps, median 46.1
FSR 3.1 at 100% 55.2-55.8 fps 45.1-47.0 fps, median 45.7
FSR 3.1 Quality 60.0 fps (the cap) 46.5-52.3 fps, median 51.6
Slowest frame of each 600 21.6-24.5 ms 27.7-40.8 ms
VRAM 7.8-8.1 of 10.8 GB 8.8-10.0 of 10.9 GB
A frame about 2,600 draws, 185-195 passes about 2,800 draws, 206 passes
Uploads while standing still 0 about 2 a frame (0.4 MB)

The comparison is approximate. I stood near the same spot in Kokiri Forest, but not exactly on it (about 2,800
draws a frame against about 2,600). Longshot's build also started from fresh options, and Auto (this PC) put the card in
tier 4; I didn't run Longshot with Hookshot's defaults to separate the two. Each setting ran once, not in interleaved
pairs. The gap holds across all three upscaler settings, so it is in the game's drawing rather than the upscaler.

v3.0.0-rc.3, Off:      vk: 20400 presents, 46.1 fps; a frame: 2804 draws (0 skipped), 138 dispatches, 206 passes, ...
v3.0.0-rc.3, Quality:  vk: fsr3 on the chain: 4800 frames drawn, 1707x960 -> 2560x1440, ...
                       vk: 31200 presents, 51.7 fps; a frame: 2818 draws (0 skipped), 138 dispatches, 206 passes, ...
v3.0.0-rc.3, Off:      vk: status: 600 frames, the slowest 31.5 ms; VRAM 9.0 of 10.9 GB

Testing

Tested: both builds in the game on this PC, as above: past the opening cutscene on v2.2.0-rc.1, and a loaded save in
Kokiri Forest on v3.0.0-rc.3. No device lost on either; 10-14 programs failed in play on v2.2.0-rc.1 and 10-11 on
v3.0.0-rc.3, with nothing visibly wrong. python -I -B tests/run_all.py: 457 passed, 2 failed, 66 skipped; the two
failures (test_display_pointer.py, test_bake_portability.py) fail on main 1ef93e79 without this change too.

Not tested: where the slowdown comes from (Longshot's code or the Auto (this PC) tier 4 settings), and the same
exact position on both builds.

Needs other testers:

  • Another RDNA 2/3 card on RADV: Kokiri Forest, standing still, v2.2.0-rc.1 against v3.0.0-rc.3 at the same spot,
    expecting the same ~16% drop if it is not specific to this PC (request to open)

Screenshots

Nothing on screen: documentation only.

Risk and rollback

None: documentation only.

Reviewer notes

The second row is the one worth a look: a slowdown on Linux AMD between the two releases, though only measured
approximately.

Issues

See NotRexed/Hookshot#3014 (the earlier PR for the first row, gone with that repository).

Checklist

For everyone

  • One topic.
  • The title is a Conventional Commit (type(scope): description).
  • Rebased on current main.
  • python -I -B tests/run_all.py run and passing (any failure that also fails on main is named above).
  • Tested in the game, or what wasn't is named under Testing.
  • Testing filled in: what was tested, what wasn't, and what needs other testers (with test requests).
  • Measured: main and this PR at named commits, on the same PC, scene and settings, at least 2 interleaved runs each, the median reported. (Not a code change: two releases, one run each, positions not exactly the same; see Measurements.)
  • No game files, dumps, build output or generated files.
  • third_party/ untouched (a new folder with a PROVENANCE.md instead).
  • The license is OK: GPL-3.0-or-later, nothing incompatible and nothing decompiled.

If applicable

  • AI assistance recorded with Assisted-by: / Co-Authored-By: trailers.
  • A row in docs/tested-configurations.md for each measured run.
  • Docs updated.

🤖 Generated with Claude Code

## Why Linux on AMD has the fewest measured runs in tested configurations. This adds two runs on an RX 7700 XT (RADV), the first on this card: Hookshot v2.2.0-rc.1 and Longshot v3.0.0-rc.3 in the same scene. Longshot ran about 16-18% slower there, and FSR 3.1 Quality no longer reaches the 60 fps cap. ## What changed - docs: two rows in `docs/tested-configurations.md`, one per build. The v2.2.0-rc.1 row was first opened as NotRexed/Hookshot#3014 before that repository went away. ## Flags and settings None. ## Measurements These are tested-configuration rows, not a code change: the two builds are compared to each other, not main against this PR. Baseline: Hookshot v2.2.0-rc.1 @ 4aee2365 Builds: both the release AppImages; Hookshot run on 2026-11-12, Longshot on 2026-11-13 Hardware: RX 7700 XT 12 GB, Mesa RADV 26.2.6, Core i7-14700KF, 64 GB, CachyOS (kernel 7.3.3), COSMIC Wayland Scene: Kokiri Forest, standing still; 2560x1440 fullscreen output, the default 60 fps cap, mesh shaders Automatic (the compute path); upscaler Off, then AMD FSR 3.1 at 100%, then FSR 3.1 Quality (1707x960), in one session Method: about 30 s or more per setting, one run each; fps from `longshot_log.txt` / `hookshot_log.txt` "N presents, X fps" lines, the slowest frame and VRAM from the "status: 600 frames" lines How to reproduce: build each release's AppImage, set Output resolution to 2560x1440, stand still in Kokiri Forest and switch the upscaler from the Insert menu | Kokiri Forest, 2560x1440 | Hookshot v2.2.0-rc.1 @ 4aee2365 | Longshot v3.0.0-rc.3 @ 0c74253d | |---|---|---| | Upscaler Off | 55.6-56.3 fps | 42.5-52.3 fps, median 46.1 | | FSR 3.1 at 100% | 55.2-55.8 fps | 45.1-47.0 fps, median 45.7 | | FSR 3.1 Quality | 60.0 fps (the cap) | 46.5-52.3 fps, median 51.6 | | Slowest frame of each 600 | 21.6-24.5 ms | 27.7-40.8 ms | | VRAM | 7.8-8.1 of 10.8 GB | 8.8-10.0 of 10.9 GB | | A frame | about 2,600 draws, 185-195 passes | about 2,800 draws, 206 passes | | Uploads while standing still | 0 | about 2 a frame (0.4 MB) | **The comparison is approximate.** I stood near the same spot in Kokiri Forest, but not exactly on it (about 2,800 draws a frame against about 2,600). Longshot's build also started from fresh options, and Auto (this PC) put the card in tier 4; I didn't run Longshot with Hookshot's defaults to separate the two. Each setting ran once, not in interleaved pairs. The gap holds across all three upscaler settings, so it is in the game's drawing rather than the upscaler. ``` v3.0.0-rc.3, Off: vk: 20400 presents, 46.1 fps; a frame: 2804 draws (0 skipped), 138 dispatches, 206 passes, ... v3.0.0-rc.3, Quality: vk: fsr3 on the chain: 4800 frames drawn, 1707x960 -> 2560x1440, ... vk: 31200 presents, 51.7 fps; a frame: 2818 draws (0 skipped), 138 dispatches, 206 passes, ... v3.0.0-rc.3, Off: vk: status: 600 frames, the slowest 31.5 ms; VRAM 9.0 of 10.9 GB ``` ## Testing **Tested:** both builds in the game on this PC, as above: past the opening cutscene on v2.2.0-rc.1, and a loaded save in Kokiri Forest on v3.0.0-rc.3. No device lost on either; 10-14 programs failed in play on v2.2.0-rc.1 and 10-11 on v3.0.0-rc.3, with nothing visibly wrong. `python -I -B tests/run_all.py`: 457 passed, 2 failed, 66 skipped; the two failures (`test_display_pointer.py`, `test_bake_portability.py`) fail on `main` 1ef93e79 without this change too. **Not tested:** where the slowdown comes from (Longshot's code or the Auto (this PC) tier 4 settings), and the same exact position on both builds. **Needs other testers:** - Another RDNA 2/3 card on RADV: Kokiri Forest, standing still, v2.2.0-rc.1 against v3.0.0-rc.3 at the same spot, expecting the same ~16% drop if it is not specific to this PC (request to open) ## Screenshots Nothing on screen: documentation only. ## Risk and rollback None: documentation only. ## Reviewer notes The second row is the one worth a look: a slowdown on Linux AMD between the two releases, though only measured approximately. ## Issues See NotRexed/Hookshot#3014 (the earlier PR for the first row, gone with that repository). ## Checklist ### For everyone - [x] One topic. - [x] The title is a Conventional Commit (`type(scope): description`). - [x] Rebased on current `main`. - [x] `python -I -B tests/run_all.py` run and passing (any failure that also fails on `main` is named above). - [x] Tested in the game, or what wasn't is named under Testing. - [x] Testing filled in: what was tested, what wasn't, and what needs other testers (with test requests). - [ ] Measured: main and this PR at named commits, on the same PC, scene and settings, at least 2 interleaved runs each, the median reported. (Not a code change: two releases, one run each, positions not exactly the same; see Measurements.) - [x] No game files, dumps, build output or generated files. - [x] `third_party/` untouched (a new folder with a `PROVENANCE.md` instead). - [x] The license is OK: GPL-3.0-or-later, nothing incompatible and nothing decompiled. ### If applicable - [x] AI assistance recorded with `Assisted-by:` / `Co-Authored-By:` trailers. - [x] A row in `docs/tested-configurations.md` for each measured run. - [x] Docs updated. 🤖 Generated with [Claude Code](https://claude.com/claude-code)
First run on this card: native 1440p sits just under the 60 fps cap,
FSR 3.1 Quality holds it.

Co-Authored-By: Claude Opus 5.5 <[email protected]>
docs(tested): RX 7700 XT on Longshot v3.0.0-rc.3 in Kokiri Forest
Some checks are pending
CI / Conventional Commit title (pull_request) Blocked by required conditions
CI / Python source checks (ubuntu-24.04, part 1/2) (pull_request) Blocked by required conditions
CI / Python source checks (ubuntu-24.04, part 2/2) (pull_request) Blocked by required conditions
CI / SPIR-V fixtures (global narrow load/store, direct surface atomics, F2I immediate) (pull_request) Blocked by required conditions
CI / Workflow lint (actionlint) (pull_request) Blocked by required conditions
Third-party parts / As their notes say (pull_request) Blocked by required conditions
Third-party parts / The same as upstream (pull_request) Blocked by required conditions
Website / Website on Cloudflare (pull_request_target) Has been skipped
PR description / Description follows the template (pull_request_target) Successful in 5s
586907b11d
The same PC and scene as the v2.2.0-rc.1 row, rebuilt with Longshot:
about 46 fps native 1440p against 56, and FSR 3.1 Quality no longer
reaches the 60 fps cap. The position was near but not exactly the
same, and Auto (this PC) is new, so the comparison is approximate.

Co-Authored-By: Claude Opus 5.5 <[email protected]>
mynameismawky changed title from WIP: docs(tested): RX 7700 XT on CachyOS, Hookshot v2.2.0-rc.1 and Longshot v3.0.0-rc.3 in Kokiri Forest to docs(tested): RX 7700 XT on CachyOS, Hookshot v2.2.0-rc.1 and Longshot v3.0.0-rc.3 in Kokiri Forest 2026-11-13 16:46:55 +00:00
Owner

This conflicts with main in docs/tested-configurations.md: the table is frozen now (#96, and later), and new runs go in one file each under docs/tested/ (format in docs/tested/README.md), so the two rows need to become run files. Please rebase onto current main.

This conflicts with `main` in `docs/tested-configurations.md`: the table is frozen now (https://git.magicbean.dev/MagicBeanDev/Longshot/pulls/96, and later), and new runs go in one file each under `docs/tested/` (format in `docs/tested/README.md`), so the two rows need to become run files. Please rebase onto current `main`.
Owner

Thanks for the RX 7700 XT run, the numbers are worth keeping. docs/tested-configurations.md is frozen since #96, so new rows can't go in there. Could you move this run into its own file under docs/tested/ (the format is in docs/tested-configurations.md) and push that here instead?

Thanks for the RX 7700 XT run, the numbers are worth keeping. `docs/tested-configurations.md` is frozen since #96, so new rows can't go in there. Could you move this run into its own file under `docs/tested/` (the format is in `docs/tested-configurations.md`) and push that here instead?
Owner

Brought up to the ready-for-review standard (#1454): this description doesn't use the pull request template, which pr ready and the description check require. Please fill it in, including the checklist's new Tested on my own PC box: where you ran the change yourself and what you saw. This goes back to WIP until then; pr ready it again once it passes.

Brought up to the [ready-for-review standard](https://git.magicbean.dev/MagicBeanDev/Longshot/src/branch/main/CONTRIBUTING.md#ready-for-review-out-of-wip) (#1454): this description doesn't use the [pull request template](https://git.magicbean.dev/MagicBeanDev/Longshot/src/branch/main/.forgejo/pull_request_template.md), which `pr ready` and the description check require. Please fill it in, including the checklist's new *Tested on my own PC* box: where you ran the change yourself and what you saw. This goes back to WIP until then; `pr ready` it again once it passes.
HumanBean changed title from docs(tested): RX 7700 XT on CachyOS, Hookshot v2.2.0-rc.1 and Longshot v3.0.0-rc.3 in Kokiri Forest to WIP: docs(tested): RX 7700 XT on CachyOS, Hookshot v2.2.0-rc.1 and Longshot v3.0.0-rc.3 in Kokiri Forest 2026-11-17 20:27:14 +00:00
Some checks are pending
CI / Conventional Commit title (pull_request) Blocked by required conditions
CI / Python source checks (ubuntu-24.04, part 1/2) (pull_request) Blocked by required conditions
CI / Python source checks (ubuntu-24.04, part 2/2) (pull_request) Blocked by required conditions
CI / SPIR-V fixtures (global narrow load/store, direct surface atomics, F2I immediate) (pull_request) Blocked by required conditions
CI / Workflow lint (actionlint) (pull_request) Blocked by required conditions
Third-party parts / As their notes say (pull_request) Blocked by required conditions
Third-party parts / The same as upstream (pull_request) Blocked by required conditions
Website / Website on Cloudflare (pull_request_target) Has been skipped
PR description / Description follows the template (pull_request_target) Successful in 5s
This pull request has changes conflicting with the target branch.
  • docs/tested-configurations.md
Some workflows are waiting to be reviewed.
View command line instructions

Manual merge helper

Use this merge commit message when completing the merge manually.

Checkout

From your project repository, check out a new branch and test the changes.
git fetch -u 'docs/tested-config-rx7700xt:mynameismawky-docs/tested-config-rx7700xt'
git switch 'mynameismawky-docs/tested-config-rx7700xt'
Sign in to join this conversation.
No reviewers
No milestone
No project
No assignees
2 participants
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set

Reference
MagicBeanDev/Longshot!240
No description provided.