MagicBeanDev/Hookshot
Archived
Watch
2
0
Fork
You've already forked Hookshot
 0
Code Issues 125 Pull requests 205 Activity

feat(menu): Frame generation by and the FPS counter name who makes the frames #2788

Closed
github-sync wants to merge 4 commits from github/pr/2788 into main
pull from: github/pr/2788
merge into: MagicBeanDev:main
Conversation 1 Commits 4 Files changed 10 +366 -60
github-sync commented 2026-11-12 13:38:15 +00:00
Owner

@polackangry opened this pull request on GitHub · 2026-11-12 13:04 UTC · original
Merged on GitHub in 8296d07c05

With Frame generation by at Auto, nothing on screen said whether NVIDIA DLSS Frame Generation or the built-in interpolation made the frames. It falls back to the built-in one silently: without nvngx_dlssg.dll (v2.1.0 and the current nightly copy only DLSS's runtime), without a temporal Upscaler, in the game's own menus, or at Frame generation's Auto. A player on an RTX 5060 Ti couldn't tell DLSS FG wasn't running until the log said not available (0xbad00004).

What it shows

  • Frame generation by, while frame generation is on: Auto: DLSS FG, Auto: Built-in FG or Auto: FSR 3.1 FG (the FSR 3.1 interpolation at 2x). The choice itself is now Built-in FG. Off: Auto, as before.
  • Its line, when another than DLSS FG makes the frames at Auto: why, for example Not DLSS FG now: DLSS Frame Generation makes a fixed 2x to 8x, not Frame generation's Auto., or no picture from a temporal Upscaler (also the game's menus), nvngx_dlssg.dll / an RTX 40 missing, more frames than its runtime makes, NGX refusing. Every line it can have is a variant, so the help box keeps its height.
  • The FPS counter: its second line was game 60; now Base 60, DLSS FG (or FSR 3.1 FG, Built-in FG). "Base" for the game's own rate, the name frame generation tools use.

One name per engine, " FG", on the row and the counter. When #2665 adds FSR 4 as a frame generator, FSR 4 FG fits the same way (FgMaker and framegen_maker_name).

How

  • vk_framegen.cpp: framegen_maker(). Each framegen_record notes who made its set (DLSS-G's dlss_fg_frames, fsr3fi_frame, the renderer's own passes, or none) in a ring of the last 64 sets; the menu's thread reads the counts and takes the most. Nothing noted for half a second (Off): none.
  • vk_dlss.cpp: dlss_fg_skip(), the reason as a code (FgSkip in vk_dlss.h). Each return of dlss_fg_frames that leaves the set to the renderer's own names its reason (skip(FG_SKIP_...)); a set DLSS-G makes clears it; a set with nothing to blend from (a loading hitch) leaves it as it was. The build without the NGX SDK says so.
  • vk_menu.cpp: tx_framegen_by, hn_framegen_by (the reasons' texts in framegen_skip_text, so they can be translated) and the FPS counter's line.
  • menu_i18n.h: every new text in all 13 languages (Auto: %s, Base %.0f, Base %.0f, %s, Built-in FG, Not DLSS FG now: %s and each reason); Built-in and game %.0f, no longer written, let go. Product names (DLSS FG, FSR 3.1 FG, Upscaler) stay as the tables already keep them.
  • The FPS counter's second line now goes through i18n::tr: it draws with its own Tab, past ui::'s translation, so main's game %.0f was never shown translated. Its buffer is 96 bytes, for UTF-8.
  • readme.md: the row and the counter.
  • Tests: test_dlssg_runtime followed the new returns and now checks that the reasons are named and cleared; present_source_crop_test checks the maker noted for FSR 3.1's set and for the built-in fallback.

Screenshots

3440x1440, Fire Temple. Frame generation 2x: the row and the counter say DLSS FG.

Frame generation by: Auto: DLSS FG, FPS counter Base 63, DLSS FG

Frame generation Auto: the built-in one makes the frames, and the line says why.

Frame generation by: Auto: Built-in FG with the reason, FPS counter Base 71, Built-in FG

In Polish (top, 2x) and Russian (bottom, Frame generation Auto): the counter's line and the row's line translated.

The FPS counter and the row's line in Polish and Russian

Testing

In game, native Windows 11, RTX 5060 Ti 16 GB, driver 617.42, a build of this branch on main (51da046) with NVIDIA's SDK and nvngx_dlssg.dll 310.9.16, DLSS Quality, a save in the Fire Temple. The row's text read through the control bridge (option get frame_generation_by), the screenshots of the window:

Case Row FPS counter
2x, in play Auto: DLSS FG Base 63, DLSS FG, 126-129 FPS
4x, in play Auto: DLSS FG
Frame generation Auto Auto: Built-in FG + why Base 71, Built-in FG, 240 FPS
4x, the game's pause menu Auto: Built-in FG
The title screen Auto: Built-in FG
Built-in chosen, 3x Built-in FG
Off Auto

No fault through 6x, 4x, 2x, Auto and Off. Menu language Polish: Auto: Wbudowany FG, Bazowe 63, DLSS FG, Teraz bez DLSS FG: ...; Russian: База 69, Встроенный FG and its line. The log agrees: of the last 600 presents 600 made by DLSS-G at 2x.

Tests: test_menu_i18n passes (it failed on the first push: the tables still had Built-in and game %.0f). The checks that read vk_menu.cpp, vk_dlss.cpp, vk_framegen.cpp, menu_i18n.h or the readme, run here: all pass but one; test_input_background fails the same on main on this PC (it was in my earlier PRs' list of machine-only failures). A row for the run is in docs/tested-configurations.md.

Not covered: FSR 3.1 FG on screen (no AMD GPU here; its maker is checked in present_source_crop_test), the reasons for an RTX 30 card or a missing runtime on screen (read from the code paths, and the missing runtime from this PC's earlier log). The race when stepping the multiplier up is #2658's, not touched here.

🤖 Generated with Claude Code

> [@polackangry](https://github.com/polackangry) opened this pull request on GitHub · 2026-11-12 13:04 UTC · [original](https://github.com/NotRexed/Hookshot/pull/2788) > **Merged on GitHub** in https://git.magicbean.dev/MagicBeanDev/Hookshot/commit/8296d07c05aaa787fa1e8e8c0dd9f2f1163601a0 With Frame generation by at Auto, nothing on screen said whether NVIDIA DLSS Frame Generation or the built-in interpolation made the frames. It falls back to the built-in one silently: without `nvngx_dlssg.dll` (v2.1.0 and the current nightly copy only DLSS's runtime), without a temporal Upscaler, in the game's own menus, or at Frame generation's Auto. A player on an RTX 5060 Ti couldn't tell DLSS FG wasn't running until the log said `not available (0xbad00004)`. ## What it shows - **Frame generation by**, while frame generation is on: `Auto: DLSS FG`, `Auto: Built-in FG` or `Auto: FSR 3.1 FG` (the FSR 3.1 interpolation at 2x). The choice itself is now `Built-in FG`. Off: `Auto`, as before. - **Its line**, when another than DLSS FG makes the frames at Auto: why, for example `Not DLSS FG now: DLSS Frame Generation makes a fixed 2x to 8x, not Frame generation's Auto.`, or no picture from a temporal Upscaler (also the game's menus), `nvngx_dlssg.dll` / an RTX 40 missing, more frames than its runtime makes, NGX refusing. Every line it can have is a variant, so the help box keeps its height. - **The FPS counter**: its second line was `game 60`; now `Base 60, DLSS FG` (or `FSR 3.1 FG`, `Built-in FG`). "Base" for the game's own rate, the name frame generation tools use. One name per engine, "<engine> FG", on the row and the counter. When #2665 adds FSR 4 as a frame generator, `FSR 4 FG` fits the same way (`FgMaker` and `framegen_maker_name`). ## How - `vk_framegen.cpp`: `framegen_maker()`. Each `framegen_record` notes who made its set (DLSS-G's `dlss_fg_frames`, `fsr3fi_frame`, the renderer's own passes, or none) in a ring of the last 64 sets; the menu's thread reads the counts and takes the most. Nothing noted for half a second (Off): none. - `vk_dlss.cpp`: `dlss_fg_skip()`, the reason as a code (`FgSkip` in `vk_dlss.h`). Each return of `dlss_fg_frames` that leaves the set to the renderer's own names its reason (`skip(FG_SKIP_...)`); a set DLSS-G makes clears it; a set with nothing to blend from (a loading hitch) leaves it as it was. The build without the NGX SDK says so. - `vk_menu.cpp`: `tx_framegen_by`, `hn_framegen_by` (the reasons' texts in `framegen_skip_text`, so they can be translated) and the FPS counter's line. - `menu_i18n.h`: every new text in all 13 languages (`Auto: %s`, `Base %.0f`, `Base %.0f, %s`, `Built-in FG`, `Not DLSS FG now: %s` and each reason); `Built-in` and `game %.0f`, no longer written, let go. Product names (`DLSS FG`, `FSR 3.1 FG`, `Upscaler`) stay as the tables already keep them. - The FPS counter's second line now goes through `i18n::tr`: it draws with its own `Tab`, past `ui::`'s translation, so main's `game %.0f` was never shown translated. Its buffer is 96 bytes, for UTF-8. - `readme.md`: the row and the counter. - Tests: `test_dlssg_runtime` followed the new returns and now checks that the reasons are named and cleared; `present_source_crop_test` checks the maker noted for FSR 3.1's set and for the built-in fallback. ## Screenshots 3440x1440, Fire Temple. Frame generation 2x: the row and the counter say DLSS FG. ![Frame generation by: Auto: DLSS FG, FPS counter Base 63, DLSS FG](https://raw.githubusercontent.com/polackangry/Hookshot/f23d9aa3c79363f4b1f3a5b1fffde6f95dd2ce47/framegen-shown/framegen-by-dlss.png) Frame generation Auto: the built-in one makes the frames, and the line says why. ![Frame generation by: Auto: Built-in FG with the reason, FPS counter Base 71, Built-in FG](https://raw.githubusercontent.com/polackangry/Hookshot/f23d9aa3c79363f4b1f3a5b1fffde6f95dd2ce47/framegen-shown/framegen-by-auto.png) In Polish (top, 2x) and Russian (bottom, Frame generation Auto): the counter's line and the row's line translated. ![The FPS counter and the row's line in Polish and Russian](https://raw.githubusercontent.com/polackangry/Hookshot/18775d14946bf86bac62ed4c38425bc45ac1418f/framegen-shown/framegen-pl-ru.png) ## Testing In game, native Windows 11, RTX 5060 Ti 16 GB, driver 617.42, a build of this branch on main (51da046) with NVIDIA's SDK and `nvngx_dlssg.dll` 310.9.16, DLSS Quality, a save in the Fire Temple. The row's text read through the control bridge (`option get frame_generation_by`), the screenshots of the window: | Case | Row | FPS counter | |---|---|---| | 2x, in play | `Auto: DLSS FG` | `Base 63, DLSS FG`, 126-129 FPS | | 4x, in play | `Auto: DLSS FG` | | | Frame generation Auto | `Auto: Built-in FG` + why | `Base 71, Built-in FG`, 240 FPS | | 4x, the game's pause menu | `Auto: Built-in FG` | | | The title screen | `Auto: Built-in FG` | | | Built-in chosen, 3x | `Built-in FG` | | | Off | `Auto` | | No fault through 6x, 4x, 2x, Auto and Off. Menu language Polish: `Auto: Wbudowany FG`, `Bazowe 63, DLSS FG`, `Teraz bez DLSS FG: ...`; Russian: `База 69, Встроенный FG` and its line. The log agrees: `of the last 600 presents 600 made by DLSS-G` at 2x. Tests: `test_menu_i18n` passes (it failed on the first push: the tables still had `Built-in` and `game %.0f`). The checks that read `vk_menu.cpp`, `vk_dlss.cpp`, `vk_framegen.cpp`, `menu_i18n.h` or the readme, run here: all pass but one; `test_input_background` fails the same on main on this PC (it was in my earlier PRs' list of machine-only failures). A row for the run is in `docs/tested-configurations.md`. Not covered: FSR 3.1 FG on screen (no AMD GPU here; its maker is checked in `present_source_crop_test`), the reasons for an RTX 30 card or a missing runtime on screen (read from the code paths, and the missing runtime from this PC's earlier log). The race when stepping the multiplier up is #2658's, not touched here. 🤖 Generated with [Claude Code](https://claude.com/claude-code) <!-- github-sync:item:2788 -->
At Auto, nothing said whether NVIDIA DLSS Frame Generation or the
built-in interpolation made the frames: without nvngx_dlssg.dll, without
a temporal Upscaler, in the game's menus or at the Auto multiplier it is
the built-in one, silently.

- vk_framegen.cpp: framegen_maker, who made the last 64 sets' frames
  (DLSS FG, FSR 3.1 FG, Built-in FG; none after half a second without).
- vk_dlss.cpp: dlss_fg_why_not, why the last set asked of DLSS-G went to
  the renderer's own (off, the device, NGX's start, too many frames for
  its runtime, no temporal Upscaler picture, refused); a made set clears
  it, a set with nothing to blend from leaves it.
- The menu: the row reads "Auto: DLSS FG" / "Auto: Built-in FG" (the
  choice "Built-in FG"), its line says why not DLSS FG, and the FPS
  counter's second line is "Base 60, DLSS FG" (was "game 60").
- Tests: test_dlssg_runtime follows the new returns and checks the
  reasons; present_source_crop_test checks the maker of each set.

Co-Authored-By: Claude Opus 5.5 <[email protected]>
test_menu_i18n: the tables kept "Built-in" and "game %.0f", texts the
menu no longer writes.

- The texts are the menu's now: vk_dlss.cpp gives the reason as a code
  (dlss_fg_skip, FgSkip in vk_dlss.h), vk_menu.cpp framegen_skip_text
  says it, so menu_i18n.h can translate it.
- menu_i18n.h: "Auto: %s", "Base %.0f", "Base %.0f, %s", "Built-in FG",
  "Not DLSS FG now: %s" and each reason in all 13 tables; the two old
  texts let go, the counts kept.
- The FPS counter's second line goes through i18n:🇹🇷 it draws past
  ui::'s translation, so "game %.0f" was never shown translated. Its
  buffer is 96 bytes for UTF-8.

Co-Authored-By: Claude Opus 5.5 <[email protected]>
github-sync commented 2026-11-12 14:48:23 +00:00
Author
Owner

@Joey-Magnus approved on GitHub · 2026-11-12 14:35 UTC · original

I merged this into current main and ran every check that reads vk_menu.cpp, vk_dlss.cpp, vk_framegen.cpp or menu_i18n.h (33 checks, test_menu_i18n, test_dlssg_runtime and present_source_crop_test among them). All pass, and CI is green.

The threading is fine. note_maker only runs on the recording thread, at vk_framegen.cpp:898 and in the Auto branch. The ring itself isn't shared, and the menu thread only reads the relaxed maker_count atomics and the timestamp. Each early return of dlss_fg_frames now names its reason, and the !blend and closed returns deliberately leave the old one, which matches the description. The screenshots and the table of cases on an RTX 5060 Ti cover what a player sees.

One heads-up for the maintainers: this conflicts with #2665 (FSR 4 / FSR 3.1 as Frame generation by choices) in vk_framegen.cpp, vk_menu.cpp, readme.md and tests/test_framegen_preset.py. Whichever lands second needs a rebase. FgMaker and framegen_maker_name are already shaped for an "FSR 4 FG" entry, so it should be mechanical. It merges cleanly with #2658.

> [@Joey-Magnus](https://github.com/Joey-Magnus) approved on GitHub · 2026-11-12 14:35 UTC · [original](https://github.com/NotRexed/Hookshot/pull/2788#pullrequestreview-5876257022) I merged this into current main and ran every check that reads vk_menu.cpp, vk_dlss.cpp, vk_framegen.cpp or menu_i18n.h (33 checks, test_menu_i18n, test_dlssg_runtime and present_source_crop_test among them). All pass, and CI is green. The threading is fine. note_maker only runs on the recording thread, at vk_framegen.cpp:898 and in the Auto branch. The ring itself isn't shared, and the menu thread only reads the relaxed maker_count atomics and the timestamp. Each early return of dlss_fg_frames now names its reason, and the !blend and closed returns deliberately leave the old one, which matches the description. The screenshots and the table of cases on an RTX 5060 Ti cover what a player sees. One heads-up for the maintainers: this conflicts with #2665 (FSR 4 / FSR 3.1 as Frame generation by choices) in vk_framegen.cpp, vk_menu.cpp, readme.md and tests/test_framegen_preset.py. Whichever lands second needs a rebase. FgMaker and framegen_maker_name are already shaped for an "FSR 4 FG" entry, so it should be mechanical. It merges cleanly with #2658. <!-- github-sync:r:5876257022 -->
Conflicts: gpu/vk/menu_i18n.h's kLangs counts (each set to the entries its
table holds now: main's count plus this branch's 11 texts),
tests/present_source_crop_test.cpp (main's antilag_frame stub and this
branch's note_maker stub, both kept), docs/tested-configurations.md (both
sides' rows kept).

Co-Authored-By: Claude Opus 5.5 <[email protected]>
github-sync closed this pull request 2026-11-12 19:02:18 +00:00
Commenting is not possible because the repository is archived.
Reviewers
No reviewers
Labels
Clear labels   
accessibility
Barrier affecting people with disabilities

  
autorelease: pending

   
autorelease: tagged

   
bug
Something isn't working

  
dependencies
Pull requests that update a dependency file

  
documentation
Improvements or additions to documentation

  
duplicate
This issue or pull request already exists

  
enhancement
New feature or request

  
github_actions
Pull requests that update GitHub Actions code

  
good first issue
Good for newcomers

  
help wanted
Extra attention is needed

  
invalid
This doesn't seem right

  
question
Further information is requested

  
testing wanted

   
wontfix
This will not be worked on

No labels
accessibility
autorelease: pending
autorelease: tagged
bug
dependencies
documentation
duplicate
enhancement
github_actions
good first issue
help wanted
invalid
question
testing wanted
wontfix
Milestone
Clear milestone
No items
No milestone
Assignees
Clear assignees
No assignees
1 participant
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/Hookshot!2788
No description provided.