fix(vk): Arc discrete cards no longer lose Kokiri Forest to the sky: uniforms snapshotted on every Intel GPU #446

Open

nyannoying wants to merge 2 commits from nyannoying/Longshot:fix/arc-discrete-ub-snapshot into main

pull from: nyannoying/Longshot:fix/arc-discrete-ub-snapshot

merge into: MagicBeanDev:main

MagicBeanDev:main

Conversation 2 Commits 2 Files changed 9 +136 -11

nyannoying commented 2026-11-14 14:23:55 +00:00

First-time contributor

• Copy link
• Copy Markdown

Why

On Intel Arc discrete cards (tested: Arc B580), Kokiri Forest shows the sky where the ground and Link's treehouse should be, with white glowing blocks and objects that flicker in and out, from the moment you spawn.

What changed

• vk: ub_snapshot_wanted() in gpu/vk/vk_draw.cpp decides the uniform snapshot (RX_VK_UB_SNAPSHOT); uniform_range() asks it. The default is now on for every Intel GPU (vendor 0x8086), not only for Intel GPUs without VRAM mirrors (the iGPUs). The comment in uniform_range() says why: with mirrors on, the B580 still read uniforms the game had already rewritten for the next frame, so the depth prepass drew the ground with other objects' matrices and the fog pass painted sky over it.
• tests: tests/test_ub_snapshot_default.py compiles the decision alone against 27 cases (unset: on for Intel, off for AMD and NVIDIA; the switch's on/off spellings) and checks that uniform_range() asks it. It fails on main.
• docs: RX_VK_UB_SNAPSHOT row in docs/development.md (it had none).

Flags and settings

Flag	What it does	Default now	Recommendation	Helps on	Doesn't help on	Untested on
RX_VK_UB_SNAPSHOT=0/1	1: uniform buffers are copied into the stream as submitted instead of bound in the game's memory	1 on every Intel GPU (was: Intel without mirrors only); 0 elsewhere	On by default on Intel	Arc B580 (fixes the white forest and glowing blocks, and is faster); Intel iGPUs already had it	—	Arc A-series (Alchemist) discrete, Arc on Linux (Mesa)

Measurements

main: cf605c85 (2026-11-14)
this PR: fix/arc-discrete-ub-snapshot @ 9305d117
Hardware: Intel Arc B580 12 GB, driver 32.0.101.9154; Core i5-13600K; 64 GB; Windows 11 Pro (26100-series Insider, 10.0.29704)
Scene: Kokiri Forest (child), the spawn spot on Link's balcony after the opening; 2048x1152 output, game's own upscaler, frame generation off, uncapped (HOS_FPS_MAX=1000), Real time game speed
Method: 6 runs, the switch alternated (on, off, on, off, on, off), each about 50 s in free play standing still; fps from the log's vi: ... frames presented (... fps over the last 5.0 s) lines (last 8 windows of each run), GPU time from vk: GPU time ... ms a frame (last 6 lines). Median of the windows, then median of the runs.
How to reproduce: start a new game, skip the opening into Kokiri Forest, stand still; RX_VK_UB_SNAPSHOT=0 gives main's behaviour on this GPU.

Baseline: this PR @ 9305d117, RX_VK_UB_SNAPSHOT=0 (main's path), which is main's code path on an Arc discrete card. main itself was not the comparison because on the B580 it loses the device entering Kokiri Forest (the float atomic add loop, fixed in #444); both sides here carry that fix, nothing else differs. The runs were made with 47ebdc75, this change on main cf605c85; the branch has since been rebased onto main 5e072fc8 (9305d117) with the same diff and was not run in the game again: gpu/vk/vk_draw.cpp did not change between the two mains, and a main-based build has another shader stamp than this card's prepared caches, which is not paid again for an unchanged diff. The change was in every build of the later work on this card (#443, whose walks start at this spawn spot): the frames looked at there show the ground and the treehouse drawn.

	this PR @ 9305d117, RX_VK_UB_SNAPSHOT=0 (main's path)	this PR @ 9305d117 (default: 1 on Intel)
fps (median of 3 runs)	69.4 (69.5, 68.6, 69.4)	78.9 (78.2, 78.9, 79.0)
GPU time a frame	9.4 ms	6.9 ms
Ground and treehouse drawn	no (sky)	yes

u_on1 snap=1 fps: 77.6 77.9 77.5 77.1 78.5 77.9 79.2 79.3  | GPU ms: 6.9 6.8 6.9 6.8 6.9 6.8
u_off1 snap=0 fps: 69.5 69.7 69.4 69.4 69.4 69.6 69.6 69.4  | GPU ms: 9.4 9.3 9.3 9.4 9.5 9.3
u_on2 snap=1 fps: 78.2 77.0 78.8 78.9 78.7 79.3 79.3 79.3  | GPU ms: 6.9 6.9 6.9 6.9 6.9 6.9
u_off2 snap=0 fps: 68.2 68.6 67.2 68.6 68.4 69.5 69.0 69.9  | GPU ms: 9.2 9.3 9.4 9.4 9.5 9.3
u_on3 snap=1 fps: 78.5 79.1 78.9 79.5 79.1 79.8 78.7 79.0  | GPU ms: 6.9 6.9 6.9 6.9 6.9 6.8
u_off3 snap=0 fps: 43.6 48.8 41.0 49.4 61.8 69.1 69.5 69.4  | GPU ms: 9.4 9.4 9.3 9.3 9.3 9.4

Testing

Tested: Arc B580, 101.9154, Windows 11: Kokiri Forest from spawn, 8 runs with the snapshot on and 8 off (switch bisect, interleaved builds, and the measurement above). Off: ground and treehouse replaced by sky, glowing white blocks, a pot that appears and disappears between frames, in every run. On: the scene is right in every run. Other switches tried and ruled out: RX_VK_SPLIT_UBFLUSH=0 (fixes the walls, not the glow), RX_VK_EQ_BIAS=1024 and =0 (no change); RX_VK_MIRROR=0 also fixes it, since it turns the snapshot on, at about half the frame rate. python -I -B tests/run_all.py on the rebased head: 486 passed, 6 failed, 70 skipped; the 6 are test_module_ids, test_pad_layout, test_tested_configurations, test_oot_state_tap, test_fpx_sticky and test_spawn_emit, which fail the same way on main 5e072fc8 here. tests/test_ub_snapshot_default.py passes here and fails on main 5e072fc8.

Not tested: other areas than Kokiri Forest; Arc A-series discrete; Arc on Linux; long sessions; the rebased head in the game (see Baseline); a Linux build (no Linux compiler on this PC: the change is one function deciding a default in gpu/vk/vk_draw.cpp, so the template's two-platform box is left out, not ticked).

Needs other testers:

• Arc A750/A770 (Alchemist), Windows: Kokiri Forest from spawn, expecting the ground and houses drawn and no glowing blocks, and no fps loss (T-RY6D)
• Intel iGPU (Core Ultra): unchanged behaviour (the snapshot was already on there), expecting no change (T-4SPA)

Screenshots

Kokiri Forest spawn, Arc B580: posted in the first comment ("main's path" with RX_VK_UB_SNAPSHOT=0, and this PR).

Risk and rollback

Only Intel GPUs change. A uniform buffer copied each draw costs CPU time; on the B580 it measured faster overall. RX_VK_UB_SNAPSHOT=0 turns it off; reverting the commit is enough.

Reviewer notes

Why the VRAM mirrors don't cover this on the B580 is not explained yet: the mirror is made at the submission, so it should not see the game's later rewrite. Pools still waiting for a mirror or refused one bind in guest memory (see #222); that is the likely gap, left for a follow-up. RX_VK_SPLIT_UBFLUSH stays as it is on Arc discrete: the snapshot alone fixes the scene.

Issues

See #320

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.
• 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.
• Flags noted, and the switch reference updated (docs/development.md, the README).
• Screenshots for a change that shows on screen.
• A file in docs/tested/ for each new measured run (format in docs/tested-configurations.md; no appended row or assembled view).
• A game-free test added for the fix, failing on main and passing here.
• Docs updated.

🤖 Generated with Claude Code

## Why On Intel Arc discrete cards (tested: Arc B580), Kokiri Forest shows the sky where the ground and Link's treehouse should be, with white glowing blocks and objects that flicker in and out, from the moment you spawn. ## What changed - vk: `ub_snapshot_wanted()` in `gpu/vk/vk_draw.cpp` decides the uniform snapshot (`RX_VK_UB_SNAPSHOT`); `uniform_range()` asks it. The default is now on for every Intel GPU (vendor 0x8086), not only for Intel GPUs without VRAM mirrors (the iGPUs). The comment in `uniform_range()` says why: with mirrors on, the B580 still read uniforms the game had already rewritten for the next frame, so the depth prepass drew the ground with other objects' matrices and the fog pass painted sky over it. - tests: `tests/test_ub_snapshot_default.py` compiles the decision alone against 27 cases (unset: on for Intel, off for AMD and NVIDIA; the switch's on/off spellings) and checks that `uniform_range()` asks it. It fails on `main`. - docs: `RX_VK_UB_SNAPSHOT` row in `docs/development.md` (it had none). ## Flags and settings | Flag | What it does | Default now | Recommendation | Helps on | Doesn't help on | Untested on | |---|---|---|---|---|---|---| | `RX_VK_UB_SNAPSHOT=0/1` | 1: uniform buffers are copied into the stream as submitted instead of bound in the game's memory | 1 on every Intel GPU (was: Intel without mirrors only); 0 elsewhere | On by default on Intel | Arc B580 (fixes the white forest and glowing blocks, and is faster); Intel iGPUs already had it | — | Arc A-series (Alchemist) discrete, Arc on Linux (Mesa) | ## Measurements main: cf605c85 (2026-11-14) this PR: fix/arc-discrete-ub-snapshot @ 9305d117 Hardware: Intel Arc B580 12 GB, driver 32.0.101.9154; Core i5-13600K; 64 GB; Windows 11 Pro (26100-series Insider, 10.0.29704) Scene: Kokiri Forest (child), the spawn spot on Link's balcony after the opening; 2048x1152 output, game's own upscaler, frame generation off, uncapped (`HOS_FPS_MAX=1000`), Real time game speed Method: 6 runs, the switch alternated (on, off, on, off, on, off), each about 50 s in free play standing still; fps from the log's `vi: ... frames presented (... fps over the last 5.0 s)` lines (last 8 windows of each run), GPU time from `vk: GPU time ... ms a frame` (last 6 lines). Median of the windows, then median of the runs. How to reproduce: start a new game, skip the opening into Kokiri Forest, stand still; `RX_VK_UB_SNAPSHOT=0` gives main's behaviour on this GPU. Baseline: this PR @ 9305d117, `RX_VK_UB_SNAPSHOT=0` (main's path), which is main's code path on an Arc discrete card. main itself was not the comparison because on the B580 it loses the device entering Kokiri Forest (the float atomic add loop, fixed in https://git.magicbean.dev/MagicBeanDev/Longshot/pulls/444); both sides here carry that fix, nothing else differs. The runs were made with 47ebdc75, this change on main cf605c85; the branch has since been rebased onto main 5e072fc8 (9305d117) with the same diff and was not run in the game again: `gpu/vk/vk_draw.cpp` did not change between the two mains, and a main-based build has another shader stamp than this card's prepared caches, which is not paid again for an unchanged diff. The change was in every build of the later work on this card (https://git.magicbean.dev/MagicBeanDev/Longshot/pulls/443, whose walks start at this spawn spot): the frames looked at there show the ground and the treehouse drawn. | | this PR @ 9305d117, `RX_VK_UB_SNAPSHOT=0` (main's path) | this PR @ 9305d117 (default: 1 on Intel) | |---|---|---| | fps (median of 3 runs) | 69.4 (69.5, 68.6, 69.4) | 78.9 (78.2, 78.9, 79.0) | | GPU time a frame | 9.4 ms | 6.9 ms | | Ground and treehouse drawn | no (sky) | yes | ``` u_on1 snap=1 fps: 77.6 77.9 77.5 77.1 78.5 77.9 79.2 79.3 | GPU ms: 6.9 6.8 6.9 6.8 6.9 6.8 u_off1 snap=0 fps: 69.5 69.7 69.4 69.4 69.4 69.6 69.6 69.4 | GPU ms: 9.4 9.3 9.3 9.4 9.5 9.3 u_on2 snap=1 fps: 78.2 77.0 78.8 78.9 78.7 79.3 79.3 79.3 | GPU ms: 6.9 6.9 6.9 6.9 6.9 6.9 u_off2 snap=0 fps: 68.2 68.6 67.2 68.6 68.4 69.5 69.0 69.9 | GPU ms: 9.2 9.3 9.4 9.4 9.5 9.3 u_on3 snap=1 fps: 78.5 79.1 78.9 79.5 79.1 79.8 78.7 79.0 | GPU ms: 6.9 6.9 6.9 6.9 6.9 6.8 u_off3 snap=0 fps: 43.6 48.8 41.0 49.4 61.8 69.1 69.5 69.4 | GPU ms: 9.4 9.4 9.3 9.3 9.3 9.4 ``` ## Testing **Tested:** Arc B580, 101.9154, Windows 11: Kokiri Forest from spawn, 8 runs with the snapshot on and 8 off (switch bisect, interleaved builds, and the measurement above). Off: ground and treehouse replaced by sky, glowing white blocks, a pot that appears and disappears between frames, in every run. On: the scene is right in every run. Other switches tried and ruled out: `RX_VK_SPLIT_UBFLUSH=0` (fixes the walls, not the glow), `RX_VK_EQ_BIAS=1024` and `=0` (no change); `RX_VK_MIRROR=0` also fixes it, since it turns the snapshot on, at about half the frame rate. `python -I -B tests/run_all.py` on the rebased head: 486 passed, 6 failed, 70 skipped; the 6 are `test_module_ids`, `test_pad_layout`, `test_tested_configurations`, `test_oot_state_tap`, `test_fpx_sticky` and `test_spawn_emit`, which fail the same way on main 5e072fc8 here. `tests/test_ub_snapshot_default.py` passes here and fails on main 5e072fc8. **Not tested:** other areas than Kokiri Forest; Arc A-series discrete; Arc on Linux; long sessions; the rebased head in the game (see Baseline); a Linux build (no Linux compiler on this PC: the change is one function deciding a default in `gpu/vk/vk_draw.cpp`, so the template's two-platform box is left out, not ticked). **Needs other testers:** - Arc A750/A770 (Alchemist), Windows: Kokiri Forest from spawn, expecting the ground and houses drawn and no glowing blocks, and no fps loss (T-RY6D) - Intel iGPU (Core Ultra): unchanged behaviour (the snapshot was already on there), expecting no change (T-4SPA) ## Screenshots Kokiri Forest spawn, Arc B580: posted in the first comment ("main's path" with `RX_VK_UB_SNAPSHOT=0`, and this PR). ## Risk and rollback Only Intel GPUs change. A uniform buffer copied each draw costs CPU time; on the B580 it measured faster overall. `RX_VK_UB_SNAPSHOT=0` turns it off; reverting the commit is enough. ## Reviewer notes Why the VRAM mirrors don't cover this on the B580 is not explained yet: the mirror is made at the submission, so it should not see the game's later rewrite. Pools still waiting for a mirror or refused one bind in guest memory (see https://git.magicbean.dev/MagicBeanDev/Longshot/pulls/222); that is the likely gap, left for a follow-up. `RX_VK_SPLIT_UBFLUSH` stays as it is on Arc discrete: the snapshot alone fixes the scene. ## Issues See https://git.magicbean.dev/MagicBeanDev/Longshot/issues/320 ## 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). - [x] Measured: main and this PR at named commits, on the same PC, scene and settings, at least 2 interleaved runs each, the median reported. - [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] Flags noted, and the switch reference updated (docs/development.md, the README). - [x] Screenshots for a change that shows on screen. - [x] A file in `docs/tested/` for each new measured run (format in `docs/tested-configurations.md`; no appended row or assembled view). - [x] A game-free test added for the fix, failing on `main` and passing here. - [x] Docs updated. 🤖 Generated with [Claude Code](https://claude.com/claude-code)

nyannoying added 2 commits 2026-11-14 14:23:55 +00:00

fix(vk): Arc discrete cards no longer lose Kokiri Forest to the sky: uniforms snapshotted on every Intel GPU ... 9305d11754

The uniform snapshot (RX_VK_UB_SNAPSHOT) was on only for Intel GPUs without VRAM mirrors, the iGPUs. On an Arc B580
(101.9154) with mirrors on, the game still rewrote a frame's uniforms before the GPU read them: the Kokiri Forest floor
and Link's treehouse were drawn with other objects' matrices and showed the sky, with white glowing blocks and pots
that came and went. With the snapshot the scene is right, at no frame rate cost there.

ub_snapshot_wanted() decides it; tests/test_ub_snapshot_default.py checks the default and that uniform_range asks it.

Co-Authored-By: Claude Opus 5.5 <[email protected]>

docs(tested): the uniform snapshot runs on an Arc B580, six files ... 3c8174d3a1

Co-Authored-By: Claude Fable 5.2 <[email protected]>

nyannoying commented 2026-11-14 14:24:12 +00:00

Author

First-time contributor

• Copy link
• Copy Markdown

Kokiri Forest, the spawn spot on Link's balcony, Arc B580 (101.9154), the same build and spot, only the switch differs.

Kokiri Forest, the spawn spot on Link's balcony, Arc B580 (101.9154), the same build and spot, only the switch differs.  

kokiri_main.png

2.6 MiB

kokiri_fix.png

2.6 MiB

nyannoying commented 2026-11-14 14:24:29 +00:00

Author

First-time contributor

• Copy link
• Copy Markdown

The two test requests are open (T-RY6D: Arc A750/A770; T-4SPA: an Intel iGPU, expecting no change) and named in the description; both wait for a maintainer's testing wanted label. A B580 needs #444 as well to reach the forest.

The two test requests are open (T-RY6D: Arc A750/A770; T-4SPA: an Intel iGPU, expecting no change) and named in the description; both wait for a maintainer's testing wanted label. A B580 needs https://git.magicbean.dev/MagicBeanDev/Longshot/pulls/444 as well to reach the forest.

nyannoying changed title from WIP: fix(vk): Arc discrete cards no longer lose Kokiri Forest to the sky: uniforms snapshotted on every Intel GPU to fix(vk): Arc discrete cards no longer lose Kokiri Forest to the sky: uniforms snapshotted on every Intel GPU 2026-11-14 14:24:31 +00:00

mooki added the

kind

bug

priority

high

labels 2026-11-14 14:57:44 +00:00

nyannoying referenced this pull request 2026-11-14 15:02:45 +00:00

feat(vk): Intel XeSS Super Resolution as a temporal upscaler (Vulkan, the player's libxess.dll) #459

nyannoying referenced this pull request 2026-11-14 16:18:38 +00:00

Game: up-front shader preparation does not finish on Intel Arc (Windows) with compiles one at a time #407

mooki added the

area/renderer

gpu/intel

labels 2026-11-14 16:28:03 +00:00

nyannoying referenced this pull request 2026-11-14 16:52:47 +00:00

perf(vk): the Arc B-series compiles shaders side by side by default: a new area no longer freezes the picture for minutes #500

Some checks are pending

Hide checks

CI / Conventional Commit title (pull_request) Blocked by required conditions

Details

CI / Python source checks (ubuntu-24.04, part 1/2) (pull_request) Blocked by required conditions

Details

CI / Workflow lint (actionlint) (pull_request) Blocked by required conditions

Details

Third-party parts / As their notes say (pull_request) Blocked by required conditions

Details

Third-party parts / The same as upstream (pull_request) Blocked by required conditions

Details

Website / Website on Cloudflare (pull_request) Blocked by required conditions

Details

CI / Python source checks (ubuntu-24.04, part 2/2) (pull_request) Blocked by required conditions

Details

CI / SPIR-V fixtures (global narrow load/store, direct surface atomics, F2I immediate) (pull_request) Blocked by required conditions

Details

PR description / Description follows the template (pull_request_target) Successful in 4s

Details

Website / Website on Cloudflare (pull_request_target) Has been skipped

Details

This pull request doesn't have enough approvals yet. 0 of 1 approvals granted.

Some workflows are waiting to be reviewed.

This branch is out-of-date with the base branch

You are not authorized to merge this pull request.

View command line instructions

Checkout

From your project repository, check out a new branch and test the changes.

git fetch -u 'fix/arc-discrete-ub-snapshot:nyannoying-fix/arc-discrete-ub-snapshot'

git switch 'nyannoying-fix/arc-discrete-ub-snapshot'

Sign in to join this conversation.

Reviewers

No reviewers

Labels

Clear labels   

area/accessibility

Barriers for people with disabilities

  

area/audio

Sound and music

  

area/input

Controllers, keyboard, mouse

  

area/install

Installing, updating, building from your game

  

area/menu

In-game menu and settings

  

area/renderer

Vulkan renderer, shaders, effects

  

area/saves

Save files

  

area/upscaling

DLSS, FSR, frame generation

  

good first issue

A good start for a new contributor

  

gpu/amd

AMD GPUs

  

gpu/igpu

Integrated GPUs

  

gpu/intel

Intel GPUs

  

gpu/nvidia

NVIDIA GPUs

  

help wanted

Extra hands or knowledge wanted

  

kind

bug

Something doesn't work as it should

  

kind

chore

Build, CI, dependencies, tidying

  

kind

docs

Documentation

  

kind

feature

A new feature or an improvement

  

kind

performance

Frame rate, frame time, loading, memory

  

kind

question

A question rather than a change

  

needs measurements

Waiting for before/after measurements

  

platform/linux

Linux only

  

platform/steam-deck

Steam Deck

  

platform/windows

Windows only

  

preview

Maintainers: deploy a preview of this PR's website (read its changes first). Removed after the run.

  

priority

critical

Crashes, data loss, or blocks most players

  

priority

high

Important: next in line

  

priority

low

Nice to have

  

status

blocked

Waits on something else first

  

status

confirmed

Reproduced or agreed on

  

status

duplicate

Already reported or proposed

  

status

invalid

Not a problem in this project

  

status

needs-info

Waiting for more information from the reporter

  

status

wontfix

Will not be worked on

  

testing wanted

Wanted: a test on other hardware (posts a test request in #test-requests)

No labels

area/accessibility

area/audio

area/input

area/install

area/menu

area/renderer

area/saves

area/upscaling

good first issue

gpu/amd

gpu/igpu

gpu/intel

gpu/nvidia

help wanted

kind

bug

kind

chore

kind

docs

kind

feature

kind

performance

kind

question

needs measurements

platform/linux

platform/steam-deck

platform/windows

preview

priority

critical

priority

high

priority

low

status

blocked

status

confirmed

status

duplicate

status

invalid

status

needs-info

status

wontfix

testing wanted

Milestone

Clear milestone

No items

No milestone

Projects

Clear projects

No items

No project

Assignees

Clear assignees

No assignees

1 participant

Notifications

Subscribe

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!446

No description provided.