fix(vk): the start-up bar counts the compute pipelines, not 100% through them #256

Open

gab9281 wants to merge 2 commits from gab9281/Longshot:fix/startup-bar-compute into main

pull from: gab9281/Longshot:fix/startup-bar-compute

merge into: MagicBeanDev:main

MagicBeanDev:main

Conversation 2 Commits 2 Files changed 2 +22 -9

gab9281 commented 2026-11-13 17:58:18 +00:00

Member

• Copy link
• Copy Markdown

Why

On the first start after a shader preparation, the start-up screen sits at "100%, a few seconds left" for about 5 minutes (RTX 3060 Laptop), which looks like being stuck on the last shader. The bar counted only the noted pipelines, then stayed full while every compute program's pipeline was built.

What changed

• vk: precompile_warm() in gpu/vk/vk_draw.cpp lists the compute programs' keys (program_cache_compute_keys()) before it starts. Both steps count on one total, the noted pipelines and then the compute pipelines, so the bar moves through the compute step instead of sitting at 100%. The work done is the same: only the count given to the bar changes.
• tests: tests/test_warm_bounded.py checks that the compute step reports its own progress against the whole. It fails on main, where it reports progress(total, total).

Flags and settings

None.

Measurements

What changes is the number the start-up screen shows, not the time: the same pipelines are built in the same order. So the measurement is the bar's value at the start of the compute step, from the start-up log line, with the shaders' time on this PR for reference.

main: 36a12f6f (2026-11-13; precompile_warm's bar code is the same on main @ 514591a3, which this PR is on: main's renderer changes since, #234 and #260, are elsewhere in vk_draw.cpp)
this PR: fix/startup-bar-compute @ 2af5f18e (measured there; 82271725, the head now, changes only the count after a stopped noted step, which this start didn't have)
Hardware: RTX 3060 Laptop (NVIDIA 620.70.06), Ryzen 7 6800H, 15 GB, CachyOS (Linux 7.3.3), KDE Wayland
Scene: the start-up screen of the first start after a preparation (RX_VK_LOAD=create), default settings
Method: the start-up screen's vk: start-up: pipelines … of … log line once the noted pipelines are done, one start each, started through game-performance
How to reproduce: prepare the shaders (a build with --prepare-shaders), start the game, and watch the start-up screen and longshot_log.txt

	main @ 36a12f6f	this PR @ 2af5f18e
The bar when the compute step starts	24457 of 24457, 100%, "a few seconds left"	27918 of 29473, 95%
The bar through the compute step	stays at 100% (about 5 min with nothing cached)	counts up to 100%

main:    [hos host] vk: start-up: pipelines 24457 of 24457, 100% of all, a few seconds left
this PR: [hos host] vk: start-up: pipelines 27918 of 29473, 95% of all, a few seconds left
this PR: [hos host] vk: precompile: 25346 noted pipelines of 16072 programs and 9925 compute pipelines of 4127 compute programs built into the pipeline cache and let go
this PR: [hos host] vk: start-up: the shaders took 11.3 s

29473 = 25346 noted + 4127 compute programs. The two starts had different numbers of noted pipelines (main's folder was prepared at another commit), and this start had the compute pipelines cached (11.3 s), so the times aren't compared. A run-to-run timing table wouldn't show anything: no pipeline is added or removed.

Testing

Tested:

• In the game, Linux (the hardware above): built at 2af5f18e (the same bar code at 82271725, rebased on main @ 514591a3 with no conflict), a first start after a preparation. The start-up screen counted through the compute step to the end, with the log lines above.
• Game-free: python -I -B tests/test_warm_bounded.py passes, and fails on main. RX_CPM=<the CPM cache> python -I -B tests/run_all.py at 82271725 (main @ 514591a3): 517 passed, 1 failed, 15 skipped (Windows only). The failure, tests/test_menu_i18n.py (a pt_BR text of a menu line main changed), fails on main 514591a3 too.

Not tested:

• Windows: the code is the same (precompile_warm isn't platform-specific). The game builds for Windows with llvm-mingw at 2c2100d1 (4:38, the same code on the main before); not run on Windows.
• A start with nothing cached on this PR, where the compute step takes minutes: this PC had them cached by then. The count is the same either way.

Needs other testers:
None (this is the bar's count, the same on every GPU).

Screenshots

Nothing new on screen: the same start-up screen, with a percentage that keeps moving.

Risk and rollback

The progress callback gets a larger total. If the compute step is skipped (the preparation was stopped), the bar ends below 100%, as any stopped preparation does. Reverting the commit is enough.

Reviewer notes

The compute keys were already listed, after the noted pipelines; this lists them once, before.

A /code-review (high) found 7 things:

• Fixed (2c2100d1):
  • after a noted step stopped part-way, the compute step's only tick said every noted pipeline was done (total + computes_done); it now reports done + computes_done;
  • the compute keys are indexed, not copied into a second list;
  • the test pins the expression and that no tick reports total.
• Left as is, on purpose:
  • A compute pipeline costs more than a noted one, but both count as one. That is how the normal precompile (precompile_run) already counts its notes, computes and fallbacks on one total. The time left comes from the bar's rate over the last 15 s (ProgressClock), so it slows down with the step instead of saying "a few seconds" for minutes. A weight per kind would change both paths and is a follow-up.
  • Part 2's guess (prepare_guess) counts the computes play dispatched, not every kept one. The normal precompile has the same gap on a first start (computes_plan's "every kept one"). The bar never goes back, so at worst it holds still for a moment when the count replaces the guess. Same follow-up.
  • The total counts keys the worker skips (failed, TAAU). They are counted done as they are skipped, so the bar still reaches the end. The log's "built" line counts what was built.
  • The keys are listed before the noted step, even on a start that is stopped during it. This is one walk of the index under its lock, needed for the total from the start.

Issues

None: found while testing #206 (the start-up screen "stuck on the last shader").

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. (The bar's count, which doesn't vary between runs: 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 game-free test added for the fix, failing on main and passing here.
• A check under tests/ that reads changed source updated in this PR, not weakened.
• Host code builds on both Windows (llvm-mingw) and Linux, with UTF-8 paths.

🤖 Generated with Claude Code

https://claude.ai/code/session_015KgSKoUqUvce5UXG89NBgi

## Why On the first start after a shader preparation, the start-up screen sits at "100%, a few seconds left" for about 5 minutes (RTX 3060 Laptop), which looks like being stuck on the last shader. The bar counted only the noted pipelines, then stayed full while every compute program's pipeline was built. ## What changed - vk: `precompile_warm()` in `gpu/vk/vk_draw.cpp` lists the compute programs' keys (`program_cache_compute_keys()`) before it starts. Both steps count on one total, the noted pipelines and then the compute pipelines, so the bar moves through the compute step instead of sitting at 100%. The work done is the same: only the count given to the bar changes. - tests: `tests/test_warm_bounded.py` checks that the compute step reports its own progress against the whole. It fails on main, where it reports `progress(total, total)`. ## Flags and settings None. ## Measurements What changes is the number the start-up screen shows, not the time: the same pipelines are built in the same order. So the measurement is the bar's value at the start of the compute step, from the start-up log line, with the shaders' time on this PR for reference. main: 36a12f6f (2026-11-13; `precompile_warm`'s bar code is the same on main @ 514591a3, which this PR is on: main's renderer changes since, #234 and #260, are elsewhere in `vk_draw.cpp`) this PR: fix/startup-bar-compute @ 2af5f18e (measured there; 82271725, the head now, changes only the count after a stopped noted step, which this start didn't have) Hardware: RTX 3060 Laptop (NVIDIA 620.70.06), Ryzen 7 6800H, 15 GB, CachyOS (Linux 7.3.3), KDE Wayland Scene: the start-up screen of the first start after a preparation (`RX_VK_LOAD=create`), default settings Method: the start-up screen's `vk: start-up: pipelines … of …` log line once the noted pipelines are done, one start each, started through `game-performance` How to reproduce: prepare the shaders (a build with `--prepare-shaders`), start the game, and watch the start-up screen and `longshot_log.txt` | | main @ 36a12f6f | this PR @ 2af5f18e | |---|---|---| | The bar when the compute step starts | 24457 of 24457, 100%, "a few seconds left" | 27918 of 29473, 95% | | The bar through the compute step | stays at 100% (about 5 min with nothing cached) | counts up to 100% | ``` main: [hos host] vk: start-up: pipelines 24457 of 24457, 100% of all, a few seconds left this PR: [hos host] vk: start-up: pipelines 27918 of 29473, 95% of all, a few seconds left this PR: [hos host] vk: precompile: 25346 noted pipelines of 16072 programs and 9925 compute pipelines of 4127 compute programs built into the pipeline cache and let go this PR: [hos host] vk: start-up: the shaders took 11.3 s ``` 29473 = 25346 noted + 4127 compute programs. The two starts had different numbers of noted pipelines (main's folder was prepared at another commit), and this start had the compute pipelines cached (11.3 s), so the times aren't compared. A run-to-run timing table wouldn't show anything: no pipeline is added or removed. ## Testing **Tested:** - In the game, Linux (the hardware above): built at 2af5f18e (the same bar code at 82271725, rebased on main @ 514591a3 with no conflict), a first start after a preparation. The start-up screen counted through the compute step to the end, with the log lines above. - Game-free: `python -I -B tests/test_warm_bounded.py` passes, and fails on main. `RX_CPM=<the CPM cache> python -I -B tests/run_all.py` at 82271725 (main @ 514591a3): 517 passed, 1 failed, 15 skipped (Windows only). The failure, `tests/test_menu_i18n.py` (a pt_BR text of a menu line main changed), fails on main 514591a3 too. **Not tested:** - Windows: the code is the same (`precompile_warm` isn't platform-specific). The game builds for Windows with llvm-mingw at 2c2100d1 (4:38, the same code on the main before); not run on Windows. - A start with nothing cached on this PR, where the compute step takes minutes: this PC had them cached by then. The count is the same either way. **Needs other testers:** None (this is the bar's count, the same on every GPU). ## Screenshots Nothing new on screen: the same start-up screen, with a percentage that keeps moving. ## Risk and rollback The progress callback gets a larger total. If the compute step is skipped (the preparation was stopped), the bar ends below 100%, as any stopped preparation does. Reverting the commit is enough. ## Reviewer notes The compute keys were already listed, after the noted pipelines; this lists them once, before. A `/code-review` (high) found 7 things: - **Fixed** (2c2100d1): - after a noted step stopped part-way, the compute step's only tick said every noted pipeline was done (`total + computes_done`); it now reports `done + computes_done`; - the compute keys are indexed, not copied into a second list; - the test pins the expression and that no tick reports `total`. - **Left as is, on purpose:** - *A compute pipeline costs more than a noted one, but both count as one.* That is how the normal precompile (`precompile_run`) already counts its notes, computes and fallbacks on one total. The time left comes from the bar's rate over the last 15 s (`ProgressClock`), so it slows down with the step instead of saying "a few seconds" for minutes. A weight per kind would change both paths and is a follow-up. - *Part 2's guess (`prepare_guess`) counts the computes play dispatched, not every kept one.* The normal precompile has the same gap on a first start (`computes_plan`'s "every kept one"). The bar never goes back, so at worst it holds still for a moment when the count replaces the guess. Same follow-up. - *The total counts keys the worker skips (failed, TAAU).* They are counted done as they are skipped, so the bar still reaches the end. The log's "built" line counts what was built. - *The keys are listed before the noted step, even on a start that is stopped during it.* This is one walk of the index under its lock, needed for the total from the start. ## Issues None: found while testing #206 (the start-up screen "stuck on the last shader"). ## 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. (The bar's count, which doesn't vary between runs: 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 game-free test added for the fix, failing on `main` and passing here. - [x] A check under `tests/` that reads changed source updated in this PR, not weakened. - [x] Host code builds on both Windows (llvm-mingw) and Linux, with UTF-8 paths. 🤖 Generated with [Claude Code](https://claude.com/claude-code) https://claude.ai/code/session_015KgSKoUqUvce5UXG89NBgi

gab9281 changed title from WIP: fix(vk): the start-up bar counts the compute pipelines, not 100% through them to fix(vk): the start-up bar counts the compute pipelines, not 100% through them 2026-11-13 18:05:07 +00:00

gab9281 force-pushed fix/startup-bar-compute from 2c2100d102 to 8227172508 2026-11-13 19:36:53 +00:00 Compare

gab9281 added the

testing wanted

label 2026-11-13 22:29:19 +00:00

mooki added the

area/renderer

kind

bug

labels 2026-11-14 01:45:34 +00:00

mooki added the

priority

high

label 2026-11-14 04:18:23 +00:00

futondestroyer commented 2026-11-14 15:27:31 +00:00

Member

• Copy link
• Copy Markdown

Test request T-RZC9: pass (Windows startup progress)

The baseline shows 100% while compute pipelines are still being built. PR 256 includes those compute-program keys in the total and shows unfinished progress until that work completes. Both versions reached the title sequence and all four runs exited normally (code 0).

Hardware: Windows 11 Home, build 26300; Core i7-13700F, 32 GB RAM (31.8 GiB usable); RTX 4070 12 GB, NVIDIA 621.01; llvm-mingw Windows renderer build.

Baseline: main @ 514591a30d707fe9f3ad31eae148c3cc1e750bff (the PR's merge base).
Candidate: PR head @ 822717250837ccebbdcb00d6556f200766c0550a, confirmed unchanged after testing.
Built using scripts/dev_packs.py, renderer-only pack over the existing base. The host was unchanged and matched in all runs; candidate renderer SHA-256 c2bc287de365f76351e73110acb4232e393b101e04ed21ad76769bf674b0787b.

Method: four real-window starts, baseline / PR / baseline / PR. Identical runtime and compatible prepared shader caches, RX_VK_LOAD=create, 1280x720, frame generation and HDR off, audio disabled. Cleared only the disposable runtime's longshot_pipelines.bin.built and .built.warming before each start to repeat the first-play warm pass. Original saves and shared base left untouched. The existing NVIDIA driver cache persisted; this was not a cold-driver-cache or startup-speed comparison.

Observation	Baseline	PR 256
Progress total, both runs	24,512	28,639 = 24,512 noted entries + 4,127 compute-program keys
Inspected compute-stage window	24,512 / 24,512, 100%, "a few seconds left"	28,492 / 28,639, 99%, "a few seconds left"
Work completed, each run's log	24,537 noted pipelines and 9,925 compute pipelines of 4,127 compute programs	Same counts
Startup completed; exit	2/2; both code 0	2/2; both code 0

Representative log lines:

baseline: vk: start-up: pipelines 24512 of 24512, 100% of all, a few seconds left
PR:       vk: start-up: pipelines 21469 of 28639, 75% of all, about 15 s left
PR:       vk: precompile: 24537 noted pipelines of 15594 programs and 9925 compute pipelines of 4127 compute programs built into the pipeline cache and let go (0 and 0 not made, 0 whose program variant is not kept), on 16 threads in 27.0 s (the compute ones 5.4 s)

Screenshots of the actual startup windows were inspected locally. No screenshot is uploaded with this report. The compute stage lasted about 5-6 seconds with these warm caches; no claim of reproducing the reporter's five-minute cold wait, accurate ETA, or faster startup.

Checks: tests/test_warm_bounded.py passes on the candidate. Its added progress assertion, run against the baseline source, fails with precompile_warm does not count its compute pipelines in the bar's total. Renderer compilation: 0 failed.

Full game-free suite with the bundled Python dependencies: 496 passed, 4 failed, 33 skipped. The same four checks were rerun on unchanged baseline and all four also fail there: nx2/test_nup_paths.py (Windows symlink privilege), tests/test_check_third_party.py (baseline provenance/fixture failures), tests/test_menu_i18n.py (stale pt_BR text), tests/test_release.py (Windows command/line-ending fixtures). No new failing check was identified. An earlier system-Python run had 487 passed / 7 failed / 39 skipped; missing project dependencies caused the additional failures/skips, so the bundled-Python run is the final suite result.

Limits: no Linux run, no uncached-driver run, and no interrupted-preparation runtime test. This tests the submitted head on its own base, not today's main.

Tested-configuration rows for the PR author/maintainer to add as distinct docs/tested/ records:

Date	Build	OS / desktop	GPU / driver	CPU / RAM	Settings	Where	Result
2026-11-14	main 514591a3, unchanged host and renderer	Windows 11 Home 26300, windowed	RTX 4070 12 GB, 621.01	i7-13700F, 32 GB	1280x720, RX_VK_LOAD=create, FG/HDR/audio off, prepared caches, built marker cleared	First-play startup warm pass; run main-a	Total 24512; log shows 24512/24512 and 100% during compute work; completed and exited 0; no speed claim
2026-11-14	PR 256 renderer 82271725 over main 514591a3 host	Windows 11 Home 26300, windowed	RTX 4070 12 GB, 621.01	i7-13700F, 32 GB	1280x720, RX_VK_LOAD=create, FG/HDR/audio off, prepared caches, built marker cleared	First-play startup warm pass; run pr-b	Total 28639 includes 4127 compute keys; completed and exited 0; no speed claim
2026-11-14	main 514591a3, unchanged host and renderer	Windows 11 Home 26300, windowed	RTX 4070 12 GB, 621.01	i7-13700F, 32 GB	1280x720, RX_VK_LOAD=create, FG/HDR/audio off, prepared caches, built marker cleared	First-play startup warm pass; run main-c2	Screen shows 24512/24512 and 100% during compute work; completed and exited 0; no speed claim
2026-11-14	PR 256 renderer 82271725 over main 514591a3 host	Windows 11 Home 26300, windowed	RTX 4070 12 GB, 621.01	i7-13700F, 32 GB	1280x720, RX_VK_LOAD=create, FG/HDR/audio off, prepared caches, built marker cleared	First-play startup warm pass; run pr-d	Screen shows 28492/28639 and 99% before completion; completed and exited 0; no speed claim

Test request T-RZC9: pass (Windows startup progress) The baseline shows 100% while compute pipelines are still being built. PR 256 includes those compute-program keys in the total and shows unfinished progress until that work completes. Both versions reached the title sequence and all four runs exited normally (code 0). Hardware: Windows 11 Home, build 26300; Core i7-13700F, 32 GB RAM (31.8 GiB usable); RTX 4070 12 GB, NVIDIA 621.01; llvm-mingw Windows renderer build. Baseline: main @ `514591a30d707fe9f3ad31eae148c3cc1e750bff` (the PR's merge base). Candidate: PR head @ `822717250837ccebbdcb00d6556f200766c0550a`, confirmed unchanged after testing. Built using `scripts/dev_packs.py`, renderer-only pack over the existing base. The host was unchanged and matched in all runs; candidate renderer SHA-256 `c2bc287de365f76351e73110acb4232e393b101e04ed21ad76769bf674b0787b`. Method: four real-window starts, baseline / PR / baseline / PR. Identical runtime and compatible prepared shader caches, `RX_VK_LOAD=create`, 1280x720, frame generation and HDR off, audio disabled. Cleared only the disposable runtime's `longshot_pipelines.bin.built` and `.built.warming` before each start to repeat the first-play warm pass. Original saves and shared base left untouched. The existing NVIDIA driver cache persisted; this was not a cold-driver-cache or startup-speed comparison. | Observation | Baseline | PR 256 | |---|---|---| | Progress total, both runs | 24,512 | 28,639 = 24,512 noted entries + 4,127 compute-program keys | | Inspected compute-stage window | 24,512 / 24,512, 100%, "a few seconds left" | 28,492 / 28,639, 99%, "a few seconds left" | | Work completed, each run's log | 24,537 noted pipelines and 9,925 compute pipelines of 4,127 compute programs | Same counts | | Startup completed; exit | 2/2; both code 0 | 2/2; both code 0 | Representative log lines: ```text baseline: vk: start-up: pipelines 24512 of 24512, 100% of all, a few seconds left PR: vk: start-up: pipelines 21469 of 28639, 75% of all, about 15 s left PR: vk: precompile: 24537 noted pipelines of 15594 programs and 9925 compute pipelines of 4127 compute programs built into the pipeline cache and let go (0 and 0 not made, 0 whose program variant is not kept), on 16 threads in 27.0 s (the compute ones 5.4 s) ``` Screenshots of the actual startup windows were inspected locally. No screenshot is uploaded with this report. The compute stage lasted about 5-6 seconds with these warm caches; no claim of reproducing the reporter's five-minute cold wait, accurate ETA, or faster startup. Checks: `tests/test_warm_bounded.py` passes on the candidate. Its added progress assertion, run against the baseline source, fails with `precompile_warm does not count its compute pipelines in the bar's total`. Renderer compilation: 0 failed. Full game-free suite with the bundled Python dependencies: **496 passed, 4 failed, 33 skipped**. The same four checks were rerun on unchanged baseline and all four also fail there: `nx2/test_nup_paths.py` (Windows symlink privilege), `tests/test_check_third_party.py` (baseline provenance/fixture failures), `tests/test_menu_i18n.py` (stale pt_BR text), `tests/test_release.py` (Windows command/line-ending fixtures). No new failing check was identified. An earlier system-Python run had 487 passed / 7 failed / 39 skipped; missing project dependencies caused the additional failures/skips, so the bundled-Python run is the final suite result. Limits: no Linux run, no uncached-driver run, and no interrupted-preparation runtime test. This tests the submitted head on its own base, not today's main. Tested-configuration rows for the PR author/maintainer to add as distinct `docs/tested/` records: | Date | Build | OS / desktop | GPU / driver | CPU / RAM | Settings | Where | Result | |---|---|---|---|---|---|---|---| | 2026-11-14 | main 514591a3, unchanged host and renderer | Windows 11 Home 26300, windowed | RTX 4070 12 GB, 621.01 | i7-13700F, 32 GB | 1280x720, RX_VK_LOAD=create, FG/HDR/audio off, prepared caches, built marker cleared | First-play startup warm pass; run main-a | Total 24512; log shows 24512/24512 and 100% during compute work; completed and exited 0; no speed claim | | 2026-11-14 | PR 256 renderer 82271725 over main 514591a3 host | Windows 11 Home 26300, windowed | RTX 4070 12 GB, 621.01 | i7-13700F, 32 GB | 1280x720, RX_VK_LOAD=create, FG/HDR/audio off, prepared caches, built marker cleared | First-play startup warm pass; run pr-b | Total 28639 includes 4127 compute keys; completed and exited 0; no speed claim | | 2026-11-14 | main 514591a3, unchanged host and renderer | Windows 11 Home 26300, windowed | RTX 4070 12 GB, 621.01 | i7-13700F, 32 GB | 1280x720, RX_VK_LOAD=create, FG/HDR/audio off, prepared caches, built marker cleared | First-play startup warm pass; run main-c2 | Screen shows 24512/24512 and 100% during compute work; completed and exited 0; no speed claim | | 2026-11-14 | PR 256 renderer 82271725 over main 514591a3 host | Windows 11 Home 26300, windowed | RTX 4070 12 GB, 621.01 | i7-13700F, 32 GB | 1280x720, RX_VK_LOAD=create, FG/HDR/audio off, prepared caches, built marker cleared | First-play startup warm pass; run pr-d | Screen shows 28492/28639 and 99% before completion; completed and exited 0; no speed claim |

mooki added the

gpu/nvidia

platform/linux

platform/windows

labels 2026-11-14 18:22:39 +00:00

klimovich008 commented 2026-11-15 13:46:57 +00:00

Contributor

• Copy link
• Copy Markdown

Cold-cache result on Windows, RX 9070 XT, AMD driver 26.10.2: this fixes a real false timeout. With the old progress callback, All preparation stopped at the default 300-second stall watchdog while compute work was still advancing (866 to 887 programs between progress reports).

With this PR's progress change backported over main 82886e89 plus the separate pipeline-input cache fix, the recovered preparation completed with the default watchdog unchanged: 24,458 graphics and 9,687 compute pipelines, zero not made and zero missing compute variants. Compute preparation took 546 seconds across 4,058 programs. The completed prediction stage was retained from the interrupted run; this was a recovery, not one uninterrupted preparation pass.

The source-cut regression also checks that a genuinely stalled compute phase still times out. This result covers preparation; it does not claim gameplay performance or testing the untouched PR head.

Cold-cache result on Windows, RX 9070 XT, AMD driver 26.10.2: this fixes a real false timeout. With the old progress callback, All preparation stopped at the default 300-second stall watchdog while compute work was still advancing (866 to 887 programs between progress reports). With this PR's progress change backported over main 82886e89 plus the separate pipeline-input cache fix, the recovered preparation completed with the default watchdog unchanged: 24,458 graphics and 9,687 compute pipelines, zero not made and zero missing compute variants. Compute preparation took 546 seconds across 4,058 programs. The completed prediction stage was retained from the interrupted run; this was a recovery, not one uninterrupted preparation pass. The source-cut regression also checks that a genuinely stalled compute phase still times out. This result covers preparation; it does not claim gameplay performance or testing the untouched PR head.

malon referenced this pull request 2026-11-15 15:11:03 +00:00

Game: Keeps on crashing in death mountain #817

All checks were successful

Hide checks

CI / Python source checks (ubuntu-24.04, part 1/2) (pull_request) Has been skipped

Details

CI / Python source checks (ubuntu-24.04, part 2/2) (pull_request) Has been skipped

Details

CI / SPIR-V fixtures (global narrow load/store, direct surface atomics, F2I immediate) (pull_request) Has been skipped

Details

Third-party parts / The same as upstream (pull_request) Has been skipped

Details

Website / Website on Cloudflare (pull_request) Has been skipped

Details

CI / Conventional Commit title (pull_request) Successful in 4s

Details

CI / Workflow lint (actionlint) (pull_request) Successful in 8s

Details

Third-party parts / As their notes say (pull_request) Successful in 11s

Details

PR description / Description follows the template (pull_request_target) Successful in 5s

Details

Website / Website on Cloudflare (pull_request_target) Has been skipped

Details

This pull request has changes conflicting with the target branch.

• gpu/vk/vk_draw.cpp

View command line instructions

Manual merge helper

Use this merge commit message when completing the merge manually.

Merge commit title fix(vk): the start-up bar counts the compute pipelines, not 100% through them (#256)

Merge commit body Reviewed-on: https://git.magicbean.dev/MagicBeanDev/Longshot/pulls/256

fix(vk): the start-up bar counts the compute pipelines, not 100% through them (#256) Reviewed-on: https://git.magicbean.dev/MagicBeanDev/Longshot/pulls/256 Copy merge commit message

Checkout

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

git fetch -u 'fix/startup-bar-compute:gab9281-fix/startup-bar-compute'

git switch 'gab9281-fix/startup-bar-compute'

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

  

beanstalk: approved @ 1697603

Set by Beanstalk for show; the farm's database is what counts

  

beanstalk: approved @ 2c5a62e

Set by Beanstalk for show; the farm's database is what counts

  

beanstalk: approved @ 4fd82b7

Set by Beanstalk for show; the farm's database is what counts

  

beanstalk: approved @ 64d6d09

Set by Beanstalk for show; the farm's database is what counts

  

beanstalk: approved @ 6f91fae

Set by Beanstalk for show; the farm's database is what counts

  

beanstalk: approved @ 70a78f9

Set by Beanstalk for show; the farm's database is what counts

  

beanstalk: approved @ 8d361dc

Set by Beanstalk for show; the farm's database is what counts

  

beanstalk: approved @ 9d8356e

Set by Beanstalk for show; the farm's database is what counts

  

beanstalk: approved @ abf2227

Set by Beanstalk for show; the farm's database is what counts

  

beanstalk: approved @ d82f202

Set by Beanstalk for show; the farm's database is what counts

  

beanstalk: approved @ e5f1e40

Set by Beanstalk for show; the farm's database is what counts

  

beanstalk: approved @ ee8ab5f

Set by Beanstalk for show; the farm's database is what counts

  

ci: approved

A maintainer approved CircleCI for this fork PR (sticky; removed after 14 quiet days)

  

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

Affects Linux (a change for every system carries both Windows and Linux)

  

platform/macos

macOS (Apple silicon or Intel Macs)

  

platform/steam-deck

Steam Deck

  

platform/windows

Affects Windows (a change for every system carries both Windows and Linux)

  

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

  

regression

Worked in an earlier build, broken in a newer one: highest priority, above other bugs

  

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)

  

ci: approved

No labels

area/accessibility

area/audio

area/input

area/install

area/menu

area/renderer

area/saves

area/upscaling

beanstalk: approved @ 1697603

beanstalk: approved @ 2c5a62e

beanstalk: approved @ 4fd82b7

beanstalk: approved @ 64d6d09

beanstalk: approved @ 6f91fae

beanstalk: approved @ 70a78f9

beanstalk: approved @ 8d361dc

beanstalk: approved @ 9d8356e

beanstalk: approved @ abf2227

beanstalk: approved @ d82f202

beanstalk: approved @ e5f1e40

beanstalk: approved @ ee8ab5f

ci: approved

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/macos

platform/steam-deck

platform/windows

preview

priority

critical

priority

high

priority

low

regression

status

blocked

status

confirmed

status

duplicate

status

invalid

status

needs-info

status

wontfix

testing wanted

ci: approved

Milestone

Clear milestone

No items

No milestone

Projects

Clear projects

No items

No project

Assignees

Clear assignees

No assignees

3 participants

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!256

No description provided.