build(release): 38 contributed pipeline lists from #14, 12,001 pipelines; COMPUTE_MOST raised to 3584 #1688
No reviewers
Labels
No labels
area/accessibility
area/audio
area/input
area/install
area/menu
area/renderer
area/saves
area/tooling
area/upscaling
beanstalk: approved
beanstalk: failed
beanstalk: needs approval
beanstalk: running
beanstalk: tested
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
risk
high
risk
low
risk
medium
status
blocked
status
confirmed
status
duplicate
status
invalid
status
needs-info
status
wontfix
ci: approved
No milestone
No project
No assignees
1 participant
Notifications
Due date
No due date set.
Dependencies
No dependencies set
Reference
MagicBeanDev/Longshot!1688
Loading…
Reference in a new issue
No description provided.
Delete branch "HumanBean/Longshot:build/pipelines-from-14"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
Type: build · Risk: low · Tested on: none (the sources' check only) · Needs: a first start on any GPU
TL;DR
Since the fork, 40 players have attached the pipeline lists their game wrote to the Submit your pipelines issue, and none of them was in the release yet. This PR adds 38 of them (two held nothing new), so the first start prepares 12,001 more pipelines before the title: 75,921 in all, up from 63,920. That should mean fewer stutters the first time a player reaches an area. The lists pass the repository's own check, and the merge stays deterministic.
Breakdown
contrib/lists with their notes; the merged shipped list grows by 12,001 pipelines (10,625 graphics, 266 mesh, 1,110 compute programs), andCOMPUTE_MOSTgoes from 2560 to 3584 so the check accepts the 3,309 compute programs the merge now names.nx2/pipeline_states.py --checkbeforeadd, andnx2/pipeline_states.py checkpasses on the result.Why
Issue #14 asks players to send the list their game wrote, and promises that we'll merge each one with
nx2/pipeline_states.py. Forty lists arrived between 2026-11-13 and today, from NVIDIA (GTX 1650 to RTX 5090 D), AMD (RX 6600 to RX 9070 XT, an 890M) and Intel Arc B370, on Windows and Linux, at resolutions from 1536x864 to 5120x1440 and with every upscaler. Several cover a full playthrough or most of the dungeons. See #14.What changed
Each list was added with
nx2/pipeline_states.py add <name> <list> --who "<user> (#14)" ..., in the order the comments were posted. Each new source therefore holds only the pipelines that no list before it has. The notes' GPU, build and settings come from the report the game wrote into each zip (longshot_pipeline_report.txt), and the route comes from the player's comment. Where a field was left blank it is marked "not given".assets/pipeline_states/contrib/: 38<user>-<gpu>.bin.xzlists and their.mdnotes (table below).nx2/pipeline_states.py:COMPUTE_MOST2560 → 3584, with a comment giving the reason and the estimated cost.assets/pipeline_states/README.md: the limit's number.Not added: MB44's (292 pipelines) and piano_butcher's (142) lists held nothing that the earlier lists don't already have.
How the zips were handled: each attachment was downloaded into its own folder and its members were listed. All 40 held only
*_pipeline_states.bin(and, in most, the report text). Only those two members were read, nothing was run, and each list passed--check(exactly a version 2 list, within the sent-list limits) beforeadd.albeth-rtx-4070-superap1017-rtx-4090berzercker-rx-9070-xtbowen26223-rtx-5090-dchuchothecatlord-rx-6950-xtdavyn-rx-6600dinotoxic-rtx-3060-tidnf113abc-arc-b370gaborne-rx-9070-xt-2gaborne-rx-9070-xtginkumasho-rtx-4060-laptopgoulagman-rtx-5070haplessidiot-rtx-3070haruhiro88-rtx-5070idipzy-rx-6750-xt175639923122.04~b84bab8jabbaxipil-rtx-4070losava-rtx-3080mr-r00k-rtx-3080-timustardtiger-rx-9070-xtpoty87-rtx-3090razytos-rx-9070-xtrealredrich-890mrhinomcd-rtx-3080-tirnc777-rtx-5070-tisalvy931-rx-9070-xtsenhorcotrim-rx-7600silverxp-rtx-4060-tisneakymiata-rtx-3050-laptopthebobdoleexperience-rtx-4070-titreverflume-rtx-3060-laptoptrulyneochaos-rtx-4070-laptopveridor-rx-9070wasabitws-gtx-1650watauro-rtx-3060x-yohe-x-rx-6600xerphyl-rtx-3070-tiyeetbot90-rx-9070-xtzethryder-rtx-2070Measurements
Verdict: nothing measured in the game. The counts below come from
nx2/pipeline_states.py check.10fab1520623986aFlags and settings
None. (
COMPUTE_MOSTis a bound in the build tool's check, not a switch.)Testing
Tested:
nx2/pipeline_states.py --checkon each of the 40 sent lists (all version 2, all accepted), thennx2/pipeline_states.py checkon the sources after the adds (passes,merged: 75921 pipelines (compute 3309, graphics 68312, mesh 4300) of 27125 programs, version 2). The test suite was not run on this PC.Not tested: a game run. Opened with
--forceat the maintainer's request: no game code changes, only the data the first start prepares.Must be tested:
@Beanstalka cold first start of this branch against main, preparation time and memory at the title).Risks
What could break: the first start's preparation takes longer with 12,001 more pipelines. The 1,110 more compute programs are kept ready on every start: by the 2026-11-12 measurement in
nx2/pipeline_states.pythat is about 1 MB and a few ms of precompile each, so roughly 1.1 GB more memory and 2 to 3 s more per start. That matters most on 8 GB cards and integrated GPUs. If it turns out too costly, prune the compute records rather than the graphics ones.Turning it off: revert the commit, or lower
COMPUTE_MOSTand drop the newest lists.Compatibility: a player's first start after updating prepares the new pipelines once. Saves, options and caches are otherwise unchanged.
Issues
See #14 (it stays open for more lists).
Checklist
Everyone
type(scope): description) of the status line's type.can't be run by hand is in Testing's Not testable by hand, with why.
in Risks.
main.python -I -B tests/run_all.pypasses (any failure that also fails onmainis named).third_party/untouched (a new folder with aPROVENANCE.mdinstead).When it applies
Assisted-by:/Co-Authored-By:trailers.🤖 Generated with Claude Code
WIP: build(release): 38 contributed pipeline lists from #14, 12,001 pipelines; COMPUTE_MOST raised to 3584to build(release): 38 contributed pipeline lists from #14, 12,001 pipelines; COMPUTE_MOST raised to 3584