Repository navigation
fix(agc): check GPU-side indirect dispatch group counts against the device limit - #6360
Merged
adriadam10 merged 3 commits intoOct 19, 2026
Merged
Conversation
…evice limit A DISPATCH_INDIRECT read on the GPU took its group counts unchecked. A count buffer that a skipped producer never wrote (PPSA08427: 0xffffffff left by skipped depth-plane passes) launched up to 2^32 groups and kept the GPU busy until the title hung. The CPU path already refuses counts above maxComputeWorkGroupCount. - Recorder::RecordDispatchArguments copies the three counts from the import into a 12-byte indirect buffer with a one-thread kernel (DispatchArguments.comp), all three 0 when one exceeds the device limit, as RecordMeshArguments does for mesh indirect records. - VulkanDevice records the dispatch from that buffer; without the kernel the import is read directly, as before. - agc_driver_indirect_dispatch_limit: 0xffffffff groups run nothing, one group still runs. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Q8tinbP1MUpUaDsbNWnGnu
This was referenced Oct 15, 2026
Merged
This was referenced Oct 15, 2026
Closed
|
Other open pull requests touch the same code. Merging one will break the other, or both implement the same thing. If this one builds on them, list them in Depends on.
Some comparisons could not be completed because a pull request conflicts with the base branch. Review these shared files manually before concluding that there is no overlap.
|
11 tasks done
12 tasks done
This was referenced Oct 17, 2026
Contributor
|
Intel and AMD Windows data. On an Intel UHD 770 this fixes the device loss. On an AMD RX 9070 XT the new test fails with and without it, because that device allows 0xffffffff groups in X. Windows 11 Pro 26300, WinLibs GCC 15.2.0 r7, Release, every execution test built, Reported
On the RX 9070 XT, Full CTest ( AI-assisted: yes (Claude Code) |
Merged
11 tasks done
adriadam10
requested changes
Oct 18, 2026
adriadam10
left a comment
adriadam10
left a comment
Collaborator
There was a problem hiding this comment.
Thanks, the approach (clamping the GPU-side counts through a one-thread kernel, same as the mesh arguments guard) looks right. Two things before this can merge:
core/libs/prx/libSceAgcDriver/tests/execution/IndirectDispatchLimit.cpp:65: the over-limit row is a fixed{0xffffffff, 1, 1}. On devices whose X limit is0xffffffff(RX 9070 XT reports[4294967295, 65535, 65535]) that count is valid, so the guard correctly lets it through and the test fails with and without the change, as @savvaniss measured. Derive the over-limit count from the device limits (e.g.maxComputeWorkGroupCount[1] + 1in Y, or X + 1 when X < 0xffffffff).- The branch conflicts with
maininExecution/src/VulkanDevice.cpp(recordDispatch) andlibSceAgcDriver/CMakeLists.txt. TheVulkanDevice.cpphunk is not a mechanical conflict, so please rebase and re-check that the indirect buffer/offset still feedvkCmdDispatchIndirectcorrectly.
Not covered by what we can run: the hang on the title after the rebase.
recordDispatch now reads its indirect offset through HostImportOffset, which adds the import's lead. The unchecked path passes that offset to vkCmdDispatchIndirect; the clamp kernel reads from import.address plus the guest offset, since import.address already includes the lead. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The over-limit row was a fixed 0xffffffff in X, which is a valid count on devices that report that X limit (RX 9070 XT). Take the first axis whose limit is below 0xffffffff and use limit + 1 there; skip when no count can exceed the limits. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
adriadam10
approved these changes
Oct 19, 2026
TheBlenDed
pushed a commit
to TheBlenDed/AnyPS5
that referenced
this pull request
Oct 25, 2026
…atch group counts against the device limit) at e54d2f3
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
What
A
DISPATCH_INDIRECTwhose arguments stay on the GPU (vkCmdDispatchIndirecton the imported argument buffer) took its group counts unchecked. The CPU path already refuses counts abovemaxComputeWorkGroupCount, but the GPU path passed garbage straight to the device: a count buffer that a skipped producer never wrote (PPSA08427 leaves0xffffffffthere when its depth-plane passes are skipped) launched up to 2^32 groups and kept the GPU busy until the title hung.Recorder::RecordDispatchArgumentscopies the three counts from the import into a 12-byte indirect buffer with a one-thread kernel (Graphics/shaders/DispatchArguments.comp), writing all three as 0 when one exceeds the device limit, the same wayRecordMeshArgumentsalready guards mesh indirect records.VulkanDevice::recordDispatchdispatches from that buffer. If the kernel cannot be created, the import is read directly, as before, and stderr says so once.agc_driver_indirect_dispatch_limit:0xffffffffgroups run no thread; one group still runs every thread.Tested
Linux (CachyOS, GCC Release, NVIDIA GeForce RTX 3060): the new test passes; the AGC group of
ctestpasses exceptagc_driver_store_over_untracked, which fails on unmodifiedmaintoo.tools/check_conventions.pypasses;tools/pr_overlap.py --branchreports no line overlap. The hang was found running PPSA08427 on my local branch, where this change removed it; the game was not rerun on this exact rebase.Checklist
main; no other open PR implements the same functionsNotImplemented_nid_no_patch); silent stubs are listed under Silent stubs in their library's technical debt file.mdfiles or images added to the repository (attach them to this PR)🤖 Generated with Claude Code
https://claude.ai/code/session_01Q8tinbP1MUpUaDsbNWnGnu