[fma-rc]

What

Implements sceAgcAcbSetFlip. It writes the same 6-dword flip packet as sceAgcDcbSetFlip. In the PS5 6.00 libSceAgc module both exports are the same 176-byte wrapper around one libSceAgcDriver helper; the only argument that differs selects the release event (CS_DONE for ACB), which our flip packet does not model.

The second commit lets the driver execute that packet on a compute queue. Until now the PM4 validator (Pm4.cpp) accepted a flip only on queue 0 and threw graphics packet in compute queue otherwise. The flip path after validation already handles any queue: it reserves the output, hands it the frame and raises the end-of-pipe interrupt on the submitting queue, the way other compute-queue releases are delivered. Since two workers can now reach the flip, the frame serial became atomic, and a compute flip gets its own frame timing instead of taking queue 0's.

Earlier attempts #1052 and #1585 were closed because no title was shown to reach it. Open PR 6447 writes the same packet for this function among changes to 100 files; this PR is only ACB flips. Death Stranding 2 (PPSA02015) stops on the export on its loading screen, about 1085 s in:

terminate called after throwing an instance of 'std::runtime_error'
  what():  sceAgcAcbSetFlip not implemented

and, with the first commit alone, at 1333 s on the packet it writes:

terminate called after throwing an instance of 'std::runtime_error'
  what():  AGC driver: FLIP at DWORD 0: graphics packet in compute queue

Tested

Linux, on main 6a32550: agc_commands fails on main with sceAgcAcbSetFlip not implemented; agc_driver_flip (new compute-queue case) fails without the second commit with AGC driver: FLIP at DWORD 0: graphics packet in compute queue. Both pass with this PR.

test	RTX 5080 (NVIDIA 615.78.08)	Radeon 610M (RADV)	lavapipe
agc_commands	pass	pass	pass
agc_driver_flip	pass	pass	pass
agc_driver_flip_stores_small_results	pass	pass	pass
agc_driver_flip_compute	pass	pass	pass

Windows (WinLibs GCC 15.2 UCRT, no GPU): the whole suite builds and runs; the one failure is one main has there too.

Death Stranding 2 on Linux, RTX 5080: with both commits the run passes 1333 s with no refused submission, stays on the loading screen and stops at 1630 s on a heap corruption abort (corrupted size vs. prev_size) that is not in this code. The title does not reach this call on plain main: it also needs the open PRs #6437, #6297, #6277, #6196, #6530, #6300, #6727, #6492 and #6174 (none of them touches this code) and local changes for EQAA targets and image-table residency that are not ready to submit.

Checklist

• Based on current main; no other open PR implements the same functions
• One topic per PR; follow-ups go in a new PR
• No comments in code except technical debt
• Unimplemented paths throw (NotImplemented_nid_no_patch); silent stubs are listed under Silent stubs in their library's technical debt file
• No notes, investigation .md files or images added to the repository (attach them to this PR)
• New third-party code is a submodule built from source
• Generic behaviour, not specific to one title
• Depends on: none
• Unblocks: PPSA02015
• GPUs tested: RTX 5080 (NVIDIA 615.78.08), Radeon 610M (RADV, Mesa), lavapipe; Arch Linux
• AI-assisted: yes

[github-actions]

Progress report (6a32550 → 66cc672)

📈 System libraries: 89.46% (+0.02%, +1 implemented functions)

✅ 1 implemented

Library	Name
libSceAgc	sceAgcAcbSetFlip

[MarcZX10]

The conventions failure is only the PR-template Depends on field: the validator requires Depends on: #N or Depends on: none directly on the checklist line. This PR lists integration-only prerequisites in prose, so please keep the checklist line in the accepted form (likely Depends on: none, since you state there is no code dependency) and move the runtime/integration context to the description above. The failed run reported this as its sole finding; eligibility, intake and progress report pass.

[github-actions]

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.

Pull request	Shared files not compared
PR 4448	core/libs/prx/libSceAgcDriver/Execution/src/Driver/Packets/PacketExecution.cpp
PR 5115	core/libs/prx/libSceAgcDriver/Execution/src/Driver/Packets/PacketExecution.cpp
PR 5431	core/libs/prx/libSceAgcDriver/Execution/src/Driver/Packets/PacketExecution.cpp
PR 5925	core/libs/prx/libSceAgcDriver/Execution/src/Driver/Packets/PacketExecution.cpp
PR 6188	core/libs/prx/libSceAgcDriver/Execution/src/Driver/Packets/PacketExecution.cpp core/libs/prx/libSceAgcDriver/Execution/src/Pm4.cpp
PR 6292	core/libs/prx/libSceAgcDriver/Execution/src/Driver/Packets/PacketExecution.cpp
PR 6501	core/libs/prx/libSceAgcDriver/Execution/src/Driver/Packets/PacketExecution.cpp
PR 6549	core/libs/prx/libSceAgcDriver/Execution/src/Pm4.cpp
PR 6615	core/libs/prx/libSceAgcDriver/Execution/src/Driver/Packets/PacketExecution.cpp
PR 6946	core/libs/prx/libSceAgcDriver/Execution/src/Driver/Packets/PacketExecution.cpp
PR 6947	core/libs/prx/libSceAgcDriver/Execution/src/Driver/Packets/PacketExecution.cpp
PR 7642	core/libs/prx/libSceAgcDriver/Execution/src/Driver/Packets/PacketExecution.cpp

[fma-rc]

Thanks, fixed: the checklist line is now Depends on: none and the prerequisites are listed in the description. The conventions check passes on the edit.

[adriadam10]

Checked the ACB flip packet against sceAgcDcbSetFlip, the compute-queue flip path (no flip hold on non-zero queues, own frame timing, EOP raised on the submitting queue) and the new agc_commands / agc_driver_flip cases, which fail on main with the errors quoted in the description; merged.