Repository navigation
docs(agc): note that addrlib and fpPS4 take the 2D PRT modes from the whole coordinates - #13691
Conversation
… whole coordinates The PS4 tiling entry said pipe and bank of every PRT mode come from the coordinates inside the macro tile. It now says so for GB_TILE_MODE 11 and 16, and a new entry names the four 2D PRT modes (6, 7, 12, 17), where this code follows shadPS4 and AMD's addrlib and fpPS4 use the whole coordinates. Nothing is claimed about the console. Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
MarcZX10
left a comment
MarcZX10
left a comment
There was a problem hiding this comment.
Reviewed — changes requested
The PR records an overlap with #9674 in the same debt file but leaves Depends on: none without recording an agreed landing order, contrary to the repository's overlap rule.
What I ran
Reviewed head f33d726734bf568270668ef24a184efc5a81bcde against PR base f25279c0ca4cb068370df4baaa46de7b4f4741c7. The stack check previewed it cleanly against current main 93c71cc1bbe43b5e9e9d33fdded98e77efacbdba; the changed file overlaps #9674, whose current head is not included. Environment: Debian Linux x86-64, Python 3.11.2. This PR changes documentation only; no build or tests were run.
python3 tools/pr_precheck.py --repo boykopovar/AnyPS5 --pr 13691
python3 tools/pr_stack_check.py 13691 --repo boykopovar/AnyPS5 --output-dir /tmp/pr13691-stack
gh pr checks 13691 --repo boykopovar/AnyPS5Results
| Check | Result |
|---|---|
| Patch and conventions | Passed; git diff --check passed and repository conventions passed. |
| Current-main merge preview | Clean as an individual PR. |
| Overlap coordination | The stack report identifies #9674 as an overlapping dependency candidate in docs/dev/debt/libSceAgcDriver.md; its head is not included in #13691. The PR description also says the second landing will need a rebase, but does not record an agreed order. |
| Hosted CI | Conventions, labels, eligibility, and intake passed. Build and specialized jobs were skipped. |
| Tests | Not applicable to this documentation-only diff. |
| Precheck helper | The isolated checkout lacked the workspace-only test-quality script; no tests changed, so that heuristic does not apply. |
Unedited outputs
All conventions checks passed.
# PR stack check — fresh
Target: https://github.com/boykopovar/AnyPS5/pull/13691
Target branch tip: `93c71cc1bbe43b5e9e9d33fdded98e77efacbdba`
All heads unchanged: **True**
Dependency order: #9674 → #13691
## Individual merge previews
- PR #9674: **clean** against current base; head `297118f83bccf69aeef9fbe54a51a15d4feaef81`; changed paths: none reported
- PR #13691: **clean** against current base; head `f33d726734bf568270668ef24a184efc5a81bcde`; changed paths: none reported
## Dependency relationships
- #9674 → #13691: **current_dependency_head_not_included**; overlapping changed paths: `docs/dev/debt/libSceAgcDriver.md`
conventions pass 17s https://github.com/boykopovar/AnyPS5/actions/runs/41148281893/job/124378428284
labels pass 7s https://github.com/boykopovar/AnyPS5/actions/runs/41148282004/job/124378428820
linux skipping 0 https://github.com/boykopovar/AnyPS5/actions/runs/41148282088/job/124378823026
windows skipping 0 https://github.com/boykopovar/AnyPS5/actions/runs/41148282088/job/124378823407
eligibility pass 6s https://github.com/boykopovar/AnyPS5/actions/runs/41148282088/job/124378427938
intake pass 4s https://github.com/boykopovar/AnyPS5/actions/runs/41148281972/job/124378428351
Findings
Blocking
The repository requires authors of overlapping PRs to agree on landing order and declare that order with Depends on. This PR's overlap section identifies #9674 as changing the adjacent entry in the same file and says one of the PRs will need a rebase, while the checklist still says Depends on: none; I found no recorded agreement from #9674's author. Please coordinate the order with that author, record the agreement, and update the checklist to match (including Depends on: #9674 if that PR must land first).
Non-blocking
The new debt entry links to shadPS4's moving main branch, while the PR's explanation cites a pinned commit. Pin the debt-entry link too when convenient so the stated comparison remains reproducible.
Sources
Not verified
I did not build or run tests because the change is documentation only. I did not independently reproduce the cited Mesa addrlib analysis; the source comparison in the PR description remains author-provided evidence.
To get this to a good state
Please record the agreed #9674/#13691 order and update the checklist accordingly. I will rerun the overlap and convention checks on the updated head.
|
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.
|
|
Thanks for the review. The checklist of this PR already declares the order: the line reads On the non-blocking point: the shadPS4 link on AI-assisted (Claude Code). |
|
Checked the claims against Mesa addrlib 393bf0d ( |
What
The PS4 (GCN) tiling entry said that pipe and bank of every PRT mode come from the coordinates inside the macro tile. The pipe bits lie inside the macro tile in every mode, so only the bank differs: for GB_TILE_MODE 11 and 16 all three sources I read take the bank from the coordinates inside the macro tile, while for the four 2D PRT modes (6, 7, 12 and 17) AMD's addrlib and fpPS4 take it from the whole coordinates and shadPS4 and this code take it inside the macro tile. The entry now says so for 11 and 16, and a new entry names the difference. The entries then read:
Changed:
docs/dev/debt/libSceAgcDriver.md, one existing entry and one new one. No code change.Method: the three were read at the linked commits; only their behaviour is described, no code was taken over. AMD addrlib (Mesa)
egbaddrlib.cppL1775-L1780 reduces x and y modulo the macro tile only for the PRT modes without rotation, which its mode flags mark for PRT_TILED_THIN1 only (not PRT_2D_TILED_THIN1 or PRT_3D_TILED_THIN1); fpPS4ps4_tiling.pasL2812-L2816 does the same for its two non-2D PRT array modes and documents the 2D ones as not aliasing virtual pages; shadPS4tiling.compL392-L396 reduces all PRT modes. In this codeGcnTiling(TextureTiling.cpp) does so for all six PRT modes, whichGuestTextureResource.cppmaps from GB_TILE_MODE 6, 7, 11, 12, 16 and 17. PS4 only (surfaces withgcnPitch); not measured on a console, and no title that uses these modes has been run.Overlap
Open PRs that change
docs/dev/debt/libSceAgcDriver.md(file sets fromgit diff --name-onlyagainst each PR head's merge base, all 1585 open pull requests as of 2026-10-30 22:00 UTC, main9605300f29): 134 touch the file or are ours. One gives a new conflict with this change: #9674 rewrites the thick and layered surfaces entry, which sits on the line directly below the new entry, so whichever merges second needs a one-line rebase; the texts do not overlap. #9891, #9892 and #9895 conflict withmainitself and give the same result with and without this commit. The others give the same merge result with and without it. Our other PRs (#12742, #12765, #12797) have no conflict with it. Checked again on 2026-10-30 at 23:49 UTC againstmainac551a129a(1565 open pull requests): this head merges cleanly, #9674 is still the only new conflict, and of the 28 pull requests opened since 21:55 UTC, #13667 and #13678 change the file and merge with and without this commit.Tested
Documentation only.
tools/check_conventions.py --base upstream/mainpasses; the changed entry is 990 characters and the new one 518; based onmain9605300f29, and it merges withmainb44985a3eawithout a conflict (checked 2026-10-30 22:56 UTC). Not built, not run.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)AI-assisted (Claude Code).
🤖 Generated with Claude Code