Repository navigation
Conversation
|
Thanks for tracking this down. At
|
|
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.
|
8b088b9 to
ad5089d
Compare
|
Rebased onto current
The two Windows check failures were also fixed:
Verification on the rebased head (RTX 4080 SUPER, Vulkan 1.4.363):
Not run on Windows or a PS5; the CI Windows job will cover the former. AI-assisted: yes. |
✅ NVIDIA GPU test: passedPR head
Unstable on main (2, these also failed in recent main runs on this machine; not counted against this PR)Failing on main too (3, not counted against this PR)Validated but not compared (1, not in the main baseline's validation pass)Validation IDs already present on main (5 IDs in 483 tests)Skipped (45)ReproduceAutomated build and test on a contributor's NVIDIA machine; it does not replace upstream CI. A new comment is posted for each new push. Full logs. |
ad5089d to
48d4ad2
Compare
|
Rebased onto current The conflict was main's new 1D-array layer-row work against this flatten path. Focused |
✅ NVIDIA GPU test: passedPR head
Unstable on main (3, these also failed in recent main runs on this machine; not counted against this PR)Failing on main too (3, not counted against this PR)Validated but not compared (1, not in the main baseline's validation pass)Validation IDs already present on main (4 IDs in 512 tests)Skipped (44)ReproduceAutomated build and test on a contributor's NVIDIA machine; it does not replace upstream CI. A new comment is posted for each new push. Full logs. |
48d4ad2 to
9bf22f3
Compare
|
Rebased onto current Focused AGC/shader tests with
AI-assisted: yes (Claude Code). |
✅ NVIDIA GPU test: passedPR head
Unstable on main (1, these also failed in recent main runs on this machine; not counted against this PR)Failing on main too (3, not counted against this PR)Validated but not compared (1, not in the main baseline's validation pass)Validation IDs already present on main (4 IDs in 525 tests)Skipped (49)ReproduceAutomated build and test on a contributor's NVIDIA machine; it does not replace upstream CI. A new comment is posted for each new push. Full logs. |
9bf22f3 to
588e87b
Compare
|
Rebased onto current
|
|
Rebased onto current
Verified locally (Linux x86-64, GCC 16.2.1, Release, RTX 4080 SUPER) with |
|
A re-run on RDNA 4 with AMD's Windows driver, at the current head: the new Windows 11 Pro 10.0.22631, AMD Radeon RX 9070 XT (driver 32.0.32015.7029), MSYS2 UCRT64 g++ 16.1.0, Release,
A skip like the one in AI-assisted: yes (Claude Code) |
|
Added the 8192-layer skip in Verified:
|
✅ NVIDIA GPU test: passedPR head
Unstable on main (3, these also failed in recent main runs on this machine; not counted against this PR)Failing on main too (3, not counted against this PR)Validated but not compared (1, not in the main baseline's validation pass)Validation IDs already present on main (13 IDs in 546 tests)Skipped (49)ReproduceAutomated build and test on a contributor's NVIDIA machine; it does not replace upstream CI. A new comment is posted for each new push. Full logs. |
✅ NVIDIA GPU test: passedPR head
Failing on main too (3, not counted against this PR)Validated but not compared (1, not in the main baseline's validation pass)Validation IDs already present on main (13 IDs in 546 tests)Skipped (49)ReproduceAutomated build and test on a contributor's NVIDIA machine; it does not replace upstream CI. A new comment is posted for each new push. Full logs. |
✅ AMD GPU test: passedPR head
Unstable on main (1, these also failed in recent main runs on this machine; not counted against this PR)Failing on main too (8, not counted against this PR)Validated but not compared (1, not in the main baseline's validation pass)Validation IDs already present on main (8 IDs in 546 tests)Skipped (60)ReproduceAutomated build and test on a contributor's AMD machine; it does not replace upstream CI. A new comment is posted for each new push. Full logs. |
|
Merged current Focused checks on the merged head, with
The previous |
✅ NVIDIA GPU test: passedPR head
Unstable on main (4, these also failed in recent main runs on this machine; not counted against this PR)Failing on main too (3, not counted against this PR)Validated but not compared (1, not in the main baseline's validation pass)Validation IDs already present on main (9 IDs in 557 tests)Skipped (54)ReproduceAutomated build and test on a contributor's NVIDIA machine; it does not replace upstream CI. A new comment is posted for each new push. Full logs. |
✅ AMD GPU test: passedPR head
Unstable on main (1, these also failed in recent main runs on this machine; not counted against this PR)Failing on main too (7, not counted against this PR)Validated but not compared (1, not in the main baseline's validation pass)Validation IDs already present on main (4 IDs in 557 tests)Skipped (66)ReproduceAutomated build and test on a contributor's AMD machine; it does not replace upstream CI. A new comment is posted for each new push. Full logs. |
23173ee to
bfb13db
Compare
|
Rebased onto current Focused tests on this machine (NVIDIA RTX 4080 SUPER, AI-assisted: yes. |
|
Windows and NVIDIA Blackwell run of the current head: head
No title was run. AI-assisted: yes. Claude Code ran the builds and tests and drafted this comment; a separate review agent checked it. The numbers are from the runs on this machine. |
✅ NVIDIA GPU test: passedPR head
Unstable on main (3, these also failed in recent main runs on this machine; not counted against this PR)Failing on main too (3, not counted against this PR)Passed on retry (1, failed in the parallel run, passed when re-run with a 360 s timeout; not counted against this PR)Validated but not compared (1, not in the main baseline's validation pass)Validation IDs already present on main (9 IDs in 592 tests)Skipped (56)ReproduceAutomated build and test on a contributor's NVIDIA machine; it does not replace upstream CI. A new comment is posted for each new push. Full logs. |
|
Verified the current head |
✅ AMD GPU test: passedPR head
Failing on main too (9, not counted against this PR)Passed on retry (1, failed in the parallel run, passed when re-run with a 360 s timeout; not counted against this PR)Validated but not compared (1, not in the main baseline's validation pass)Validation IDs already present on main (4 IDs in 592 tests)Skipped (64)ReproduceAutomated build and test on a contributor's AMD machine; it does not replace upstream CI. A new comment is posted for each new push. Full logs. |
|
The three review points are addressed at head
Verified locally on the RTX 4080 SUPER with |
# Conflicts: # core/libs/prx/libSceAgcDriver/Execution/src/VulkanDevice.cpp
|
Merged Verification on this host (NVIDIA RTX 4080 SUPER,
|
What
A 1D image array with more slices than
maxImageArrayLayerscould not be bound:ArrayLayerWindoweither threw or exposed only a window of it. God of War Ragnarök's present compute shader uses a256x1, 5120-layer, linear R8 1D array withimage_load_mip; on this devicemaxImageArrayLayersis 2048.This treats a 1D array that is only loaded or stored (not sampled) as the equivalent 2D image of
height = layers:PrepareImageModes, a non-sampled 1D array has its base dimension changed toDim2Dandflattened1DArrayset before SPIR-V emission. The plan, the specialized module and the binding all use the 2D binding (the binding number isbase + dimension, which is why a runtime mode that changed only the mode index produced mismatched bindings).DescriptorBindingcarries a per-elementimageFlattened1DArrayflag; the driver flattens only those elements, not every Image2D-over-1D-array window (the existing window path still serves sampled arrays and 2D-instruction/1D-T# views).height = last - base + 1, base address advanced by the viewed slices,depth = 0,baseArray = 0); anything else still throws.Limitations, also in the debt entry: sampled 1D arrays are not flattened (sampling would filter across layers); a loaded 1D array with mipmaps, tiled storage or DCC now throws instead of using the window path; 2D arrays past the limit remain unsupported. No console measurement exists for sampling a flattened array.
Tested
Linux x86-64 (CachyOS), GCC 16.2.1, Release, NVIDIA GeForce RTX 4080 SUPER (driver 615.78.8.0):
agc_driver_graphics(includes the new flatten unit test: base and windowed flatten),agc_driver_image_array_window_past_layer_limit,agc_driver_image_array_sample_on_volumeandagc_driver_image_packed_horizontal_gatherpass withANYPS5_REQUIRE_VULKAN=1,ANYPS5_NO_SHADER_CACHE=1andVK_INSTANCE_LAYERS=VK_LAYER_KHRONOS_validation; 0VUID-lines.image view of more array slices than maxImageArrayLayersandspecialized module bindings disagree with the prepared planfailures are gone. Together with fix(shader): declare compute invocation builtins used by lowering #14870 (compute builtins) the run advances to a different, unrelated blocker,vertex shader does not export position.tools/check_conventions.py --base origin/mainandgit diff --checkare clean;tools/pr_overlap.py --branch feat/shader-flatten-large-1d-arraysreports no overlap with open pull requests.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)