From 63d39623e351bb536c525bd01ca2327e1e046fae Mon Sep 17 00:00:00 2001 From: milescoviello Date: Wed, 21 Oct 2026 16:09:50 +0000 Subject: [PATCH 1/9] fix(build): run guest_unused_import_dependencies without the ASan link order check The test loads guest_module_args_host into a relinked guest, so in an ANYPS5_ENABLE_ASAN build the ASan runtime is not first in the initial library list and the guest stops before it runs. The other relinker tests that load that host already take verify_asan_link_order=0 through add_asan_test_options; this one was added after them. Co-Authored-By: Claude Opus 5.5 --- core/relinker/CMakeLists.txt | 1 + 1 file changed, 1 insertion(+) diff --git a/core/relinker/CMakeLists.txt b/core/relinker/CMakeLists.txt index fa78fbaa42..6d0a7af013 100644 --- a/core/relinker/CMakeLists.txt +++ b/core/relinker/CMakeLists.txt @@ -622,6 +622,7 @@ if(BUILD_TESTING) add_asan_test_options(guest_module_args verify_asan_link_order=0) add_asan_test_options(guest_deferred_dependencies verify_asan_link_order=0) add_asan_test_options(guest_module_fini verify_asan_link_order=0) + add_asan_test_options(guest_unused_import_dependencies verify_asan_link_order=0) add_test(NAME guest_needed_nested_modules COMMAND ${Python3_EXECUTABLE} ${CMAKE_CURRENT_SOURCE_DIR}/relinker/tests/test_guest_needed_nested_modules.py $) add_test(NAME guest_zero_initializer COMMAND ${Python3_EXECUTABLE} From 1f87b51ea5ad473e5a0b51b30f58d5ae3736f205 Mon Sep 17 00:00:00 2001 From: milescoviello Date: Wed, 21 Oct 2026 16:32:20 +0000 Subject: [PATCH 2/9] fix(agc): value-initialize a pipeline journal entry before decoding it parse() declared each LoadedEntry without an initializer. Several State members (hasColorTarget, topology, viewport, negativeOneToOne, cullMode, frontFace, blend) have no default member initializer, so the graphics recipe of a compute record, which is never decoded, held indeterminate values. Moving the entry into the loaded list then read them: UBSan reports "load of value 6, which is not a valid value for type 'bool'" in State's move constructor in agc_driver_pipeline_journal. Co-Authored-By: Claude Opus 5.5 --- core/libs/prx/libSceAgcDriver/Graphics/src/PipelineJournal.cpp | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/core/libs/prx/libSceAgcDriver/Graphics/src/PipelineJournal.cpp b/core/libs/prx/libSceAgcDriver/Graphics/src/PipelineJournal.cpp index d0669ed238..e18e5bb667 100644 --- a/core/libs/prx/libSceAgcDriver/Graphics/src/PipelineJournal.cpp +++ b/core/libs/prx/libSceAgcDriver/Graphics/src/PipelineJournal.cpp @@ -621,7 +621,7 @@ private: if (decoded) loaded.indexModules.insert(identity); } else if (record.kind == GraphicsRecord || record.kind == ComputeRecord) { if (decode && !loaded.indexRecords.contains(record.payloadHash)) { - LoadedEntry entry; + LoadedEntry entry{}; entry.kind = record.kind; bool decoded = false; if (record.kind == GraphicsRecord) { From 789a5f686ffd1c26ba83c6de412ebbb27d23608f Mon Sep 17 00:00:00 2001 From: Evgeniy Kazarin Date: Sun, 18 Oct 2026 06:28:39 +0300 Subject: [PATCH 3/9] perf(agc): defer small host imports until first GPU use --- .../Graphics/src/GuestBufferMemory.cpp | 5 ++- .../prx/libSceAgcDriver/tests/Recorder.cpp | 2 +- .../tests/execution/HostImportWindows.cpp | 44 +++++++++++++++++++ 3 files changed, 48 insertions(+), 3 deletions(-) diff --git a/core/libs/prx/libSceAgcDriver/Graphics/src/GuestBufferMemory.cpp b/core/libs/prx/libSceAgcDriver/Graphics/src/GuestBufferMemory.cpp index 074681e607..053fb10c01 100644 --- a/core/libs/prx/libSceAgcDriver/Graphics/src/GuestBufferMemory.cpp +++ b/core/libs/prx/libSceAgcDriver/Graphics/src/GuestBufferMemory.cpp @@ -3441,10 +3441,11 @@ bool ImportMappedRanges(const Context& context, const GuestAllocations::Mapped& } else if (state.refreshedGeneration > generation) { return true; } - if (!adopting) current = GuestAllocations::GuestAllocationsAcquire_nid_postfix(); for (const auto& mapped : ranges) { const auto range = mapped.lock(); - if (range == nullptr || leasedRangeOwner(current, range->address) != range) continue; + if (range == nullptr || range->bytes <= 0x10000) continue; + if (current.empty()) current = GuestAllocations::GuestAllocationsAcquire_nid_postfix(); + if (leasedRangeOwner(current, range->address) != range) continue; try { if (!GuestMemory::GpuMutex().HeldByThisThread()) { const auto run = importRun(current, range->address, std::numeric_limits::max(), effectiveBudget(context, state)); diff --git a/core/libs/prx/libSceAgcDriver/tests/Recorder.cpp b/core/libs/prx/libSceAgcDriver/tests/Recorder.cpp index 97f02e7f2f..45df34ee97 100644 --- a/core/libs/prx/libSceAgcDriver/tests/Recorder.cpp +++ b/core/libs/prx/libSceAgcDriver/tests/Recorder.cpp @@ -2371,7 +2371,7 @@ void remappedImportTests(const Device& device) { void hostImportRetirementTests(const Device& device) { const auto& context = device.GetContext(); - constexpr std::size_t UnitBytes = 65536; + constexpr std::size_t UnitBytes = 2 * 65536; constexpr std::size_t AllocationBytes = 4 * UnitBytes; constexpr std::size_t CopyBytes = 256; #ifdef _WIN32 diff --git a/core/libs/prx/libSceAgcDriver/tests/execution/HostImportWindows.cpp b/core/libs/prx/libSceAgcDriver/tests/execution/HostImportWindows.cpp index e6d9bd6b30..35e4f777cd 100644 --- a/core/libs/prx/libSceAgcDriver/tests/execution/HostImportWindows.cpp +++ b/core/libs/prx/libSceAgcDriver/tests/execution/HostImportWindows.cpp @@ -21,6 +21,14 @@ #include #include +namespace AgcDriver { + +struct VulkanDeviceTestAccess { + static Graphics::Context ContextOf(const VulkanDevice& device) { return device.graphicsContext(); } +}; + +} + namespace { namespace Graphics = AgcDriver::Graphics; @@ -163,6 +171,41 @@ Outcome ResidentMappingsImportAfterCollection() { return Outcome::Passed; } +Outcome SmallMappingsImportOnFirstUse() { + GuestBlock small(GuestAlignment); + GuestBlock large(2 * GuestAlignment); + auto device = OpenVulkanTestDevice(); + if (!device) return Outcome::NoDevice; + const auto context = AgcDriver::VulkanDeviceTestAccess::ContextOf(*device); + if (context.hostImportAlignment == 0) return Outcome::NoImport; + for (auto* guest : {&small, &large}) { + Graphics::ClearHostImports(device->Device()); + GuestAllocations::Mapped ranges; + for (const auto& range : GuestAllocations::GuestAllocationsAcquire_nid_postfix()) { + if (range->address == guest->Address()) ranges.push_back(range); + } + Require(ranges.size() == 1, "mapped import test lost its registered range"); + { + std::lock_guard lock(AgcDriver::GuestMemory::GpuMutex()); + Require(Graphics::ImportMappedRanges(context, ranges, GuestAllocations::GuestAllocationsGeneration_nid_postfix(), true), "mapped import test could not adopt the device"); + } + Require(Graphics::HostImportCovers(context, guest->Address(), ProbeBytes) == (guest == &large), "mapping did not defer exactly the small import"); + Require(Fill(*device, guest->Address()), "first GPU use could not import the mapped buffer"); + Require(Graphics::HostImportCovers(context, guest->Address(), ProbeBytes), "first GPU use left the mapped buffer unimported"); + constexpr std::array pattern{0x11111111u, 0x22222222u, 0x33333333u, 0x44444444u}; + for (std::size_t offset = 0; offset < ProbeBytes; offset += sizeof(std::uint32_t)) { + std::uint32_t actual = 0; + std::memcpy(&actual, guest->Data() + offset, sizeof(actual)); + Require(actual == pattern[(offset / sizeof(actual)) % pattern.size()], "first GPU use changed the fill result"); + } + for (std::size_t offset = ProbeBytes; offset < GuestAlignment; ++offset) { + Require(guest->Data()[offset] == 0, "first GPU use overwrote bytes outside the fill"); + } + } + Graphics::ClearHostImports(device->Device()); + return Outcome::Passed; +} + } int main() { @@ -176,6 +219,7 @@ int main() { const auto windows = WindowsFollowTheirRange(); Require(windows == Outcome::Passed, "host import windows: the device stopped importing guest memory between two scenarios"); Require(ResidentMappingsImportAfterCollection() == Outcome::Passed, "resident import test failed"); + Require(SmallMappingsImportOnFirstUse() == Outcome::Passed, "mapped import test lost its Vulkan device or host import support"); std::puts("host import windows tests passed"); return 0; } catch (const std::exception& error) { From e0764b9e14cb9eece34c9807c0098b75d7613766 Mon Sep 17 00:00:00 2001 From: Evgeniy Kazarin Date: Wed, 21 Oct 2026 10:18:44 +0800 Subject: [PATCH 4/9] test(agc): map-time import tests for ranges of 64 KiB left to first GPU use The import-requester test now counts the map requester on a 128 KiB range and asserts that a 64 KiB range is not imported at map time and is imported, for the access requester, at its first use. The over-budget map test maps 128 KiB so it still reaches the budget check. Co-Authored-By: Claude Opus 5.5 --- .../libSceAgcDriver/tests/BdaResources.cpp | 31 +++++++++++++++---- 1 file changed, 25 insertions(+), 6 deletions(-) diff --git a/core/libs/prx/libSceAgcDriver/tests/BdaResources.cpp b/core/libs/prx/libSceAgcDriver/tests/BdaResources.cpp index e4ea60da83..324c9bf90e 100644 --- a/core/libs/prx/libSceAgcDriver/tests/BdaResources.cpp +++ b/core/libs/prx/libSceAgcDriver/tests/BdaResources.cpp @@ -979,7 +979,7 @@ void importNeighbouringEdgesTests(const Context& context) { // A range mapped while the imports are at the budget is refused at map time without retiring an import: the map // observer holds no GPU lock, so it must not evict (before, it evicted the least recently used import to fit). void importAtMapOverBudgetTests(const Context& context) { - constexpr std::size_t bytes = 65536; + constexpr std::size_t bytes = 2 * 65536; auto importing = context; importing.hostImportAlignment = bytes; void* kept = ::operator new(bytes, std::align_val_t{bytes}); @@ -1039,17 +1039,32 @@ void importRequesterTests(const Context& context, const BdaTestAccess& access) { const auto counted = [](ImportRequester requester) { return HostImportCounters().requesters[static_cast(requester)]; }; const auto made = [&](ImportRequester requester) { return counted(requester).made; }; Require(std::string(ImportRequesterName(ImportRequester::Map)) == "map" && std::string(ImportRequesterName(ImportRequester::Build)) == "build" && std::string(ImportRequesterName(ImportRequester::Access)) == "access" && std::string(ImportRequesterName(ImportRequester::Join)) == "join" && std::string(ImportRequesterName(ImportRequester::Segment)) == "segment", "an import requester has the wrong name"); - change(0, true); - { + const auto mappedRange = [&](std::size_t index) { GuestAllocations::Mapped ranges; const auto lease = GuestAllocations::GuestAllocationsAcquire_nid_postfix(); for (const auto& range : lease) { - if (range->address == address(0)) ranges.push_back(range); + if (range->address == address(index)) ranges.push_back(range); } Require(ranges.size() == 1, "the requester test's mapped range is not registered"); + return ranges; + }; + { + GuestAllocations::Mutation mutation; + mutation.Add(block + 6 * piece, 2 * piece, true, true, true); + } + { const auto before = made(ImportRequester::Map); - static_cast(ImportMappedRanges(importing, ranges, GuestAllocations::GuestAllocationsGeneration_nid_postfix(), true)); - Require(made(ImportRequester::Map) == before + 1 && HostImportCovers(importing, address(0), piece), "a range imported at map time was not counted for the map requester"); + static_cast(ImportMappedRanges(importing, mappedRange(6), GuestAllocations::GuestAllocationsGeneration_nid_postfix(), true)); + Require(made(ImportRequester::Map) == before + 1 && HostImportCovers(importing, address(6), 2 * piece), "a range imported at map time was not counted for the map requester"); + } + change(0, true); + { + const auto map = made(ImportRequester::Map); + const auto access = made(ImportRequester::Access); + static_cast(ImportMappedRanges(importing, mappedRange(0), GuestAllocations::GuestAllocationsGeneration_nid_postfix(), true)); + Require(made(ImportRequester::Map) == map && !HostImportCovers(importing, address(0), 16), "a range of 64 KiB was imported at map time"); + Require(HostImportFor(importing, address(0), 16) != nullptr && HostImportCovers(importing, address(0), piece) && made(ImportRequester::Access) == access + 1 && made(ImportRequester::Map) == map, + "a range of 64 KiB left to its first GPU use was not imported there, or not counted for that use's requester"); } change(1, true); { @@ -1127,6 +1142,10 @@ void importRequesterTests(const Context& context, const BdaTestAccess& access) { const auto counts = HostImportCounters(); Require(counts.live >= 4 && counts.liveBytes == HostImportLiveBytes(), "the live import counts disagree with the live bytes"); for (std::size_t index = 0; index < 6; ++index) change(index, false); + { + GuestAllocations::Mutation mutation; + mutation.Remove(block + 6 * piece); + } ClearHostImports(importing.device); ::operator delete(block, std::align_val_t{piece}); } From a1cee90f52ccfa025b6a2b152b4795f7a310fe67 Mon Sep 17 00:00:00 2001 From: Evgeniy Kazarin Date: Thu, 22 Oct 2026 08:57:01 +0800 Subject: [PATCH 5/9] test(agc): keep the map-failure memo test on a range imported at map time importMappedFailureGenerationTests (#8802) maps a single 64 KiB range and expects ImportMappedRanges to try its import. Ranges of 64 KiB are now left to their first GPU use, so the probe saw no allocation. The fixture maps 128 KiB as one window, which is still imported at map time. Co-Authored-By: Claude Opus 5.5 --- core/libs/prx/libSceAgcDriver/tests/BdaResources.cpp | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/core/libs/prx/libSceAgcDriver/tests/BdaResources.cpp b/core/libs/prx/libSceAgcDriver/tests/BdaResources.cpp index 324c9bf90e..0f28c680cd 100644 --- a/core/libs/prx/libSceAgcDriver/tests/BdaResources.cpp +++ b/core/libs/prx/libSceAgcDriver/tests/BdaResources.cpp @@ -809,7 +809,7 @@ private: void importMappedFailureGenerationTests(const Context& context) { ClearHostImports(context.device); - WindowedRange range(context, WindowedRange::piece, 1); + WindowedRange range(context, 2 * WindowedRange::piece, 1); ImportAllocationProbe probe(range.importing); probe.failure = VK_ERROR_INVALID_EXTERNAL_HANDLE; GuestAllocations::Mapped mapped; From c242694d099fa8cb1cd44eb40dceed9ce1699ae3 Mon Sep 17 00:00:00 2001 From: MarcZX10 <67475268+MarcZX10@users.noreply.github.com> Date: Fri, 23 Oct 2026 16:31:10 +0000 Subject: [PATCH 6/9] test(tools): cover Unicode environment reports --- tools/tests/test_env_reference.py | 59 +++++++++++++++++++++++++++++++ 1 file changed, 59 insertions(+) diff --git a/tools/tests/test_env_reference.py b/tools/tests/test_env_reference.py index f56a933b14..51fac32289 100644 --- a/tools/tests/test_env_reference.py +++ b/tools/tests/test_env_reference.py @@ -3,6 +3,7 @@ import io import sys import tempfile import unittest +from unittest import mock from pathlib import Path sys.path.insert(0, str(Path(__file__).resolve().parents[1])) @@ -164,6 +165,64 @@ class EnvReferenceTests(unittest.TestCase): (core / "3rdparty" / "Ignored.cpp").write_text('static bool b = IsEnvironmentSet("APS5_IGNORED");\n', encoding="utf-8") self.assertEqual(set(env_reference.reads([core], BOTH)), {"APS5_TRACE_GPU"}) + def test_main_reconfigures_streams_for_unicode_report_paths(self): + class Stream: + def __init__(self): + self.encoding = "ascii" + self.errors = "strict" + self.text = "" + + def reconfigure(self, encoding=None, errors=None): + self.encoding = encoding + self.errors = errors + + def write(self, text): + text.encode(self.encoding, self.errors) + self.text += text + return len(text) + + def flush(self): + pass + + root = self.fixture() / "source-Ж" + root.mkdir() + core = root / "core" + core.mkdir() + (core / "Sample.cpp").write_text('static bool a = std::getenv("APS5_TRACE_GPU") != nullptr;\n', encoding="utf-8") + out, err = Stream(), Stream() + with mock.patch.object(env_reference.sys, "stdout", out), mock.patch.object(env_reference.sys, "stderr", err): + self.assertEqual(env_reference.main(["--root", str(core)]), 0) + self.assertEqual((out.encoding, out.errors), ("utf-8", "backslashreplace")) + self.assertEqual((err.encoding, err.errors), ("utf-8", "backslashreplace")) + self.assertIn(str(core / "Sample.cpp"), out.text) + + def test_main_reports_unicode_scan_root_with_non_utf8_stderr(self): + class Stream: + def __init__(self): + self.encoding = "ascii" + self.errors = "strict" + self.text = "" + + def reconfigure(self, encoding=None, errors=None): + self.encoding = encoding + self.errors = errors + + def write(self, text): + text.encode(self.encoding, self.errors) + self.text += text + return len(text) + + def flush(self): + pass + + root = Path(tempfile.gettempdir()) / "missing-source-Ж" + out, err = Stream(), Stream() + with mock.patch.object(env_reference.sys, "stdout", out), mock.patch.object(env_reference.sys, "stderr", err): + self.assertEqual(env_reference.main(["--root", str(root)]), 1) + self.assertEqual((out.encoding, out.errors), ("utf-8", "backslashreplace")) + self.assertEqual((err.encoding, err.errors), ("utf-8", "backslashreplace")) + self.assertIn(str(root), err.text) + def test_main_rejects_a_scan_root_that_is_not_a_directory(self): root = self.fixture() for name in ("missing", "core/Sample.cpp"): From 918bbde6faf58fe89b8506fb6cd70869b18e6de9 Mon Sep 17 00:00:00 2001 From: MarcZX10 <67475268+MarcZX10@users.noreply.github.com> Date: Fri, 23 Oct 2026 16:31:23 +0000 Subject: [PATCH 7/9] fix(tools): reconfigure environment report streams --- tools/env_reference.py | 3 +++ 1 file changed, 3 insertions(+) diff --git a/tools/env_reference.py b/tools/env_reference.py index b7277df3bb..b4640417ac 100644 --- a/tools/env_reference.py +++ b/tools/env_reference.py @@ -160,6 +160,9 @@ def check(found, doc, prefixes, out): def main(argv=None): + for stream in (sys.stdout, sys.stderr): + if hasattr(stream, "reconfigure"): + stream.reconfigure(encoding="utf-8", errors="backslashreplace") parser = argparse.ArgumentParser( description="list the environment names the code reads, or check them against a reference document") parser.add_argument("--check", action="store_true", From da5c03c20f8d33a78c27608a2f5840cb77e3a145 Mon Sep 17 00:00:00 2001 From: MarcZX10 <67475268+MarcZX10@users.noreply.github.com> Date: Fri, 23 Oct 2026 16:42:45 +0000 Subject: [PATCH 8/9] test(tools): normalize Windows report paths --- tools/tests/test_env_reference.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/tools/tests/test_env_reference.py b/tools/tests/test_env_reference.py index 51fac32289..388160602c 100644 --- a/tools/tests/test_env_reference.py +++ b/tools/tests/test_env_reference.py @@ -194,7 +194,7 @@ class EnvReferenceTests(unittest.TestCase): self.assertEqual(env_reference.main(["--root", str(core)]), 0) self.assertEqual((out.encoding, out.errors), ("utf-8", "backslashreplace")) self.assertEqual((err.encoding, err.errors), ("utf-8", "backslashreplace")) - self.assertIn(str(core / "Sample.cpp"), out.text) + self.assertIn((core / "Sample.cpp").as_posix(), out.text) def test_main_reports_unicode_scan_root_with_non_utf8_stderr(self): class Stream: From 371652fd0e87047b399cccd12e0a2299ae3186dd Mon Sep 17 00:00:00 2001 From: milescoviello Date: Mon, 19 Oct 2026 22:25:04 +0000 Subject: [PATCH 9/9] fix(agc): draw a PS_ITER_SAMPLE program once per pixel into a one-sample color target Co-Authored-By: Claude Opus 5.5 --- core/libs/prx/libSceAgcDriver/Graphics/src/State.cpp | 3 ++- core/libs/prx/libSceAgcDriver/tests/Graphics.cpp | 6 +++++- docs/dev/debt/libSceAgcDriver.md | 2 +- 3 files changed, 8 insertions(+), 3 deletions(-) diff --git a/core/libs/prx/libSceAgcDriver/Graphics/src/State.cpp b/core/libs/prx/libSceAgcDriver/Graphics/src/State.cpp index 56436ff48f..e5fad3bdb1 100644 --- a/core/libs/prx/libSceAgcDriver/Graphics/src/State.cpp +++ b/core/libs/prx/libSceAgcDriver/Graphics/src/State.cpp @@ -1492,9 +1492,10 @@ State DecodeState(const QueueState& queue, bool vertexWorkgroup) { color.uintExport = slotExport == 7; if (color.samples == 1 && result.samples > 1 && result.colors.empty() && !result.depth && (read(cx, 0x292) & 1u) == 0) { const auto sampleInputs = read(cx, 0x1b3) & read(cx, 0x1b4) & (ShaderRecompiler::PixelInputBit(ShaderRecompiler::PixelInput::Ancillary) | ShaderRecompiler::PixelInputBit(ShaderRecompiler::PixelInput::SampleCoverage)); - Require(!result.perSampleShading && sampleInputs == 0 && (read(cx, 0x203) & 0x100u) == 0, "a one-sample color target under a multisampled rasterizer without MSAA_ENABLE is unsupported with PS_ITER_SAMPLE, a sample index or coverage input, or a sample-mask export"); + Require(sampleInputs == 0 && (read(cx, 0x203) & 0x100u) == 0, "a one-sample color target under a multisampled rasterizer without MSAA_ENABLE is unsupported with a sample index or coverage input, or a sample-mask export"); result.samples = 1; result.customSampleLocations = false; + result.perSampleShading = false; } if (color.samples > 1 && result.samples == 1 && result.colors.empty() && !result.depth && !result.alphaToCoverage && (read(cx, 0x292) & 1u) == 0) { Require(result.conservativeRasterization == VK_CONSERVATIVE_RASTERIZATION_MODE_DISABLED_EXT || (read(cx, 0x1b3) & 0x44u) == 0, "conservative rasterization with centroid interpolation of a draw raised to a multisampled target is unsupported"); diff --git a/core/libs/prx/libSceAgcDriver/tests/Graphics.cpp b/core/libs/prx/libSceAgcDriver/tests/Graphics.cpp index abdb669923..00390f67fb 100644 --- a/core/libs/prx/libSceAgcDriver/tests/Graphics.cpp +++ b/core/libs/prx/libSceAgcDriver/tests/Graphics.cpp @@ -824,7 +824,7 @@ void msaaEnableClearOneSampleTargetTests() { queue = drawC(); queue.context[0x1b3] |= ShaderRecompiler::PixelInputBit(ShaderRecompiler::PixelInput::Ancillary); requireOneSample(queue, "draw C with an enabled but unaddressed ANCILLARY input"); - constexpr std::string_view refused = "a one-sample color target under a multisampled rasterizer without MSAA_ENABLE is unsupported with PS_ITER_SAMPLE, a sample index or coverage input, or a sample-mask export"; + constexpr std::string_view refused = "a one-sample color target under a multisampled rasterizer without MSAA_ENABLE is unsupported with a sample index or coverage input, or a sample-mask export"; for (const auto input : {ShaderRecompiler::PixelInput::Ancillary, ShaderRecompiler::PixelInput::SampleCoverage}) { queue = drawC(); queue.context[0x1b3] = queue.context[0x1b4] = 2u | ShaderRecompiler::PixelInputBit(input); @@ -832,6 +832,10 @@ void msaaEnableClearOneSampleTargetTests() { } queue = drawC(); queue.context[0x293] = 0x10000u; + queue.context[0x1b3] = queue.context[0x1b4] = 2u | ShaderRecompiler::PixelInputBit(ShaderRecompiler::PixelInput::PerspectiveSample); + requireOneSample(queue, "draw C with PS_ITER_SAMPLE and a sample-interpolated input"); + Require(!DecodeState(queue).perSampleShading, "PS_ITER_SAMPLE into a one-sample color target without MSAA_ENABLE asked for sample shading"); + queue.context[0x1b3] = queue.context[0x1b4] = 2u | ShaderRecompiler::PixelInputBit(ShaderRecompiler::PixelInput::Ancillary); expectFailure([&] { DecodeState(queue); }, refused); queue = drawC(); queue.context[0x1c4] = 9; diff --git a/docs/dev/debt/libSceAgcDriver.md b/docs/dev/debt/libSceAgcDriver.md index 7a8f6ad52d..9e87ef8f93 100644 --- a/docs/dev/debt/libSceAgcDriver.md +++ b/docs/dev/debt/libSceAgcDriver.md @@ -37,7 +37,7 @@ - [Point geometry output](../../../core/libs/prx/libSceAgcDriver/Graphics/src/State.cpp) (AGC driver) - a geometry program with `VGT_GS_OUT_PRIM_TYPE = 0` emits mesh points with `OutputPoints` and scalar `PrimitivePointIndicesEXT` as required by [SPV_EXT_mesh_shader](https://github.khronos.org/SPIRV-Registry/extensions/EXT/SPV_EXT_mesh_shader.html). Each packed primitive export contributes its first 9-bit vertex index and its null bit, using the same NGG primitive packing as line and triangle outputs. The subgroup reserves at least `GE_MAX_OUTPUT_PER_SUBGROUP` point slots, as the line path does; the console budget is not measured. Points use the register size or the misc-X point-size export and clamp interpreted by the vertex-point path; those register, clamp and rasterization inferences remain unmeasured on the console. Point sprites remain unsupported. Point output from a vertex workgroup still throws. Cocoon (PPSA08766) reaches this path with triangle-strip input. - [Line-strip geometry output](../../../core/libs/prx/libSceAgcDriver/Graphics/src/State.cpp) (AGC driver) - a geometry program with `VGT_GS_OUT_PRIM_TYPE = 1` (line strip) runs as a mesh shader with `OutputLinesEXT`: each primitive export is one line whose two vertex indices are packed like the three of a triangle, as Mesa's NGG lowering exports line strips (`ac_nir_lower_ngg_gs.c`), and each input primitive gets `VGT_GS_MAX_VERT_OUT - 1` lines where a triangle strip gets `- 2` triangles. A subgroup gets at least `GE_MAX_OUTPUT_PER_SUBGROUP` line slots; whether the console does the same is unmeasured. A fast-launch subgroup takes its line count from `GE_NGG_SUBGRP_CNTL.PRIM_AMP_FACTOR`. Line output counts as lines for polygon offset decode. The lines are drawn 1 pixel wide with Vulkan default rasterization; `PA_SU_LINE_CNTL` is not read, and console line rules are not measured. - [Alpha-to-coverage](../../../core/libs/prx/libSceAgcDriver/Graphics/src/State.cpp) (AGC driver) - only `DB_ALPHA_TO_MASK = 0xaa01` is accepted: the enable bit with the four quad offsets at 2 and no `OFFSET_ROUND`, the undithered setting [radeonsi](https://gitlab.freedesktop.org/mesa/mesa/-/blob/bffd91947efc1932da99aca69fc540d6d4201b0c/src/gallium/drivers/radeonsi/si_state.c#L371-L381) programs. Draws are single-sampled, so the pixel shader [writes](../../../core/shader/recompiler/SpirvBackend/src/SpirvModuleEmitter.cpp) `SampleMask` 0 when its MRT0 alpha is below one half or NaN and all ones otherwise, and runs with late fragment tests. That rule was measured on a Radeon 660M (gfx1035, RADV with `RADV_DEBUG=noatocdithering`, which programs the same register) for 0, -0, 0.25, the two floats below one half, one half, the two above, 0.75, 1, 2, -1, both infinities and both NaNs, and is pinned in [AlphaToCoverage.cpp](../../../core/libs/prx/libSceAgcDriver/tests/execution/AlphaToCoverage.cpp); it is not measured on gfx1013. Vulkan's `alphaToCoverageEnable` is not used: the NVIDIA driver 615.78.08 dithers single-sample coverage (an alpha of 0.25 writes half the pixels of an RTX 3050 Laptop target). Other offsets, `OFFSET_ROUND`, a sample-mask export, and an MRT0 that is missing, swizzled, packed or exported without a floating-point alpha throw. A draw without a color target on MRT0 keeps alpha-to-coverage from its MRT0 export, as [radeonsi exports MRT0 as `32_AR` for alpha-to-coverage without a color buffer](https://gitlab.freedesktop.org/mesa/mesa/-/blob/mesa-24.0.0/src/gallium/drivers/radeonsi/si_state_shaders.cpp#L2425-L2431). A draw that enables alpha-to-coverage with no MRT0 export still throws: what the console covers then is not measured. Default dithered `DB_ALPHA_TO_MASK = 0x18701` additionally uses a per-quad sample-mask table [measured on gfx1036](https://github.com/boykopovar/AnyPS5/pull/4451#issuecomment-6124848921) for eight standard samples with FP16_ABGR or 32_ABGR exports; other sample counts, custom locations, EQAA combinations and dithered UNORM16/SNORM16 exports throw. The dithered rule is not measured on gfx1013. -- [A one-sample target under a multisampled rasterizer without MSAA_ENABLE](../../../core/libs/prx/libSceAgcDriver/Graphics/src/State.cpp) (AGC driver) - `PA_SC_AA_CONFIG.MSAA_NUM_SAMPLES` > 0 with `PA_SC_MODE_CNTL_0.MSAA_ENABLE` clear, one-sample color targets and no depth target, or a one-sample depth target with no color target or one-sample color targets, is drawn as a one-sample draw at the pixel center. Without the bit every sample is taken to be at the pixel center (see the entry for a multisampled target rasterized without MSAA_ENABLE below), and a one-sample target is taken to keep the coverage of sample 0, so a one-sample depth target is tested and written once per pixel with the depth at the pixel center. That is the rule [`VK_AMD_mixed_attachment_samples`](https://github.com/KhronosGroup/Vulkan-Docs/blob/193ebe195f47ffc5004683d1659e71c725658c4d/chapters/fragops.adoc#L2510-L2516) gives a color target with fewer samples than the rasterizer, but Vulkan reaches it only beside a depth or stencil target with the rasterizer's sample count ([VUID-VkGraphicsPipelineCreateInfo-subpass-01505](https://github.com/KhronosGroup/Vulkan-Docs/blob/193ebe195f47ffc5004683d1659e71c725658c4d/chapters/pipelines.adoc#L3033-L3041)), so here it is an analogy, and for a one-sample depth target an extension of it. [shadPS4](https://github.com/shadps4-emu/shadPS4/blob/15abf68c9b52c4b04d094f2837719fdb871021f3/src/video_core/renderer_vulkan/vk_graphics_pipeline.cpp#L127-L128) also rasterizes with the largest sample count of the bound depth and color targets ([L416-L417](https://github.com/shadps4-emu/shadPS4/blob/15abf68c9b52c4b04d094f2837719fdb871021f3/src/video_core/renderer_vulkan/vk_pipeline_cache.cpp#L416-L417), [L477-L480](https://github.com/shadps4-emu/shadPS4/blob/15abf68c9b52c4b04d094f2837719fdb871021f3/src/video_core/renderer_vulkan/vk_pipeline_cache.cpp#L477-L480)), and reads `PA_SC_AA_CONFIG` only for a `SAMPLE_COVERAGE` input ([L174-L177](https://github.com/shadps4-emu/shadPS4/blob/15abf68c9b52c4b04d094f2837719fdb871021f3/src/video_core/renderer_vulkan/vk_pipeline_cache.cpp#L174-L177)). This is not measured on hardware. An `ANCILLARY` or `SAMPLE_COVERAGE` pixel input and a sample-mask export throw in a pixel program that runs, and so does `PS_ITER_SAMPLE` with one-sample color targets and no depth target, because what the console gives them under a multisampled configuration is not known; a depth-only draw that skips its pixel program is drawn with the null program, and its `PS_ITER_SAMPLE` and pixel input bits are ignored. With a one-sample depth target, a pixel program with `PS_ITER_SAMPLE` and neither input runs once per pixel; if the console runs it once per sample, its atomics and append or consume counters count differently. Points and alpha-to-coverage throw as for a multisampled target. With a one-sample depth target, `DB_EQAA.INTERPOLATE_COMP_Z` is accepted: with every sample at the pixel center, Z interpolated to any sample position is the center's Z, which the one-sample draw tests. That rests on the same inference; neither [PAL](https://github.com/GPUOpen-Drivers/pal/blob/c5e800072a32f68b6ccc4422936d96167c6e0728/src/core/hw/gfxip/gfx9/gfx9MsaaState.cpp#L117-L118) nor [radeonsi](https://gitlab.freedesktop.org/mesa/mesa/-/blob/761d5c8a54f93b3d43608f718910ed742983dd0b/src/gallium/drivers/radeonsi/si_state.c#L3314-L3316) sets the bit, and its meaning is not documented. Line and point lists (`VGT_PRIMITIVE_TYPE` 1-3, even when a geometry program turns them into triangles), `INTERPOLATE_SRC_Z` and overrasterization (`OVERRASTERIZATION_AMOUNT`, `ENABLE_POSTZ_OVERRASTERIZATION`) still throw. METAL GEAR SOLID: Peace Walker (PPSA18216) draws a one-sample color target under a 4x configuration this way; the first draw of its opening movie binds a one-sample depth target and no color target under the same configuration. Gran Turismo 7 (PPSA01317) draws its main pass with a one-sample depth target, `DB_EQAA` 0x00050000 and a 4x configuration. +- [A one-sample target under a multisampled rasterizer without MSAA_ENABLE](../../../core/libs/prx/libSceAgcDriver/Graphics/src/State.cpp) (AGC driver) - `PA_SC_AA_CONFIG.MSAA_NUM_SAMPLES` > 0 with `PA_SC_MODE_CNTL_0.MSAA_ENABLE` clear, one-sample color targets and no depth target, or a one-sample depth target with no color target or one-sample color targets, is drawn as a one-sample draw at the pixel center. Without the bit every sample is taken to be at the pixel center (see the entry for a multisampled target rasterized without MSAA_ENABLE below), and a one-sample target is taken to keep the coverage of sample 0, so a one-sample depth target is tested and written once per pixel with the depth at the pixel center. That is the rule [`VK_AMD_mixed_attachment_samples`](https://github.com/KhronosGroup/Vulkan-Docs/blob/193ebe195f47ffc5004683d1659e71c725658c4d/chapters/fragops.adoc#L2510-L2516) gives a color target with fewer samples than the rasterizer, but Vulkan reaches it only beside a depth or stencil target with the rasterizer's sample count ([VUID-VkGraphicsPipelineCreateInfo-subpass-01505](https://github.com/KhronosGroup/Vulkan-Docs/blob/193ebe195f47ffc5004683d1659e71c725658c4d/chapters/pipelines.adoc#L3033-L3041)), so here it is an analogy, and for a one-sample depth target an extension of it. [shadPS4](https://github.com/shadps4-emu/shadPS4/blob/15abf68c9b52c4b04d094f2837719fdb871021f3/src/video_core/renderer_vulkan/vk_graphics_pipeline.cpp#L127-L128) also rasterizes with the largest sample count of the bound depth and color targets ([L416-L417](https://github.com/shadps4-emu/shadPS4/blob/15abf68c9b52c4b04d094f2837719fdb871021f3/src/video_core/renderer_vulkan/vk_pipeline_cache.cpp#L416-L417), [L477-L480](https://github.com/shadps4-emu/shadPS4/blob/15abf68c9b52c4b04d094f2837719fdb871021f3/src/video_core/renderer_vulkan/vk_pipeline_cache.cpp#L477-L480)), and reads `PA_SC_AA_CONFIG` only for a `SAMPLE_COVERAGE` input ([L174-L177](https://github.com/shadps4-emu/shadPS4/blob/15abf68c9b52c4b04d094f2837719fdb871021f3/src/video_core/renderer_vulkan/vk_pipeline_cache.cpp#L174-L177)). This is not measured on hardware. An `ANCILLARY` or `SAMPLE_COVERAGE` pixel input and a sample-mask export throw in a pixel program that runs, because what the console gives them under a multisampled configuration is not known; a depth-only draw that skips its pixel program is drawn with the null program, and its `PS_ITER_SAMPLE` and pixel input bits are ignored. A pixel program with `PS_ITER_SAMPLE` and neither input runs once per pixel; if the console runs it once per sample, its atomics and append or consume counters count differently. Points and alpha-to-coverage throw as for a multisampled target. With a one-sample depth target, `DB_EQAA.INTERPOLATE_COMP_Z` is accepted: with every sample at the pixel center, Z interpolated to any sample position is the center's Z, which the one-sample draw tests. That rests on the same inference; neither [PAL](https://github.com/GPUOpen-Drivers/pal/blob/c5e800072a32f68b6ccc4422936d96167c6e0728/src/core/hw/gfxip/gfx9/gfx9MsaaState.cpp#L117-L118) nor [radeonsi](https://gitlab.freedesktop.org/mesa/mesa/-/blob/761d5c8a54f93b3d43608f718910ed742983dd0b/src/gallium/drivers/radeonsi/si_state.c#L3314-L3316) sets the bit, and its meaning is not documented. Line and point lists (`VGT_PRIMITIVE_TYPE` 1-3, even when a geometry program turns them into triangles), `INTERPOLATE_SRC_Z` and overrasterization (`OVERRASTERIZATION_AMOUNT`, `ENABLE_POSTZ_OVERRASTERIZATION`) still throw. METAL GEAR SOLID: Peace Walker (PPSA18216) draws a one-sample color target under a 4x configuration this way; the first draw of its opening movie binds a one-sample depth target and no color target under the same configuration. Gran Turismo 7 (PPSA01317) draws its main pass with a one-sample depth target, `DB_EQAA` 0x00050000 and a 4x configuration. - [MSAA_ENABLE on a single-sample target](../../../core/libs/prx/libSceAgcDriver/Graphics/src/State.cpp) (AGC driver) - `PA_SC_MODE_CNTL_0.MSAA_ENABLE` with `PA_SC_AA_CONFIG.MSAA_NUM_SAMPLES = 0` is drawn like the same state without the bit when sample 0 of the four `PA_SC_AA_SAMPLE_LOCS_PIXEL_*_0` registers is at the pixel center and the draw rasterizes filled triangles or rectangles (`VGT_PRIMITIVE_TYPE` 7 or 17 as well as the triangle types) with alpha-to-coverage and conservative rasterization off. [RADV](https://gitlab.freedesktop.org/mesa/mesa/-/blob/745709a000482a4a24c91faef5c5a40bd653a863/src/amd/vulkan/radv_cmd_buffer.c#L13606-13609) programs this state for one-sample custom sample locations on GFX10 and later, where the bit makes the rasterizer use the programmed location of sample 0 ([061bc615](https://gitlab.freedesktop.org/mesa/mesa/-/commit/061bc6151a11597f720535f00e79a5752e24aff6)); that the bit changes nothing else for such a draw is inferred from that and is not measured on hardware. EA Sports UFC 5 (PPSA03541) reaches a rect-list draw in this state. Other locations of sample 0, absent location registers, points, lines, alpha-to-coverage and conservative rasterization throw - [A multisampled target rasterized without MSAA_ENABLE](../../../core/libs/prx/libSceAgcDriver/Graphics/src/State.cpp) (AGC driver) - not measured on hardware: `PA_SC_AA_CONFIG.MSAA_NUM_SAMPLES` > 0 with `PA_SC_MODE_CNTL_0.MSAA_ENABLE` clear puts every color, depth and stencil sample at the pixel center, as in GL without multisampling ([ARB_multisample](https://github.com/KhronosGroup/OpenGL-Registry/blob/6af574a14089/extensions/ARB/ARB_multisample.txt#L185), [RADV 061bc615](https://gitlab.freedesktop.org/mesa/mesa/-/commit/061bc6151a11)). Lines cover whole pixels by the host's line rule, not the console's aliased one ([D3D11.3](https://github.com/microsoft/DirectX-Specs/blob/f37b0d6033a1/d3d/archive/D3D11_3_FunctionalSpec.htm#L13705)). Depth sample tests and occlusion counts are unknown. Rasterized points, alpha-to-coverage, and Z interpolation, overrasterization or a sample-mask export with depth throw. PPSA06888 draws a 2x target, PPSA18216 a 4x depth and color pair and 4x lines. - [sceAgcDriverGetEqContextId](../../../core/libs/prx/libSceAgcDriver/Eq/src/Query.cpp) (libSceAgcDriver) - returns `ident`: an end-of-pipe interrupt is delivered only to the registration whose id equals the raising queue, so `ident` is that context's id (titles register 0 or 1 for graphics and `0x20 + 8 * pipe + queue` for compute). The console's value is not known; KytyPS5 returns the event's `data`, which it fills with the interrupt's context id, whereas `data` here is the EV_CLEAR trigger count. Other event filters throw