HH45137
e3fac5e9fa
[Fix]: Add Task/Mesh shader enum mapping
...
Map the Task and Mesh shader stages to their NV OpenGL enum values in ProgramEnumConverter so shader stage conversion handles task and mesh pipelines correctly.
2026-10-12 00:41:40 +08:00
HH45137
3823fe528c
[Feat, Test]: Add NV mesh shader support
...
Advertise GL_NV_mesh_shader in the DirectVulkan backend and expose the new task/mesh shader stages in the GL-to-MG and glslang conversion layers. This also wires the mesh task drawing entry points and adds a focused MeshShader test covering extension discovery and task shader compilation.
2026-10-12 00:20:53 +08:00
swung0x48
08124c99f1
[Test] (trace): replace the create-indirect fixture with minecraft-1.21.1-neoforge-create-indirect-in-world-align1024, a derived trace and not the original capture - the desktop capture bound Flywheel's scatter lists at 16-byte SSBO offsets that Adreno 830 and Mali's Vulkan driver (64) reject, and fixture_tools/realign_buffer_ranges.cpp moves the 1,270 binds off a 1024-byte boundary onto 1024-aligned slots (3,812 calls inserted, target call 504631 -> 508443, host SSIM unchanged to 9 digits on both backends)
2026-09-30 10:00:26 -04:00
swung0x48
f973008c4d
[Fix] (Espryt): a partial buffer flush no longer takes the range-invalidating map while an upload-ring copy into the same store may still be queued - Adreno 830 loses the queued copy, and create-instancing drew whole Embeddium terrain sections empty
2026-09-30 04:07:41 -04:00
swung0x48
7030d307be
[Test] (Magma): UnboundUniformBlockScenario - a dispatch and a draw whose program declares a uniform block at an empty binding point (never bound, or its buffer deleted while bound) run and produce their fallback, and a buffer bound there afterwards is read, on both backends
2026-09-29 19:58:41 -04:00
swung0x48
853c5f1264
[Fix] (Magma): a uniform block declared at an empty binding point reads a zeroed block instead of dereferencing a null buffer
2026-09-29 19:55:53 -04:00
swung0x48
c14283ee2d
[Docs] (Magma): the trim threshold's comment states the bound its argument proves - a growth adds at most twice what the slot held, so the slot stays over a third full
2026-09-29 10:58:33 -04:00
swung0x48
9dccabf03c
[Test] (Magma): descriptor pool growth and trim - the pure sizing policy adds O(log N) pools for an N-set frame, and DescriptorPoolGrowthScenario's 4096-distinct-set frame lands whole on both backends, grows at most 7 pools, and is given back after quiet frames and re-grown while they are still in flight
2026-09-29 10:34:16 -04:00
swung0x48
53fa16c2c2
[Fix] (Magma): a frame slot whose descriptor pools stay under a quarter full for 120 frames is handed back whole - its set cache cleared, grown pools destroyed, base pool reset and every memo holding a set handle dropped - so rd12's loading burst no longer lives for the whole process (live sets 15,042 -> 39, pools 65 -> 6); MOBILEGL_MAGMA_DESCRIPTOR_TRIM_FRAMES (0 = never) and a test-visible pool census
2026-09-29 10:34:16 -04:00
swung0x48
c2de6576ad
[Fix] (Magma): a frame slot's next descriptor pool doubles its LARGEST pool of the same update-after-bind flavour, capped at 4096 sets - doubling the first (oldest, smallest) one made every grown pool the same size, so rd12's 15,016-set loading frame grew one slot to 60 pools (1x128 + 59x256); now 9 (64 ... 4096, 4096, 4096)
2026-09-29 10:34:15 -04:00
swung0x48
9c32d9fa68
[Test] (Magma): OversizedRecordingSplitScenario carries state across the split - a depth occluder, a stencil mark, an additive blend destination, a 4x multisample target resolved after it, occlusion and primitives-generated queries and a transform-feedback capture left open across it, each checked on both backends and in the budget-64 lane
2026-09-29 09:44:58 -04:00
swung0x48
4d3d677a7f
[Test] (Magma): OversizedRecordingSplitScenario - a 1024-draw frame, one pixel and one uniform value per draw, lands whole on both backends; the DirectVulkan.SplitRecording lane pins the draw budget to 64 and asserts the split fired
2026-09-29 09:44:58 -04:00
swung0x48
d20c2d01ae
[Fix] (Magma): a frame's recording is submitted every 16384 draws/dispatches and continues on a fresh command buffer, with at most frames-in-flight + 1 submissions outstanding - rd12's 981k-draw loading frame no longer grows one command buffer until Adreno's 16 KiB driver mappings exhaust vm.max_map_count (MOBILEGL_MAGMA_MAX_DRAWS_PER_COMMAND_BUFFER, 0 = unbounded)
2026-09-29 09:44:58 -04:00
swung0x48
21c7e8cd63
[Test] (MG_IntegrationTest, MG_Test): scattered writes over a GPU clear on both backends - two distant 2x2 boxes, a hundred boxes, boxes filling 3/4 of their union and three quadrants spanning the level, the 3/4 and level-spanning cases again at the image-bindable re-mint, and a MipmapStorage footprint fuzz against a reference bitmap
2026-09-29 09:06:07 -04:00
swung0x48
9524c688ca
[Fix] (Espryt, Magma): scattered glTexSubImage writes upload exactly their texels - MipmapStorage keeps an exact dirty footprint beside its rect list (a write log once a merge or a fold would widen the list past the writes), Magma always uploads it, Espryt uploads it box by box (through the unpack ring too) once the driver texture may hold GPU writes and merges it at the image-bindable re-mint; the union box, the 3/4 cut-off, the 96-rect fold and a level-spanning union no longer put the stale shadow over GPU-written texels (two 2x2 writes over a GPU clear: 41 of 64 texels red -> 0)
2026-09-29 09:06:07 -04:00
swung0x48
c60639db1d
[Fix] (Magma): eglSwapInterval reaches the swapchain - 0 picks IMMEDIATE, else MAILBOX, else FIFO; >= 1 FIFO; < 0 FIFO_RELAXED, else FIFO; rebuilt after the next present when the mode changes (MC 26.3 vsync on: 190-290 fps -> 120)
2026-09-29 05:51:18 -04:00
swung0x48
b1d60523e4
[Fix] (Magma): an upload into an image the open recording already wrote, or that still has a queued clear, is submitted after that work - a glTexSubImage after a clear or a draw into the level no longer ends up under it
2026-09-29 04:28:55 -04:00
swung0x48
8966d5c294
[Fix] (Espryt, Magma): a partial write into a level whose upload is converted (widened RGB formats, Magma's RGB->RGBA expansion) or reshaped (Espryt 1D arrays) uploads only its dirty region - the rest of the stale shadow no longer lands over GPU-drawn texels
2026-09-29 04:28:22 -04:00
swung0x48
568090f09f
[Fix] (Espryt): the image-bindable re-mint reads every level back from the driver texture before replaying it, so GPU-written levels survive the first glBindImageTexture - iris-derivative SSIM 0.996196552 -> 0.996376198
2026-09-29 03:06:45 -04:00
swung0x48
fff9d639f0
[Fix] (DirectVulkan, tools/trace_replay): preserve resident buffer ordering and coherent trace copies - fix intermittent geometry and UI corruption in Minecraft 26.3-rc-3
...
Stage updates to adopted GPU-resident buffers on the command timeline instead of overwriting memory that earlier draws still read. Wait for submitted copies as well as pending commands before CPU readback.
Commit coherent mapped writes before Android trace buffer copies, and retire buffer shadows on deletion so copies see current vertices and name reuse cannot leave dangling dirty-shadow entries.
Validated draw ordering, submitted-copy readback, traced buffer copies, and mapped-buffer deletion/name reuse on Redmi Adreno 830. Minecraft 26.3-rc-3 Magma passed movement, HUD, and menu checks and exited normally; the new trace parsed all 64,410 frames without warnings.
2026-09-16 01:22:11 -04:00
swung0x48
9eae98581f
[Test] (SelfTest, POST): probe upload/draw ordering in persistently mapped vertex arenas - native GLES controls distinguish mapped destination corruption from staging and synchronization failures
...
Add the persistent buffer ordering probe to the GLES POST's known-driver-bug inventory. Queue updates and draws into independent FBOs before reading them back, covering SubData and copies from both coherent persistent and ordinary staging buffers. Retry fresh mapped allocations for intermittent corruption.
Require passing never-mapped and fully serialized controls before reporting a finding. Include finish-before, map-then-unmap and barrier diagnostics, preserve caller GL state, and treat setup, allocation and GL errors as inconclusive. This adds detection and reporting only; no rendering workaround is enabled.
Validation: all 46 DriverBugProbesTest tests pass, including 11 new ordering, control, collector and cleanup cases. The Android API 26 / NDK 27 native probe detects all three upload paths on Mali-G1-Ultra r54p1 with clean controls and no GL errors. llvmpipe reports no finding after 240 mapped FBO checks per upload path.
2026-09-07 00:44:08 -04:00
swung0x48
d7655247f7
[Fix, Test] (DirectGLES, Integration): rebind VAOs when an adopted buffer is respecified - the immediate retire path forgot the buffer-id generation, so cached vertex and element bindings kept the deleted store
...
Advance the buffer-id generation when Ops_Respecify retires immutable storage on the context thread, matching the existing adoption and deferred-retirement paths.
Add pixel regression coverage for unchanged VBO/IBO attachments across same-size redefinition, growth and shrinkage, including bound and unbound VAOs sharing a vertex arena and an index arena returning to shadow storage.
2026-09-06 22:55:06 -04:00
swung0x48
50fb13430f
[Merge] (MG_State, ShaderTranspiler, tools/cts): land the write-map landing and open-ended fp64 storage block fixes that take KHR-Single-GL45.subgroups to 100% on Magma
2026-09-05 05:43:34 -04:00
swung0x48
d4f8adcf6d
[Tools, Test] (tools/cts, MG_Test): make the CTS runner's chunk timeout idle-based so a healthy 20-minute invocation is no longer killed and its in-flight case mis-recorded as a crash, and link MSVC test executables with /WHOLEARCHIVE so the dllimport gl* references in GetProcAddress.cpp resolve
2026-09-05 05:36:50 -04:00
swung0x48
795e08f7e6
[Fix, Test] (ShaderTranspiler): flatten fp64 storage blocks whose last member is a runtime array - the pass declined them, so the fp64 demotion re-derived ArrayStride 4 over the application's 8/16/32-byte double buffer and every double/dvecN data[] SSBO read raw words
2026-09-05 05:36:49 -04:00
swung0x48
1e7ecab4db
[Fix, Test] (MG_State, BufferObject): land a non-persistent write map's staged bytes into a GPU-resident store at unmap and explicit flush instead of dropping them - SSBO binding and large-store adoption make resident stores reachable through glMapBufferRange, so every per-draw re-initialisation was silently lost
2026-09-05 05:36:48 -04:00
swung0x48
81b17c0b75
[Test] (ShaderTranspiler, Integration): pin the five reworked shapes - the read-only capture stage through the real link, the seeded carrier's ESSL declaration, the metadata-driven block redeclaration and its decline, the control-stage-less evaluation decline, and a carrier clearing an i64vec4
2026-08-28 06:22:12 -04:00
swung0x48
d1edf765f5
[Fix] (Link, Async): carry the resolved gl_PointSize capture request into the SPIR-V handoff and let a deferred verdict name its own severity - the demotion read the request off a reflection slice phase A never fills it into, so its forced carrier was dead code, and its decline reason replayed at a level no shipped build keeps
2026-08-28 06:22:12 -04:00
swung0x48
97e07190ac
[Fix] (ShaderTranspiler): seed a forced point-size carrier nothing writes and decline a control stage whose live clip/cull distance would still print gl_PointSize - an ES front end deletes a never-written output, and SPIRV-Cross redeclares that block from member decorations rather than access
2026-08-28 06:22:11 -04:00
swung0x48
a4dcdf989e
[Fix] (ShaderTranspiler): stop the point-size carrier landing on a location a live varying owns, and decline the evaluation stage a synthesized pass-through control stage cannot feed - an i64vec4 counted as one location, and a located input carrier trips both backends' pass-through guard
2026-08-28 06:22:11 -04:00
swung0x48
1c113e4b26
[Fix] (ShaderTranspiler): count a 64-bit integer vector as two locations in the XFB flattener's LocationSpan - it read 64-bitness off the element's float type alone, so an i64vec3/4 member packed the members after it onto locations it already owns
2026-08-28 06:22:11 -04:00
swung0x48
bf9cfb3079
[Test] (Integration): PointSizeDemotionScenario - the demoted value chain is client-invisible in both configurations, with pinned per-backend lanes and a log-latch arming guard
2026-08-28 06:22:10 -04:00
swung0x48
e1818d497a
[Test] (ShaderTranspiler): pin the point-size demotion's shapes - capability stripped, carriers named and located past the program's varyings, byte-identical no-ops, whole-struct-copy decline, and the two new L1 key bits
2026-08-28 06:21:31 -04:00
swung0x48
d7f66722d1
[Fix] (ShaderTranspiler, Link, DirectGLES, DirectVulkan): demote tessellation/geometry gl_PointSize to an ordinary varying where the device cannot host the built-in - the value survives for gl_in reads and by-name capture, both backends' declines stay for shapes the pass refuses, and the verdict rides the L1 key
2026-08-28 06:21:30 -04:00
swung0x48
92dc41ebf9
[Test] (DirectVulkan, SelfTest): pin the probe's fence-timeout teardown against a fake driver, the never-worse arming refusal, and paused-span patch/instanced counting against an unpaused control
2026-08-28 06:19:41 -04:00
swung0x48
feea131d8b
[Fix] (DirectVulkan, SelfTest): honour the primitives-generated probe's leak-on-timeout contract in both of its callers, count paused-span draws the frontend cannot price, refuse a substitute measured worse than the stream query, and derive the POST row's failure clause from the measurement
2026-08-28 06:19:40 -04:00
swung0x48
19f4402fbf
[Test] (DirectVulkan): pin the primitives-generated probe's verdict and override mapping, and hold the reroute's GL answers and arming observable on the integration lane
2026-08-28 06:19:40 -04:00
swung0x48
1350031368
[Fix] (DirectVulkan): count GL_PRIMITIVES_GENERATED for transform-feedback-inactive draws - a bring-up probe measures the silent stream query and reroutes such draws through the dedicated primitives-generated query or a clipping-statistics pool, reported from the Vulkan POST
2026-08-28 06:19:39 -04:00
swung0x48
0ee3384b22
[Fix] (Espryt): land a SubData into an adopted store as a GPU-ordered copy - the in-place coherent write tore the frames still reading the old section bytes
2026-08-28 04:50:07 -04:00
swung0x48
ba3f8d6774
[Test] (Integration): pin the adopted mesh-arena store - cross-frame SubData visibility, readback identity, and the GPU-written readback
2026-08-28 04:19:11 -04:00
swung0x48
3327784fd0
[Fix] (Espryt): adopt mesh-arena-sized stores into coherent persistent maps at definition, and stop the flush tiers from re-synchronizing them
2026-08-28 04:19:11 -04:00
swung0x48
ff426da3a9
[Fix, Test] (MG_State, DirectVulkan): order host writes to an adopted store after recorded GPU work - a SubData issued after a dispatch landed in coherent memory before the deferred dispatch executed, so its increments overwrote the newer bytes; un-skip the DirectVulkan half of the SubData-after-dispatch scenario
2026-08-28 03:01:15 -04:00
swung0x48
08419a1fe6
[Chore] (Config): point the two stale comments at the renamed feature fields
2026-08-28 00:47:35 -04:00
swung0x48
5d51372c44
[Chore] (Config): triage backend-scoped env toggles under MOBILEGL_ESPRYT_ and MOBILEGL_MAGMA_ prefixes
2026-08-28 00:38:17 -04:00
swung0x48
7fd4550968
[Fix] (Espryt): copy before any FBO attach in the packed16 probe - attaching the array relayouts it to the plain order and was neutralizing the subject
2026-08-28 00:22:16 -04:00
swung0x48
971537058e
[Fix] (Espryt): probe every allocation recipe of the packed16 shape - the Mali layout heuristic inverts between contexts, so one recipe cannot speak for the storage
2026-08-28 00:22:16 -04:00
swung0x48
dd98c450ad
[Test] (SelfTest): model the params-before-upload escape so a params-first probe regression reads as a red test
2026-08-28 00:22:15 -04:00
swung0x48
5dbbbbd7eb
[Fix] (Espryt): allocate the packed16 probe's textures uploads-first, the order a minted backend texture performs and the one the Mali layout choice keys on
2026-08-28 00:22:15 -04:00
swung0x48
5d140a41ce
[Fix] (Espryt): re-arm the packed16 probe on the allocation-scoped mirror the device actually has and let any mirrored level trigger the widening
2026-08-28 00:22:15 -04:00
swung0x48
bf376b230f
[CI] (Integration): rerun the buffer scenarios with the map flush disabled so the upload-ring tier keeps coverage
2026-08-28 00:15:15 -04:00
HH45137
swung0x48