Program 203 reuses prefixSumCache for a second subgroup reduction before every workgroup invocation has consumed the first result. Add a fingerprint-gated SPIR-V pass that inserts the missing Workgroup acquire-release barrier while preserving native subgroup operations.
Keep the repair opt-in behind MOBILEGL_ITERATIONRP_FIX_BARRIER, cover insertion, pass-through, and idempotence, and enable it together with the existing iterationRP subgroup repairs for the matching Linux and Android CI retraces.
Six more verified defects from the residual memo/cache mechanisms:
- Program resource cache (DirectVulkan reflection): glShaderStorageBlockBinding
deliberately does not bump the backend state version, and the SSO pipeline
composite is unnamed so the by-name in-place patch can never reach its slot -
the composite kept serving pre-rebind SSBO bindings. The cache now keys on the
program's block-binding version; a binding-only change re-applies the overrides
by name instead of re-running spirv-reflect. SetShaderStorageBlockBinding also
gains the equality bail-out its uniform-block sibling has, so the composite
mirror's replay stops churning the version every draw.
- LinkProgram's allowVSOnlyPrograms function-static latch never set its own
initialized flag (dead memo, re-read every call) - and completing it would have
frozen a per-backend capability across re-initialization. Replaced with a fresh
per-link read from the null-checked active backend.
- Query object registry: drained at full library teardown (DestroyAllQueryObjects,
mirroring DestroyAllSyncObjects) - undeleted queries and their backend wrappers
leaked across Destroy/Initialize cycles, stale ids stayed IsQuery == GL_TRUE in
the re-initialized library, and a later delete could hand the old backend's
wrapper to a different backend's DeleteBackendQuery.
- Converted vertex streams and the host-side EBO max-index scan now SyncGpuWrites
before reading the coherent mapping: XFB/SSBO/image writes are merely recorded
at that point, so the conversion read pre-write bytes (the restart-index
rewrite already synced; these two host reads did not).
- Zero-stride converted bindings: both converters rejected stride 0, making the
factory's documented single-element conversion unreachable and silently
dropping every draw using such a binding; the stride is substituted with the
element size for the one-element case.
- DemoteFloat64Pass block relayout: measurement queued into the module eagerly,
so a mid-struct failure left a half-relaid-out block (compacted offsets before
the failing member, 64-bit offsets after) while claiming the block was left
alone. Decoration writes are now collected and committed only when the whole
block measures successfully.
The previous commit's fingerprint was pinned to one array's incidental
dimensions - workgroup exactly 32x16x1, element exactly vec2, length
exactly 32 - which is the auto-exposure reduction and nothing else. The
pack ships the same idiom twice:
- auto-exposure: 32x16 (512 invocations), shared vec2 prefixSumCache[32]
- RTW warp: 1024 invocations, shared float prefixSumCache[64]
so the warp kept writing 128 subgroups into 64 entries on an 8-lane
device and the retrace stayed bit-identically wrong (ssim 0.027902).
Key the fingerprint on the pack's idiom instead of one array's shape: a
workgroup array of 32-bit floats indexed by gl_SubgroupID, fed by a
subgroup scan, whose declared length is below ceil(invocations / native
width). Three properties keep that a targeted repair rather than a
general array resizer:
- the index must BE gl_SubgroupID (through OpCopyObject, a signedness
OpBitcast, or a spill whose every store is that id), so an index
masked or clamped into range is left alone;
- the >= 16-lane early-out is retained, so every module on the devices
the pack was written for passes through byte-identical;
- growth is certified against maxComputeSharedMemorySize using a
natural-alignment layout model, and declined outright when a
declaration cannot be sized, so a patched module can never fail
pipeline creation where the original would not have.
Verified against the shaders the CI trace actually contains: of the 14
compute modules in the fixture exactly these two change, the other
twelve are byte-identical, and all fourteen pass spirv-val. The
integration scenario grows a second case for the 1024-invocation shape;
both abort with heap corruption when the patch is disabled.
iterationRP's Program 203 declares shared vec2 prefixSumCache[32] for a
512-invocation workgroup indexed by gl_SubgroupID; any device narrower
than 16 lanes partitions into more than 32 subgroups and the pack writes
shared memory out of bounds (heap corruption on lavapipe's CPU
rasterizer, ssim 0.028 on the CI retrace). Fix it where the fault lies -
in the fixture - and keep the GL contract sound everywhere else:
- FixIterationRPSubgroupScratchPass: fingerprint-gated SPIR-V pass that
grows exactly that array to ceil(invocations/width) entries on sub-16-lane devices; every other module passes through byte-identical.
- DeriveNumSubgroupsPass stays default-on for the Adreno topology bug
and is made spec-sound: pipelines request REQUIRE_FULL_SUBGROUPS
whenever the workgroup shape makes the flag legal (computeFullSubgroups
enabled, local_size_x a multiple of the native width, subgroup count
within maxComputeWorkgroupSubgroups).
- EmulateSubgroupsPass: 32-lane virtual-subgroup lowering kept in-tree
as a last resort, enabled only by MOBILEGL_MAGMA_EMULATE_SUBGROUP=1 on
devices with no native subgroup support; fails closed on extended
subgroup instructions and on modules whose added scratch would exceed
maxComputeSharedMemorySize.
- IterationRPFirstReductionScenario skips gracefully outside the pack's
16..256-lane source domain; the new IterationRPScratchFixScenario runs
the fixture-shaped reduction on any width and asserts the exact
width-independent total. DriverPost keeps reporting FAIL on
out-of-domain devices.
- Program203 -> IterationRP rename throughout; the per-trace
num_subgroups_quirk plumbing is removed from the trace replayer, JNI
chain, and CI workflows.
Complementary Reimagined would not load through Espryt on Mali: Iris got
GL_FRAMEBUFFER_UNSUPPORTED building its composite framebuffer, because
colortex1 is RGB8_SNORM and colortex2 is RGB16F - three-channel formats
that no real ES driver can render to (EXT_render_snorm covers R/RG/RGBA
only, and the float extensions exclude the RGB forms). The frontend's
probe cache diagnosed this correctly and then had nothing to offer: the
NoThreeChannelRenderTarget widening machinery existed but was gated to
multisample targets alone. llvmpipe turns out to refuse most of the same
attachments - CI retrace stayed green only because a replay never
branches on glCheckFramebufferStatus - so this was never a desktop-vs-
device split, just an unlit path.
The widening now applies to every color-attachable image, renderbuffers
included, riding the driver-probe branch so the native format is still
tried first and substituted only on refusal. One ThreeChannelWidening
table owns the widened (internalformat, format, type) triple per source
format - the previous per-case branches disagreed with each other and
could emit an unuploadable (RGBA16F, GL_RGB, GL_BYTE) combination or
widen into another three-channel format the driver refuses just the
same. Uploads repack three-component client data to four with the
format's own one in the alpha channel (127 is not 1 for RGB8I - the
integer arms carry integer ones); readback drops the synthetic alpha,
derived from the actual image being read, not the bound framebuffer,
so glGetTexImage through a scratch FBO cannot be confused by an
unrelated widened attachment.
Stored alpha on a widened attachment is now an invariant 1.0 rather
than an accident: the color-mask sync clears the alpha bit per draw
buffer (glColorMaski for MRT mixes), and clears route through
glClearBufferfv with alpha substituted on widened slots only -
scissored clears inherit the discipline for free, integer color
buffers keep their explicit integer-clear path, and glGet still
answers the application's own mask. GL_DST_ALPHA blending, blits and
readback therefore all see 1.0 without further interception.
DriverPost grows the rows this bug earned: EXT_color_buffer_float
detection (previously unreferenced anywhere) with a FAIL row when
absent, the missing EXT_render_snorm row, and a three-channel-
attachment row that reports one representative per widening class -
graded so a half-float-only driver warns about the 32-bit float gap
instead of being declared unsupported.
Gates: 606/606 unit at default and with the async kill switch; full
retrace, both backends - the complementary fixtures now run with the
widening ACTIVE on llvmpipe and pass with a slightly better SSIM than
before; ext caselist DirectGLES holds 3914/4867 with zero set drift
while 54 cases move from NotSupported to genuinely passing; on the
Mali-G77 device, Complementary Reimagined builds its pipeline and
renders in-world through Espryt (md5-verified build), BSL still green.
A new ThreeChannelAttachmentScenario pins the frontend answer -
COMPLETE where it used to say UNSUPPORTED - on the real driver.
swung0x48
