P3b/P4b R-5 (and R-3's cheap half), landing under the P7 program (PLAN-PH-P34B-P7.md
1.2, wave 2-D package D2).
The rule had SIX statements and nothing compared them, and two of the four prose ones
were wrong in the same way: MG_Pipe/MGPipe.h and MG_Backend/DirectGLES/Managers.h both
said "THREE OF THEM HAVE A DEPENDENCY" (there are six rows) and both said "the mirror
pairs (10 without 11, ...) are all fine" - when 10-without-11 is precisely D-K2's FOURTH
row (ID-14/ID-15), refused by the server and withheld by the client, with PipeFill.cpp
saying so out loud: "The brief's original 'bit 10 without 11 is fine' is WITHDRAWN for
P4a as built." Neither wrong comment could fail anything; Managers.h's copy sits in the
header of the file that implements the refusal.
The rows, once, in MG_Pipe/SubsystemDeps.def (X-macro, FatalFamilies.def's convention),
expanded by MGPipe.h into kMGPipeSubsystemDependencies[] / MGPipeSubsystemRequires() /
MGPipeSubsystemDependenciesAreSet() with four static_asserts. Six rows: 8->7, 9->10,
10->{7,11}, 11->10, 12->none (a ROW, not an absence), 13->7. Bits 8 and 13 are the two
the old comment did not know about; bit 13's row lived fifteen lines further down in the
same header, which is how "THREE OF THEM" survived two phases.
THE TWO READERS DO NOT YET READ IT, deliberately: MG_Impl/Pipe/PipeFill.cpp is the
contract package's file for the whole phase (TextureEmit.h states that rule) and
MG_Backend/DirectGLES/Managers.cpp belongs to wave 2-D package D1 while it is in flight.
Switching them is two one-line changes for the integrator. The anti-drift property does
not wait: MG_Test/Backend/DirectGLES/SubsystemDepsTest.cpp is the first thing in the tree
that calls BOTH readers at the same mask - sweeping the phase constants, the two
documented refusal lanes and one mask per row - and compares each against the .def. It
lives in that binary because that binary links both sides.
Both comparisons are one-directional, each for a measured reason. The server resolvers
also fold in the family's own bit and D-C3's wire-caps refusal, so a resolver may say NO
for its own reasons but never YES where the table refuses. MGPipeP4aFamilyEmits is a
QUADRUPLE and the first version of this case asserted equality and reddened at 0x1fff on
the sampler and program families because a bare unit binary registers no consumer - the
assertion being wrong, not the product, and recorded in the note because the temptation
to assert equality here is strong.
Red-once (R-16), twice: dropping the fourth row from the .def fails at COMPILE time
(MGPipe.h's static_asserts); making the server's ResolveSamplerSubsystemArm stop checking
bit 10 (throwaway, reverted) reds the runtime case and names the mask - 0x9ff, which is
exactly the lane CMakeLists.txt already pins for G12.
R-3's cheap half: CONTRACT-P5.md claimed the image-access, sampler-view-target and
draw-buffer encodings as contract by pointing at package-header LINE RANGES, and all
three had drifted (ImageEmit.h:146-159 no longer holds the encoding at all). All three
rows now state the encoding by enumerator and function name, and the two confusable
siblings are named. The remaining half - MGPSamplerView::Target still has no Count bound,
no ...IsValid and no Fatal for an out-of-range value - is stated in the row rather than
closed, being an MG_Pipe change outside this package's footprint.
Managers.h:690-695's copy of the wrong prose is NOT corrected here: it is package D1's
file, and its two defects are named in notes/p34b/espryt-d2.md so the integrator can take
them in one line.
G1 after this change: pull .text 10822147 = 0xa52203, nm --defined-only identical to
~/w7/p7-before/pull-syms.txt (30570 symbols) - the new inline constexpr table does not
reach the pull build.
Every place that stated "P5 does not chunk" as current behaviour except the
Fatal texts (previous commit) now records the stage chunk budget and what it
does not cover.
- CONTRACT-P5.md section 5: the STAGE_MB row names MGPipeStageChunkBytes() =
clamp(segment/4, 4096, segment), the two buffer walks (1e7c372e) and one
texture level's whole-width slabs (9469d48e, assembled by
StagedTextureStore::AdoptRun), and the Fatal that stays the backstop for a row
with no cut; the RING_MB row says the record bound is about the record's own
bytes.
- BRIEF-P5.md's R-10 section is no longer headed "does not chunk": it keeps the
record bound and the maxrec proof obligation unchanged and records that
content is cut, with the encoder Fatal as the fallback.
- ROADMAP.md's staging debt row and open question 11, ARCHITECTURE.md's
record-bound bullet and SEG_STAGE table note, README.md's knob row,
MEASUREMENTS.md section 6.3 and CURRENT_STAGE_PROGRESS.md's open item: the
explicit MOBILEGL_IPC_STAGE_MB=256 profile is no longer needed for the two cut
paths, and the carrier design for the rest stays P8.
- Historical results (notes/**/p5-results/, p5b/) are left exactly as recorded.
- The round-2 report guessed the apply thread's ~48% mutex time was the backend
pendingMutex; the symbolized call graph at 56a77348 says it is the idle poll in
ServerLoop::ApplyThreadMain taking m_controlMutex twice per iteration
(ControlIsPending in the ready lambda, PumpControlRequest), with DrainRing at
29% of the thread. Every package below is cut from that graph, not from self%.
- T (transport): the control mailbox gets an atomic shadow bit m_controlPosted so
the idle poll takes no lock (handshake, C2 critical section and the
NOT_INITIALIZED answer unchanged); Doorbell::Wait spins on a once-calibrated
iteration budget and reads no clock in steady state (the park phase alone
computes deadlines); client/server wait and park counts ride the stats line
as wait[srv= srvpark= cli= clipark=] and the wire ledger as cliwait=/clipark=.
- D (role guards): OnApplyThread() is inline - one relaxed atomic key against the
caller's thread pointer (aarch64 __builtin_thread_pointer, else
std::this_thread::get_id) - and m_applyThreadId, the function-static guard and
the pthread_equal call are gone; the three scope depth counters stop being
thread_local (their only reader is the apply-thread guard, so they count only
when OnApplyThread() is true; a GL-thread scope no longer exempts the apply
thread, pinned by a new case); the guard chains are ordered cheapest-first.
- B (persistent maps): the shadow allocator rounds the SIZE to a page multiple, so
TrackWriteMap aligns OUTWARD when the shadow is page-granular - a sub-page map
is one protected page, edges and their per-draw XXH3 disappear; the inward arm
stays as the fallback. Review found a hole the edge hashes had hidden since
round 1: a map that faulted but was not consumed by the walk that advanced the
epoch was never pushed. The walk now drains every marked member before the
binding walk, registration bumps the epoch so a fresh map ships once on the
draw path, and a non-4 KB kernel page refuses the mprotect arm by name.
- C (client per verb): an image-unit high-water mark (push build only, fed by
glBindImageTexture) replaces the 192-unit sweep per draw; the residual fill's
7-term supplied conjunction is a memo keyed on its four inputs; the server's
63-field stamp per verb is a constexpr per-class mask table.
- generate_mipmap no longer skips its wait under MOBILEGL_IPC_BATCH_WAITS: its sink
is the backend GenerateMipmap, which reads MGB_CTX->GetActiveTextureUnit(), and
a batched publish let the next validate move the serial under it -
Fatal{UnmigratedPipeInput, "GetActiveTextureUnit@ReadPixels"} on the gate tree.
The rule is written at the site: a record may skip its wait only if its apply
reads nothing the residual fill writes. The texture-death scenario gets the
framebuffer case's fence before it reads the object_death tally.
- Redmi 2f7cbe2e, FCL + Minecraft 26.3-rc-3, VD12, CPU pinned: inproc 64.3 fps
(client 15.0 ms/frame, apply 12.4) -> 103-114 fps (client 8.7-9.2 ms/frame on a
mid core, apply 7.0-7.4); monolith 115 (120 Hz vsync-capped runs) / 206 (one
uncapped run, 5.0 ms/frame on the big core). Gates: unit 2203/2203,
integration-split 111/111 (the two formerly flaky cases 6/6 on repeat),
generators, include closure, dirty surface and doc citations green.
VD12 Minecraft on the Redmi (FCL, DirectGLES): the split's in-world frame rate
13 -> 60-66 fps against monolith's 110-114, and persistent-map wire bytes
25 MB/frame -> ~0.23 MB/frame. The residual split-specific costs are named at
the end, not hidden: the R-1 serialization (WaitForApplied ~12%, P11's
versioned-inputs design) and a split-only mutex complex still under
investigation.
Persistent-map push, three arms picked per buffer at map time:
- mprotect dirty tracker: the mapped range's page-aligned interior is
protected PROT_READ; the app's next write faults, the SIGSEGV handler marks
the page in the slot's bitmap and unprotects, and the push ships one 64 KB
block per block containing a faulted page. Ownership is three tests -
SEGV_ACCERR only (a userspace raise and an unmapped page chain verbatim),
inside a live tracked range, and not a refault on a still-set bit (a write
we unprotected succeeds, so that is a non-write access and it crashes
honestly). The alignment is INWARD: an outward-aligned edge page holds
foreign bytes, and a signal-blocking worker thread faulting on one is a
process kill - that was the ctest-only SmallRing segfault. Edge spans
(at most two per map) are hashed per push and ship their containing block
only on change (unconditional edges measured 112 MB/frame). Re-arm is one
mprotect per contiguous run. The first push ships everything once.
- hash-suppress arm (MOBILEGL_IPC_PERSISTENT_HASH_SUPPRESS, default 1):
per-block XXH3 against the last push's hashes; untracked buffers only.
- PushDrawConsumers: only the buffers the upcoming draw/dispatch can read
(VAO attribs/element, indexed binding points, indirect/parameter slots,
TexBuffer backing - the last also retires the per-frame-cadence cloud
twitch), plus the fault-epoch skip: an unmoved epoch proves no interior
byte changed, tracked edges are hashed in place, and any hash-arm member
vetoes. The membership map caches the tracker slot pointer (a 64-slot
rescan per buffer per draw was measured and removed).
MOBILEGL_IPC_BATCH_WAITS (default 1; MOBILEGL_PIPE_VERIFY forces 0): a
value-class record (kCtxState/kCtxCso/kCtxObject with no reply slot) is
published without waiting for its own apply; the barrier is taken at the next
pull-reading verb, the only place BARRIER-PULLED fields are read, so the R-1
fence posts are untouched and only the round trips between them change. Two
consequences are pinned rather than argued: kCtxObject verbs the tests read
synchronously get explicit fences (GenerateMipmapFieldsCross,
FramebufferDeathCrosses), and a frontend object whose last SharedPtr drops on
the apply thread - impossible before batching - now enqueues its
(kind, lifetimeId) on the deferred destroy queue and the GL thread replays
the DestroyAndFree helper at the verb hooks (CopyImageLayeredScenario aborted
Fatal{RoleViolation, MGPipeSlots} without it).
Doorbell::Wait's spin reads the clock once per 64 yields: on the workload
device a steady_clock read is a real syscall (no vDSO) and the per-iteration
reads were ~20% of the whole process at ~1000 waits/frame.
Gates: gen_pipe --check/--self-test green; unit 2188/2188;
integration-split 110/111, the one red being GenerateMipmapDepthPixels whose
DirectGLES and DirectVulkan MONOLITH siblings flake identically with every
code path of this package provably inert (PushIsArmed false, no ClientSession,
deferral's monolith early-out).