summaryrefslogtreecommitdiffstats
path: root/STATE-CURRENT.md
diff options
context:
space:
mode:
Diffstat (limited to 'STATE-CURRENT.md')
-rw-r--r--STATE-CURRENT.md61
1 files changed, 61 insertions, 0 deletions
diff --git a/STATE-CURRENT.md b/STATE-CURRENT.md
index 0414c56..f1d8313 100644
--- a/STATE-CURRENT.md
+++ b/STATE-CURRENT.md
@@ -1,5 +1,66 @@
# Open-IOV Current State
+> **Snapshot: 2026-11-16 ~00:35 (SESSION 157 continued; HEAD `88d5005a` ==
+> `vgpu/master`; working tree clean; supersedes the 2026-11-15 ~16:35 entry
+> below).**
+> The graphics front crossed a second major gate since the last snapshot:
+> **the `DRAW_IAB` macro now executes end to end**, and the remaining wall is
+> the **store** (the draw's vertex buffer and render-target visibility).
+>
+> - c788 (module `6aeb20c7`) is the milestone: the chained saved-window
+> release fired (`MME-CARRY-FIRE mode=release held=68 win=240 push=0x6d6d6500`)
+> -> `MME-PACK-REARM gen=5` -> `MME-SA-REARM 7 slot(s)` ->
+> `MME-DRAW-PLAN slot=34 src_off=0x334 dst_off=0x204 span=14` ->
+> `MME-DRAW-MOVE` -> `MME-FAST ... macro-status=00122266(mpc=614 mac=0x22
+> run=1)` — the draw call runs. Stores still 0 (`NOWRITE-CONFIRMED
+> wrote=0`; two operand pages changed).
+> - TCS-off (`45dbcdcd`, eng65) clears the DS latch entirely (the
+> `0x84000420` was the synthesized TCS bind); c792 has no `DS-FAST` line.
+> - The `DRAW_IAB` body was decoded with the public Mesa MME toolchain
+> (`mme_tu104_sim`): slot 34 (`0x265`) is a 25-instruction macro that emits
+> **8 draws of 3 vertices** (`SET_VERTEX_ARRAY_START` 0..21,
+> `DRAW_VERTEX_ARRAY` 3). The macro is correct; the no-store cause is
+> downstream.
+> - eng66-68 (`ef2e68b9`, `dab4a02c`, `3481e62d`): DRAW-STATE/CALL-ARGS
+> census, synthetic full-RT viewport, release chained after any GR launch.
+> - eng69-74 (`5ff47a95`..`88d5005a`): vertex-stream census/persistence and
+> `sema_gl_vs_autobind` (scan a launch data page for a float2 vertex block
+> and prepend VS0 + the Turing size pair on release saves); `GUEST-WIT`
+> widened to launch pages and the big-surface alias pages.
+> - c807/c811/c817/c819/c820/c822/c823: the replay is now **syntactically
+> complete** — call order, SA[34], DS (TCS-off clean), viewport 1920x1080,
+> RT0/RT1 alias, nonzero `VS0`/`SZ0`, `mac=0x22` — yet the render target
+> still receives nothing (`wrote=0`, `0/1025`) and `GUEST-WIT` shows only
+> CB/GR-context/report pages written. The 151-word DRAW window carries no
+> vertex format/location/size at all, and the client's captured pushes
+> never bind the stream directly (the only `0x1c00` census hits are
+> SPH-payload false positives).
+> - **Remaining hard blockers:** (a) the true per-draw vertex buffer is
+> unidentified (the eng71 scan's CB-page candidate is a heuristic, score 8;
+> a nonzero stale `VS0` also stores nothing); (b) render-target
+> store/alias visibility is unproven (the big-surface alias pages do not
+> populate the WIT, so an RT store would be invisible today). eng74 (`88d5005a`)
+> dumps the scanned candidate blob (`VS-BLOB`) and records the
+> surface-alias RT pages in the `GUEST-WIT`, so both blockers are
+> measurable on the next boot.
+> - Throughput: the pack-completion and shallow-boot/X-start lotteries
+> continue (c783-c785, c789-c791, c804 shallow; c823 ran 4 frames black).
+>
+> CUDA front unchanged (backgrounded while graphics holds the device; its
+> 10-30 status stands): `41343c63` green (`victory_mail` `VADD bad=0` +
+> `STATUS OK`; `vecadd_loop` 100 launches PASS); the fresh-boot battery's red
+> class is D2H-stage with release-discipline fixes staged; the SESSION 143
+> Vulkan fence regression is not re-measured.
+>
+> Live at check (read-only): host uptime 14 h 32 m (boot 2026-11-15 ~10:03);
+> `vgsp` loaded (refcount 0) + mdev
+> `3b5523a9-ef3a-4f83-8091-da29dce26265` + QEMU cycle c823 (PID 109722, 12 GiB
+> file-backed guest, kvmfr ivshmem, since 19:36); X alive (display-channel
+> polls advancing); `nvidia-smi` reports no GPU (the claim is held); pstore
+> empty; no Xid/Oops; no reboot during the check. The open-iov.org wiki was
+> refreshed alongside this snapshot (11 pages, revs 24199-24209; 68 pages
+> audited, 0 problems).
+>
> **Snapshot: 2026-11-15 ~16:35 (SESSION 157 continued; HEAD `6aeb20c7` ==
> `vgpu/master`; the working tree carries only this docs refresh; supersedes
> the 2026-11-15 ~13:55 entry below).**