marfrit/panvk-bifrost
1
0
Fork 1
Code Issues 1 Pull Requests Actions Packages Projects Releases Wiki Activity

iter19 close + r5 full sweep scoreboard

Browse Source 
Adds iter19 campaign-close doc (mesa-panvk-bifrost r7 ship, XFB
packed-varying channel-extract fix). Also lands the r5 full
dEQP-VK sweep scoreboard (2.26M tests, 97.65% runnable pass rate)
that surfaced the holes_vert crash in the first place.

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
This commit is contained in:
Markus Fritsche
2026-05-25 11:55:53 +02:00
parent 6019cce7d1
commit 02b212b29b
2 changed files with 176 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files
cts-results/r5_full_sweep_2026-05-25.md
+78
View File
@@ -0,0 +1,78 @@
# r5 full dEQP-VK sweep — 2026-05-25
Driver under test: `mesa-panvk-bifrost-26.0.6.r5-1` (fragmentStoresAndAtomics flip)
Hardware: PineTab2 / RK3566 / Mali-G52 r1 MC1 (PAN_ARCH 7)
CTS: vulkan-cts-1.3.10.0 (commit 7aa3551)
Run host: ohm
Run window: 2026-05-24 ~22:00 → 2026-05-25 02:48 CEST (~5h wall)
## Corrected scoreboard
| Metric | Count | % of total |
|-----------------------|--------------|------------|
| Total tests | 2,258,378 | 100.00 |
| Passed | 553,447 | 24.51 |
| Failed | 13,318 | 0.59 |
| NotSupported | 1,691,613 | 74.90 |
| **Pass rate (runnable)** | - | **97.65%** |
| Group status | Count |
|------------------------------------|-------|
| Clean groups (0 failures) | 41/53 |
| Groups with failures | 12/53 |
| Watchdog-killed (data recovered) | 3/53 |
The 3 watchdog-killed groups (`api`, `subgroups`, `transform_feedback`) hit dEQP's
internal watchdog timer and got SIGKILL'd (exit=137), but the per-group `.qpa`
files contain all completed test results — totals were re-derived from those
qpa files directly. The wrapper's `elapsed=3s` field is misleading; the real
runs went the full distance.
## Failed groups (sorted by fail count)
| Group | Fails | Notes |
|---------------------|--------|--------------------------------------------------------------|
| image | 10,445 | 78% of all fails. Single biggest target. |
| subgroups | 2,342 | Bifrost is 4- or 8-wide warps; many fails likely arch-gap |
| glsl | 306 | |
| memory_model | 106 | Bifrost memory ordering — expected partial coverage |
| draw | 68 | |
| transform_feedback | 27 | All `resume_*` — **documented iter17 by-design gap** |
| compute | 8 | |
| pipeline | 5 | |
| descriptor_indexing | 4 | |
| api | 3 | |
| spirv_assembly | 3 | |
| info | 1 | |
## Watchdog-stop tests (last test attempted in each group)
| Group | Last test | Verdict |
|--------------------|-----------------------------------------------------------------------------|------------------|
| api | `dEQP-VK.api.device_init.create_instance_device_intentional_alloc_fail.basic` | watchdog (total time) |
| subgroups | `dEQP-VK.subgroups.clustered.graphics.subgroupclusteredadd_ivec3` | watchdog (touch interval) |
| transform_feedback | `dEQP-VK.transform_feedback.simple.holes_vert` | **userspace coredump, reproducible** — see #107 |
## transform_feedback 27 fails — all known by-design
Every fail is shape `dEQP-VK.transform_feedback.simple.resume_*` —
3 stream-count cells × {`resume`, `resume_beginqueryindexed_streamid_0`,
`resume_endqueryindexed_streamid_0`} × {256, 512, 131072}. All fail at
`vktTransformFeedbackSimpleTests.cpp:949` with `received:N expected:0`.
These match the iter17 closeout note ("remaining 81 fails are by-design
resume_* tests, transformFeedbackDraw=false"). The set here is a strict
subset of the 81; this run only hit the 27 cells that this sweep configuration
exercised. **No r5 regression on iter17 XFB.**
## Next actions
- **Phase 0 on `holes_vert` coredump** — userspace-only (dmesg clean), reproducible in isolation against r6 too. Tracked as task #107.
- **r7 candidate sort** — `image` group's 10,445 fails are the largest mass; needs Phase 0 evidence sort to see if it's one root cause replicated across format/dim cells or many.
## Provenance
- Raw per-group qpa + run.log: `ohm:/home/mfritsche/cts-results/r5_full/`
- Summary log: `ohm:/home/mfritsche/cts-results/r5_full/summary.log`
- Driver: `ohm:/usr/lib/panvk-bifrost/libvulkan_panfrost.so` (r5 then r6)
Aggregation script: see git history of this file.
iter19_campaign_close_2026-05-25.md
+98
View File
@@ -0,0 +1,98 @@
# iter19 — XFB store channel-extract fix for packed varyings (campaign close)
Shipped: 2026-05-25
PR: https://git.reauktion.de/marfrit/marfrit-packages/pulls/96
Merge commit: 902de73a02d9
Package: `mesa-panvk-bifrost-26.0.6.r7-1-aarch64`
## What shipped
Single 21+/3- line patch to `src/panfrost/vulkan/panvk_vX_xfb_lower.c`
(`0007-panvk-bifrost-xfb-component-base-fix.patch`). Eliminates a
reproducible SIGSEGV in `vkCreateGraphicsPipeline` for any shader with
XFB-bound varyings declared at non-zero `layout (component=N)`.
## How surfaced
The r5 full dEQP-VK sweep (2,258,378 tests, 97.65% runnable pass rate)
on 2026-05-24/25 included 27 fails in `transform_feedback` plus a
SIGKILL on the chunk when `transform_feedback.simple.holes_vert` hit
`FATAL ERROR: Test program crashed`. Isolated repro produced a userspace
SIGSEGV with no kernel GPU fault (dmesg clean), pointing at a pure
libvulkan_panfrost bug.
## Root cause
`lower_xfb_output_iter17` (and identically upstream
`pan_nir_lower_xfb.c::lower_xfb_output`, which carries a `// TODO`)
computed the source-channel mask as `mask << channel_idx`. `channel_idx`
is the varying-location component (0..3) but `src` only contains
channels starting at `nir_intrinsic_component(intr)`. For a scalar
declared `layout (component=2) flat out float vegeta`, NIR emits
`store_output src=<vec1>, component=2`, and the lowering computed
`mask << 2` against a 1-component src — out-of-range; the malformed
nir_def then segfaulted during downstream NIR constant-folding in
`nir_constant_expressions.c::evaluate_*`.
The `assert(nir_intrinsic_component(intr) == 0)` precondition was
inherited from upstream Mesa as a documented `// TODO`; release builds
(`-DNDEBUG`) elided it, turning the precondition violation directly
into a SIGSEGV.
## Fix
1. Compute `src_channel = channel_idx - nir_intrinsic_component(intr)`
and use `mask << src_channel` instead.
2. Convert both elided asserts to explicit release-mode early-return
guards (closes the same elision class as the original bug).
3. Add a dispatcher-side comment explaining why `i*2+j` is the
varying-location component index.
## Verification
| Family | Result |
|----------------------------------------------|-------------------|
| `transform_feedback.simple.holes_vert` | Crash → Fail (color) |
| `transform_feedback.simple.holes_extra_draw_vert` | Crash → Fail (color) |
| `transform_feedback.simple.basic_*` | 36/36 Pass |
| `transform_feedback.simple.depth_clip_*` | 1 Pass + 4 NotSupp |
| `transform_feedback.simple.lines_or_triangles*` | 16 NotSupp |
| `transform_feedback.simple.holes_geom*` | NotSupp (no GS on G52) |
Zero new regressions on previously passing tests.
## Process
Full 8-step bugfix process followed:
- Phase 0 — characterize: `iter19/phase0_holes_vert_close.md`
- Phase 1 — source map: `iter19/phase1_holes_vert_situation.md`
- Phase 2 — root-cause + ranked options: `iter19/phase2_holes_vert_situation.md`
- Phase 3 — implement + inner-test: `iter19/phase3_holes_vert_close.md`
- Phase 4 — wider verify (focused subset)
- Phase 5 — 2nd-model review: **APPROVE WITH CHANGES (non-blocking)** → defensive-guards added
- Phase 7 — final retest: clean
- Phase 8 — ship: r7 PR → CI green → marfrit repo published → ohm `pacman -S` upgraded → smoke test confirms Crash→Fail
## Three-point ship-check verdict
1. ✓ PR #96 merged at `902de73a02d9`
2. ✓ CI runs #1380 (`mesa-panvk-bifrost-aarch64`) and #1381 (`mesa-panvk-bifrost-video-aarch64`) both `success`
3. ✓ `pacman -Q mesa-panvk-bifrost` on ohm reports `26.0.6.r7-1`; installed lib BuildID `9f8dacfc...`; holes_vert no longer SIGSEGVs
## Open follow-ups (not blocking r7 ship)
- **iter20: holes_vert color-check residual** — when `panvk_per_arch(nir_lower_xfb)` removes the original store_output post-XFB-lowering, varyings that are *both* XFB-bound *and* read by the fragment shader lose their rasterizer-side path. holes_vert's fragment shader sees `goku`/`vegeta` as undef → outputs black instead of expected blue. Naive "keep store_output" probe fixed color but blew up compile-time on `max_output_components_128` — needs a more nuanced fix.
- **iter20+: pre-existing latent crashes** unmasked by r7 — `dEQP-VK.transform_feedback.simple.max_output_components_{64,128,256}` coredump on shipped r6 baseline too (confirmed independently); they were never reached during the r5 sweep because the watchdog killed transform_feedback after holes_vert.
## Lineage
| rev | What | Date |
|-----|-----------------------------------------------|------------|
| r1 | KHR_robustness2 + nullDescriptor + nullDescriptor on Bifrost | iter8 |
| r2 | has_vk1_1/has_vk1_2 = true on Bifrost | iter9 |
| r3 | VK_EXT_transform_feedback (iter13) | iter13 |
| r4 | XFB primitive decomposition (iter17) | iter17 |
| r5 | fragmentStoresAndAtomics = true | 2026-05-23 |
| r6 | VK_EXT_legacy_dithering | 2026-05-25 |
| **r7** | **XFB packed-varying channel-extract fix** | **2026-05-25** |