nx2: notes 123-130 (fences, lock handoffs, translation cache crash, macro parameters, counter polling, worker wake-ups, memo compaction, measurement pitfalls)

1 files changed, 10 insertions, 0 deletions

diff --git a/NX2_NOTES.md b/NX2_NOTES.md
index 28f58aed43..4eab9301ef 100644
--- a/NX2_NOTES.md
+++ b/NX2_NOTES.md
@@ -184,6 +184,14 @@ With the patched decoder, games 1/3/4 only leave ~60-110 rare SIMD encodings in
 | 126 | same abort (row 125), then inside `ExecuteVirtualMachineMapBufferEx` | NVN describes and maps images another server made with gsv 428 `nn::gpu::sdk::QueryNativeImage` and 426 `MapNativeImage`(handle, u32, address space, u8 permission). Reply: words 3-4 GPU address (426), 5-6 offset in the memory, 7-38 a `GpuOunceUserImageDescriptor`, 39 flags. The SDK requires descriptor byte 0x00 == 1, byte 0x20 == 1, u32 size at 0x2c != 0 (offset at 0x28); NVN reads (w-1)|(h-1)<<16 at 0x08 and the layout kind at 0x24 (1: formats 0x25/0x37/0x38, 2: 0x99-0x9c; sdk+0x6bb960) | both answered from the pool's preference and memory; block height/pitch at 0x10/0x14 are a guess (only blank camera frames use them) |
 | 127 | abort 2034-0000 at main+0xcded94 (~125 s, after the normal graphics boot) | `nn::camera::sdk::CreateCameraContext` sends gsv 601() then 602(u64 1), which must return the camera context session (handle word 3, result word 4); without it the SDK sends 603 (destroy) and fails | 601 stub, 602 returns a `nx2:camera` stub session. Title 1 (1.8.0) then plays |
 | 128 | (title 1, 1.8.0, Ryzen 5 2600 + RTX 3050 6 GB) 8 fps | Eden defaults: gpu_accuracy High, uncompressed ASTC, reactive flushing on | gpu_accuracy Low, astc_recompression Bc3, use_reactive_flushing off, async shaders on: ~20 fps, uneven. Bushes/cacti flicker (row 97). Boot also froze once at "Failed to allocate 20480 MiB of backing memory": the 12 GB layout needs 20 GB of free commit up front |
+| 123 | (title 1) 5,500-7,000 fences a second, one per compute release, each made in full (both cache locks twice, a fence object, a fencing-thread wake-up) even with nothing to wait for: SignalFence 7.6% of the GPU thread | with fast GPU fences the release callback runs at once; ~20% of the fences had no downloads, queries or operations. OverlapRangeSet::Subtract restarted its search after every erase | empty fences skipped (one in 64 still made), one lock per cache per fence, single-pass Subtract (9905916): SignalFence 3.9% |
+| 124 | GPU thread ~9% spinning on, sleeping on and waking for the cache locks | the fencing thread took the buffer cache once per fence to write 250-330 MB/s of downloaded buffer data back (13,000-20,000 ranges a second) | signaled fences released together (31cd7b2, `EDEN_NX2_FENCE_BATCH=0` off): spinning 7.0% -> 5.5%; fence cost per call unchanged within noise (longer holds). Open: the downloads themselves |
+| 125 | crash every few minutes: BindHostGraphicsUniformBuffer -> ReadBlockUnsafe, access violation at 0x600 (2 of 4 runs) | DeviceMemoryManager's translation cache was shared by all threads; Map on a service thread (nx2:gpu 211 job mappings, logged at both crashes) cleared it between the GPU thread's null check and its copy | per-thread cache invalidated by a generation counter (1a497fa): 0 crashes in 6 runs |
+| 126 | macro parameters appended to a std::vector per call (~2% of the GPU thread) | ~1,000,000 calls a second, nearly all whole in one method run | stack copy (3caf8ea, `EDEN_NX2_MACRO_COPY=1` off). Passing the command list words in place is unsafe: they are often guest memory read in place, and the memo stores the parameters after the call |
+| 127 | main core: ~23% of its samples read CNTPCT (~17 million reads a second) | the game polls the counter while it waits for the GPU; a spinning thread slows its SMT sibling (8-core/16-thread Ryzen) | PAUSE between counter reads less than ~1.7 us apart (8abeb0b, `EDEN_NX2_CNTPCT_PAUSE=0` off); effect not separable from run-to-run noise |
+| 128 | Vulkan worker woken with a system call every 8 draws (~2% of the GPU thread) | DispatchWork notified even when the worker was busy | every 32 draws, notify only a waiting worker (d9d7e94) |
+| 129 | macro memo arena (16 MiB) cleared every ~2 s | records stored again with another shape left their old copies | compaction (9a37142); live records are ~12 MiB, so most compactions still end in a clear; hits stay ~97% (46% on the last call's slot) |
+| 130 | measurements | a stuck `nvidia-smi -l` loop had spun one core for 12 h (every run since the evening of 10-28 had one core less); the attract loop shows different courses each run (frame means vary +-15%); stopped test loops kept running and overlapped | compare per-call costs (`C:\dev\nx2work\zones.py` on `EDEN_NX2_GPU_PROFILE` logs); `ab_env.sh` and `hangcheck.sh` refuse to start while eden-cli runs; `stopall.ps1` stops every loop |
 
 Run settings used (in `build/bin/user/config/sdl2-config.ini`): `memory_layout_mode=4`
 (12 GB). Debug knobs: `EDEN_NX2_TC_EXPECTED_MIB` / `EDEN_NX2_TC_CRITICAL_MIB` (collector thresholds),
@@ -193,6 +201,8 @@ faulting stack as module+offset, then a minidump), `EDEN_NX2_AUTO_STUB`, `EDEN_N
 `EDEN_NX2_GPU_PROFILE` (GPU-thread zone timers every 10 s), `EDEN_NX2_SAMPLE=<threads>` with
 `_START`/`_SECONDS`/`_OUT` (sampling profiler; `C:\dev\nx2work\folded.py` reads its .folded stacks),
 `EDEN_NX2_MME_RECORD=<file>` (macro calls for `C:\dev\nx2work\mme_test`),
+`EDEN_NX2_FENCE_BATCH=0`, `EDEN_NX2_MACRO_COPY=1`, `EDEN_NX2_CNTPCT_PAUSE=0` (turn rows 124, 126, 127 off),
+`EDEN_NX2_MME_MEMO_STATS` (memo counters),
 `EDEN_NX2_FRAME_CAPTURE=<dir>` (every 120th presented frame as TGA), `EDEN_NX2_PB_CHECK`
 (pushbuffer entries changed between submit and execution), `EDEN_NX2_QUERY_TRACE` (every GPU
 report/semaphore write). `C:\dev\crashrun.sh` runs eden-cli under cdb and prints the crash stack;