pin v2 payload: engine 2610141212001, cuda 2610141455001, cuda12 2610141455002, sbsa 2610141455003, vulkan 2610141455004, macos 2610141455005, media 2610071902001

.gitattributes

@@ -285,3 +285,12 @@ components/sbsa/2610131352003/ggml-cuda-sbsa-aarch64.so filter=lfs diff=lfs merg
 components/vulkan/2610131352004/ggml-vulkan-x86_64.so filter=lfs diff=lfs merge=lfs -text
 components/vulkan/2610131352004/ggml-vulkan-win-x86_64.dll filter=lfs diff=lfs merge=lfs -text
 components/macos/2610131352005/ggml-metal-aarch64.dylib filter=lfs diff=lfs merge=lfs -text
+components/engine/2610141212001/opencoti-0.10.5-c13-2610141212001 filter=lfs diff=lfs merge=lfs -text
+components/engine/2610141212001/opencoti-0.10.5-c13-2610141212001.exe filter=lfs diff=lfs merge=lfs -text
+components/cuda/2610141455001/ggml-cuda-x86_64.so filter=lfs diff=lfs merge=lfs -text
+components/cuda/2610141455001/ggml-cuda-win-x86_64.dll filter=lfs diff=lfs merge=lfs -text
+components/cuda12/2610141455002/ggml-cuda-cu12-x86_64.so filter=lfs diff=lfs merge=lfs -text
+components/sbsa/2610141455003/ggml-cuda-sbsa-aarch64.so filter=lfs diff=lfs merge=lfs -text
+components/vulkan/2610141455004/ggml-vulkan-x86_64.so filter=lfs diff=lfs merge=lfs -text
+components/vulkan/2610141455004/ggml-vulkan-win-x86_64.dll filter=lfs diff=lfs merge=lfs -text
+components/macos/2610141455005/ggml-metal-aarch64.dylib filter=lfs diff=lfs merge=lfs -text

components/cuda/2610141455001/BUILD_INFO.cuda.md

@@ -0,0 +1,32 @@
+# cuda 2610141455001
+
+Component of the opencoti pin format v2 (docs/protocols/PIN_FORMAT.md).
+
+Built from the engine source of build 2610141255001; gated floor (engine-min) 2610141212001; gated with 2610141212001, 2610141255001.
+
+| file | platform | kind | sha256 | bytes |
+|---|---|---|---|---|
+| `ggml-cuda-x86_64.so` | x86_64 | cuda | `1beda05b1dbf406796975bedfdd55cdd51ad4bc07696ba28c3aa0ca0e3c8acbb` | 757231168 |
+| `ggml-cuda-win-x86_64.dll` | win-x86_64 | cuda | `889e797020e3a3b9ef7d8b781109d3d2a927db287af3167261e6e030e208ee3a` | 731318272 |
+
+ABI `ggml` = `bac5050ddc8c58c51f0b059a7f9753b33af0ed79d355090e7c9a17ffaaaf340a` (exported) — sha256 of this list:
+
+```
+llama.cpp/ggml/include/ggml-alloc.h 94e4cd069b9313b2ceb35dacec901981e0bb478d8bb31035b7126be091998c23
+llama.cpp/ggml/include/ggml-backend.h 764d1640ad62579513c4837539708f89e8603c9dbeae053a13d1cbd84a82c865
+llama.cpp/ggml/include/ggml-cuda.h 674ea064021cf5b9970e78f1209e513dd23350b94a44a705a27eec1423fe4b0a
+llama.cpp/ggml/include/ggml-metal.h 322f36cd30f3e9e7aad7b5b9bc63078012fd0b7706ac4e24f5721b604f3d8980
+llama.cpp/ggml/include/ggml-neo-pipeline.h 190f39e183542d16588a3104d8cb4359bf52fc7a62afe8d8cc4c822f4085086d
+llama.cpp/ggml/include/ggml-oc-kvarn-range.h 76635fdd8978632f670640ddf5125a666441e3d47c680bb8b5c0a099e4861202
+llama.cpp/ggml/include/ggml-oc-ledger.h c1be11d69a57a1e721f1ae8720b87887d239ac0183d4277fbaa07f1be257d20c
+llama.cpp/ggml/include/ggml-vulkan.h f9df4f7161a84ce7600d53c3fac8a62837d5716676c9f38cd340699ef1c070d9
+llama.cpp/ggml/include/ggml.h ecfbf114f9e9151d22e30ef6d5139570c5c41b595843ca4f99125bd0196709aa
+llama.cpp/ggml/include/opencoti-devtype.h 94dea946173449bb76bffd7f733ab822fe658fe7198992de750ac012d29308ec
+llama.cpp/ggml/src/ggml-backend-impl.h f9fdfe6bae61d1387593b40fc0f90ceee490c6a339ec6d7388c51e4fca318d8f
+llamafile/gpu_backend.h 4929c36c034b6589e88653bcb7a5c5a225743a07320f64c0857d92bdaf048208
+```
+
+CUDA 13. Six generations (sm_75, 80, 86, 89, 90, 120); `120` is the sm_120f kernel set. Built from the clean-room tree of
+chain 0662 on a cloud runner, from the c12 round's object cache (the interface is unchanged): the six-generation library compiled 36 of its 405 units (the KVarN decode instances 0657 changed) and reused 369 from the c12 round's cache by content; the Windows DLL compiled the same 36 units plus ggml-backend.cpp under wine on bs2 and reused the c12 release's pinned objects by content (the set the content diff of the two source trees names). The Windows DLL is
+the same six generations (wine lane). Carries the lazy KVarN stage of 0657.
+Gated with the 2610141255001 engine: bs2 (sm_120) Bonsai 2 gate 56 / 56 and the slot gate 6 / 6; the glibc 2.28 image; the Windows DLL on eleven2go (RTX 3090, sm_86): the Windows GPU gate's CUDA cases, 77.4 tok/s against the c12 control's 77.5 with the same 64 tokens, no ABI mismatch, memory on the device the counter names, no spill. sm_75 / 80 / 89 / 90 were not run on a card at this cut (sm_86 = the RTX 3090, sm_120 = bs2).

components/cuda/2610141455001/ggml-cuda-win-x86_64.dll

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:889e797020e3a3b9ef7d8b781109d3d2a927db287af3167261e6e030e208ee3a
+size 731318272

components/cuda/2610141455001/ggml-cuda-x86_64.so

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1beda05b1dbf406796975bedfdd55cdd51ad4bc07696ba28c3aa0ca0e3c8acbb
+size 757231168

components/cuda12/2610141455002/BUILD_INFO.cuda12.md

@@ -0,0 +1,29 @@
+# cuda12 2610141455002
+
+Component of the opencoti pin format v2 (docs/protocols/PIN_FORMAT.md).
+
+Built from the engine source of build 2610141255001; gated floor (engine-min) 2610141212001; gated with nothing — NOT run.
+
+| file | platform | kind | sha256 | bytes |
+|---|---|---|---|---|
+| `ggml-cuda-cu12-x86_64.so` | x86_64 | cuda12 | `6e9f87bbff798cf4b33f2504aab819da5b33b91c334dc2003a7c30d3f9ae32e5` | 261655040 |
+
+ABI `ggml` = `bac5050ddc8c58c51f0b059a7f9753b33af0ed79d355090e7c9a17ffaaaf340a` (exported) — sha256 of this list:
+
+```
+llama.cpp/ggml/include/ggml-alloc.h 94e4cd069b9313b2ceb35dacec901981e0bb478d8bb31035b7126be091998c23
+llama.cpp/ggml/include/ggml-backend.h 764d1640ad62579513c4837539708f89e8603c9dbeae053a13d1cbd84a82c865
+llama.cpp/ggml/include/ggml-cuda.h 674ea064021cf5b9970e78f1209e513dd23350b94a44a705a27eec1423fe4b0a
+llama.cpp/ggml/include/ggml-metal.h 322f36cd30f3e9e7aad7b5b9bc63078012fd0b7706ac4e24f5721b604f3d8980
+llama.cpp/ggml/include/ggml-neo-pipeline.h 190f39e183542d16588a3104d8cb4359bf52fc7a62afe8d8cc4c822f4085086d
+llama.cpp/ggml/include/ggml-oc-kvarn-range.h 76635fdd8978632f670640ddf5125a666441e3d47c680bb8b5c0a099e4861202
+llama.cpp/ggml/include/ggml-oc-ledger.h c1be11d69a57a1e721f1ae8720b87887d239ac0183d4277fbaa07f1be257d20c
+llama.cpp/ggml/include/ggml-vulkan.h f9df4f7161a84ce7600d53c3fac8a62837d5716676c9f38cd340699ef1c070d9
+llama.cpp/ggml/include/ggml.h ecfbf114f9e9151d22e30ef6d5139570c5c41b595843ca4f99125bd0196709aa
+llama.cpp/ggml/include/opencoti-devtype.h 94dea946173449bb76bffd7f733ab822fe658fe7198992de750ac012d29308ec
+llama.cpp/ggml/src/ggml-backend-impl.h f9fdfe6bae61d1387593b40fc0f90ceee490c6a339ec6d7388c51e4fca318d8f
+llamafile/gpu_backend.h 4929c36c034b6589e88653bcb7a5c5a225743a07320f64c0857d92bdaf048208
+```
+
+CUDA 12.8, for GPUs older than sm_75 (sm_52, 61, 70). Chain 0662, from the c12 object cache. **These bytes have not been run
+on a card** (no such GPU reachable at the cut).

components/cuda12/2610141455002/ggml-cuda-cu12-x86_64.so

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6e9f87bbff798cf4b33f2504aab819da5b33b91c334dc2003a7c30d3f9ae32e5
+size 261655040

components/engine/2610141212001/BUILD_INFO.engine.md

@@ -0,0 +1,77 @@
+# engine 2610141212001
+
+Component of the opencoti pin format v2 (docs/protocols/PIN_FORMAT.md).
+
+| file | platform | kind | sha256 | bytes |
+|---|---|---|---|---|
+| `opencoti-0.10.5-c13-2610141212001` | any | bin | `6670a23e952307237d8acf58abdb0eebc656491e5e13bc679c1807356e7a6ead` | 143110004 |
+| `opencoti-0.10.5-c13-2610141212001.exe` | win-x86_64 | bin | `6670a23e952307237d8acf58abdb0eebc656491e5e13bc679c1807356e7a6ead` | 143110004 |
+
+ABI `ggml` = `bac5050ddc8c58c51f0b059a7f9753b33af0ed79d355090e7c9a17ffaaaf340a` (exported) — sha256 of this list:
+
+```
+llama.cpp/ggml/include/ggml-alloc.h 94e4cd069b9313b2ceb35dacec901981e0bb478d8bb31035b7126be091998c23
+llama.cpp/ggml/include/ggml-backend.h 764d1640ad62579513c4837539708f89e8603c9dbeae053a13d1cbd84a82c865
+llama.cpp/ggml/include/ggml-cuda.h 674ea064021cf5b9970e78f1209e513dd23350b94a44a705a27eec1423fe4b0a
+llama.cpp/ggml/include/ggml-metal.h 322f36cd30f3e9e7aad7b5b9bc63078012fd0b7706ac4e24f5721b604f3d8980
+llama.cpp/ggml/include/ggml-neo-pipeline.h 190f39e183542d16588a3104d8cb4359bf52fc7a62afe8d8cc4c822f4085086d
+llama.cpp/ggml/include/ggml-oc-kvarn-range.h 76635fdd8978632f670640ddf5125a666441e3d47c680bb8b5c0a099e4861202
+llama.cpp/ggml/include/ggml-oc-ledger.h c1be11d69a57a1e721f1ae8720b87887d239ac0183d4277fbaa07f1be257d20c
+llama.cpp/ggml/include/ggml-vulkan.h f9df4f7161a84ce7600d53c3fac8a62837d5716676c9f38cd340699ef1c070d9
+llama.cpp/ggml/include/ggml.h ecfbf114f9e9151d22e30ef6d5139570c5c41b595843ca4f99125bd0196709aa
+llama.cpp/ggml/include/opencoti-devtype.h 94dea946173449bb76bffd7f733ab822fe658fe7198992de750ac012d29308ec
+llama.cpp/ggml/src/ggml-backend-impl.h f9fdfe6bae61d1387593b40fc0f90ceee490c6a339ec6d7388c51e4fca318d8f
+llamafile/gpu_backend.h 4929c36c034b6589e88653bcb7a5c5a225743a07320f64c0857d92bdaf048208
+```
+
+ABI `media` = `b099409b3029a6e380268e31d1caefa6706569961d32e72b3543fbc73903d95c` (exported) — sha256 of this list:
+
+```
+llamafile/oc-audiocpp/audiocpp.h 6ad0b55d5f28adc9ad60a45bf9a7caa428c4f5441389a6932d6670a8e2a90d73
+llamafile/oc-codec/oc_codec_sidecar.h f2c9001d1414aa2a55533b867b79c52a52c682558e0234c5bebd30bf97b91005
+```
+
+ABI `loader` = `d6ff34022f1d3fb5af56ca0c80e3f4e552bf09f1bebdf0cb5f314a454bf38dbc` (exported) — sha256 of this list:
+
+```
+.cosmocc/4.0.2/bin/ape-m1.c 78af79e20abbb3f99a97355959b59bd58949564064adb754f0302ebabb4bf7a3
+```
+
+**c13 dev snapshot — the release candidate of c13 for integration testing.** Engine 2610141212001 is the dev-channel build of chain 0001…0662, the source the c13 release engine 2610141255001 is built from in the clean room; the gates below ran on the release engine and on the libraries of this index; this dev engine itself ran the glibc 2.28 route gate with this index's Linux CUDA and Vulkan libraries (2026-10-14, every read route twice and a clean shutdown, both PASS). EVERY GPU library of this index and the Metal library are NEW builds of chain 0662 (the `ggml` digest is c12's: a c12 library loads, but lacks the lazy slot memory); the Metal library is signed and notarised; the media sidecars and the macOS loader are the c12 files.
+
+What is new since c12 (r2):
+
+- **The CPU backend runs at the CPU's own instruction set** (0660, bug-4060): the CPU kernels were compiled once, at the
+  x86-64 baseline (SSE2, no AVX2 / FMA) — IQ-type experts on the CPU ran at a fifth of stock llama.cpp. The CPU backend is now
+  built four times (x86-64: baseline, AVX2 + FMA, AVX-512, AVX-512 + VNNI / VBMI; arm64: v8.0, v8.2 dot product, v8.2 + i8mm)
+  and the engine picks the best copy for the CPU it starts on; `OPENCOTI_CPU_VARIANT=<n>` selects a lower one. One file still
+  runs on an SSE2-only x86-64 or an armv8.0 CPU. On an EPYC 9575F (Zen 5), Qwen3.6-35B-A3B UD-IQ3_XXS on the CPU: 6 threads
+  18.9 tok/s (c12: 8.2; stock llama.cpp b10760: 20.4), 1 thread 6.7 (c12: 1.8; stock 7.5).
+- **Whistle's decoder too** (0661): the beam search of the speech-to-text model runs in the same variant; on the EPYC 30 FLEURS
+  clips on the CPU take half the time of c12, word timings unchanged; live word delay median 2.2 s (c12: 3.2 s).
+- **Parallel slots cost nothing until used** (0657, 0659, bug-4061): the default four slots no longer commit four slots'
+  recurrent state and KVarN stage at load — the default declares one live slot and each further slot takes its memory when it
+  takes a request (an explicit `-np N` still commits N). On CUDA, Vulkan, the CPU (Linux / macOS) and Metal. A Bonsai 2 27B on
+  an 8 GB card: all 65 layers on the GPU at the default (c12: 63, at a third of the speed). When the memory for a further slot
+  is not there, the server reclaims the least recently used idle slot or the request waits — never a refusal. Off switches per
+  backend: `OPENCOTI_VK_RESERVED=0`, `OPENCOTI_CPU_RESERVED=0`, `OPENCOTI_METAL_RESERVED=0`.
+- **`reasoning_effort` at the top level of a chat request reaches the template** (0658, ollama #1301): a template with levels
+  rendered its default before; `"none"` turns thinking off as before.
+- **LFM2 with more than one sequence** (0656, in c12 since r2): no abort and no wrong state with parallel requests.
+- **Generation notices — a measuring PROTOTYPE, off unless requested; not the interface** (0655): the engine can plant one
+  line telling the model how much room is left (`[N tokens left]`, rising in urgency) at a break between two pieces of prose —
+  never inside code, a command, a query or a table. The prototype takes the request fields `notice_think_left`,
+  `notice_answer_left`, `notice_think_text`, `notice_answer_text`, `notice_think_steps`, `notice_answer_steps`; none set (the
+  default) = off. These names exist for measurement and WILL change: the agreed interface — one `notices` request object,
+  announced by the feature `generation_notice_v1` — is not in this engine, and this engine does not announce that feature. Do
+  not build a client on the prototype fields.
+- **Interface**: the `ggml`, `media` and `loader` digests are those of c12. The GPU libraries and the Metal library of this
+  index are new builds (0657 changes a CUDA kernel, 0659 the Vulkan and Metal libraries); a c12 library loads on this engine
+  by digest but lacks the lazy slot memory.
+
+Gated on these bytes (2026-10-14): bs2 (RTX PRO 6000 Blackwell, sm_120) Bonsai 2 CUDA + Vulkan gate 56 / 56 — KLD against the reference CUDA base 0.000168 (CUDA) / 0.000311 (Vulkan), needle 3 / 3 at 112k, PQ2_0 tg 126.8 / PTQ1_0 140.5 tok/s, two non-Bonsai models bit-identical between the c12 and the c13 Vulkan library; the default slots cost nothing at load on CUDA, Vulkan and the CPU, 6 / 6 each (Bonsai 2 27B at a 7.1 GB budget: c12 63 or 62 of 65 layers at 52 tok/s, c13 65 of 65 at full speed, +48 MiB CUDA / +11 MiB Vulkan); the glibc 2.28 image, Vulkan and CUDA; Mac mini Whistle 254 / 254 (Metal + CPU, word error rate 17.41 % on 827 words, 16.08 % with the detector), signal stop 9 / 9, single-prompt teardown 7 / 7; eleven2go (Windows 11, RTX 3090) with this index's Windows libraries: the GPU gate 13 / 13 (CUDA 77.4 tok/s, the c12 control 77.5, the same 64 tokens; no ABI mismatch, memory figures equal to Windows' counter, no spill; Vulkan 48.4 tok/s) and Whistle on the CPU 48 / 48.
+
+Known at this cut: on one thread the CPU path is about 12 % slower than stock llama.cpp on the EPYC (6 threads: 7 %); the
+Whistle decoder reads its weights once per beam per token, so it is bound by one core's memory reads (c13 row 40.6 will read
+them once per token); no KVarN on Metal (c13 row 15.1); the standalone transcribefile / whisperfile tools keep the baseline
+CPU code.

components/engine/2610141212001/opencoti-0.10.5-c13-2610141212001

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6670a23e952307237d8acf58abdb0eebc656491e5e13bc679c1807356e7a6ead
+size 143110004

components/engine/2610141212001/opencoti-0.10.5-c13-2610141212001.exe

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6670a23e952307237d8acf58abdb0eebc656491e5e13bc679c1807356e7a6ead
+size 143110004

components/macos/2610141455005/BUILD_INFO.macos.md

@@ -0,0 +1,40 @@
+# macos 2610141455005
+
+Component of the opencoti pin format v2 (docs/protocols/PIN_FORMAT.md).
+
+Built from the engine source of build 2610141255001; gated floor (engine-min) 2610141212001; gated with 2610141255001.
+
+| file | platform | kind | sha256 | bytes |
+|---|---|---|---|---|
+| `ape-macos-aarch64` | macos-aarch64 | ape | `07f550415a59415973ef877eec99a776dfa5263b06fdb1fd9cefb7b884f0f723` | 56896 |
+| `ggml-metal-aarch64.dylib` | macos-aarch64 | metal | `82449bd0be6041f991ef001e6ea6cbeb057a238e33b75fa9ea3b9d47b5029e25` | 1652688 |
+
+ABI `ggml` = `bac5050ddc8c58c51f0b059a7f9753b33af0ed79d355090e7c9a17ffaaaf340a` (exported) — sha256 of this list:
+
+```
+llama.cpp/ggml/include/ggml-alloc.h 94e4cd069b9313b2ceb35dacec901981e0bb478d8bb31035b7126be091998c23
+llama.cpp/ggml/include/ggml-backend.h 764d1640ad62579513c4837539708f89e8603c9dbeae053a13d1cbd84a82c865
+llama.cpp/ggml/include/ggml-cuda.h 674ea064021cf5b9970e78f1209e513dd23350b94a44a705a27eec1423fe4b0a
+llama.cpp/ggml/include/ggml-metal.h 322f36cd30f3e9e7aad7b5b9bc63078012fd0b7706ac4e24f5721b604f3d8980
+llama.cpp/ggml/include/ggml-neo-pipeline.h 190f39e183542d16588a3104d8cb4359bf52fc7a62afe8d8cc4c822f4085086d
+llama.cpp/ggml/include/ggml-oc-kvarn-range.h 76635fdd8978632f670640ddf5125a666441e3d47c680bb8b5c0a099e4861202
+llama.cpp/ggml/include/ggml-oc-ledger.h c1be11d69a57a1e721f1ae8720b87887d239ac0183d4277fbaa07f1be257d20c
+llama.cpp/ggml/include/ggml-vulkan.h f9df4f7161a84ce7600d53c3fac8a62837d5716676c9f38cd340699ef1c070d9
+llama.cpp/ggml/include/ggml.h ecfbf114f9e9151d22e30ef6d5139570c5c41b595843ca4f99125bd0196709aa
+llama.cpp/ggml/include/opencoti-devtype.h 94dea946173449bb76bffd7f733ab822fe658fe7198992de750ac012d29308ec
+llama.cpp/ggml/src/ggml-backend-impl.h f9fdfe6bae61d1387593b40fc0f90ceee490c6a339ec6d7388c51e4fca318d8f
+llamafile/gpu_backend.h 4929c36c034b6589e88653bcb7a5c5a225743a07320f64c0857d92bdaf048208
+```
+
+ABI `loader` = `d6ff34022f1d3fb5af56ca0c80e3f4e552bf09f1bebdf0cb5f314a454bf38dbc` (exported) — sha256 of this list:
+
+```
+.cosmocc/4.0.2/bin/ape-m1.c 78af79e20abbb3f99a97355959b59bd58949564064adb754f0302ebabb4bf7a3
+```
+
+Apple notarisation (Accepted), submission id(s): `ca3dd0d1-2833-4a85-8aee-71b39952c641 eea7c66d-e8a4-4860-b9da-71e9416a2805`.
+
+The prebuilt APE loader (the c12 file, `loader` digest unchanged) and a NEW prebuilt Metal library, built on the Mac by the
+c13 engine from its embedded sources (0659: reserved shared buffers). Notarised: Metal library
+ca3dd0d1-2833-4a85-8aee-71b39952c641 (Accepted), loader eea7c66d-e8a4-4860-b9da-71e9416a2805.
+Gated on the Mac mini with the 2610141255001 engine and THIS signed library in the executable directory: Whistle 254 / 254 (the encoder on MTL0), signal stop 9 / 9, single-prompt teardown 7 / 7.

components/macos/2610141455005/ggml-metal-aarch64.dylib

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:82449bd0be6041f991ef001e6ea6cbeb057a238e33b75fa9ea3b9d47b5029e25
+size 1652688

components/media/2610071902001/BUILD_INFO.media.md

@@ -33,7 +33,7 @@ llamafile/oc-codec/oc_codec_sidecar.h f2c9001d1414aa2a55533b867b79c52a52c682558e
 Apple notarisation (Accepted), submission id(s): `83ee464c-3816-4b5c-9da7-cd4313840f25 aa53f9d9-14ba-443a-a059-64f82416f86a`.
 
 The twelve sidecars of c11, unchanged (same `media` digest): oc-codec, oc-audiocpp, oc-espeak, three per platform.
-Not re-gated as a set at this cut (same bytes, same digest). xollama loaded the codec sidecar beside the c12 engine on Linux x86_64 (mp3 input to Whistle, 2026-10-13).
+Not re-gated as a set at this cut (same bytes, same digest).
 
 ---
 

components/sbsa/2610141455003/BUILD_INFO.sbsa.md

@@ -0,0 +1,29 @@
+# sbsa 2610141455003
+
+Component of the opencoti pin format v2 (docs/protocols/PIN_FORMAT.md).
+
+Built from the engine source of build 2610141255001; gated floor (engine-min) 2610141212001; gated with nothing — NOT run.
+
+| file | platform | kind | sha256 | bytes |
+|---|---|---|---|---|
+| `ggml-cuda-sbsa-aarch64.so` | aarch64 | cuda | `da4b31254b54458f78e94a63c0d4cd54a3388664b468f8cb3609af06d77354a3` | 345350128 |
+
+ABI `ggml` = `bac5050ddc8c58c51f0b059a7f9753b33af0ed79d355090e7c9a17ffaaaf340a` (exported) — sha256 of this list:
+
+```
+llama.cpp/ggml/include/ggml-alloc.h 94e4cd069b9313b2ceb35dacec901981e0bb478d8bb31035b7126be091998c23
+llama.cpp/ggml/include/ggml-backend.h 764d1640ad62579513c4837539708f89e8603c9dbeae053a13d1cbd84a82c865
+llama.cpp/ggml/include/ggml-cuda.h 674ea064021cf5b9970e78f1209e513dd23350b94a44a705a27eec1423fe4b0a
+llama.cpp/ggml/include/ggml-metal.h 322f36cd30f3e9e7aad7b5b9bc63078012fd0b7706ac4e24f5721b604f3d8980
+llama.cpp/ggml/include/ggml-neo-pipeline.h 190f39e183542d16588a3104d8cb4359bf52fc7a62afe8d8cc4c822f4085086d
+llama.cpp/ggml/include/ggml-oc-kvarn-range.h 76635fdd8978632f670640ddf5125a666441e3d47c680bb8b5c0a099e4861202
+llama.cpp/ggml/include/ggml-oc-ledger.h c1be11d69a57a1e721f1ae8720b87887d239ac0183d4277fbaa07f1be257d20c
+llama.cpp/ggml/include/ggml-vulkan.h f9df4f7161a84ce7600d53c3fac8a62837d5716676c9f38cd340699ef1c070d9
+llama.cpp/ggml/include/ggml.h ecfbf114f9e9151d22e30ef6d5139570c5c41b595843ca4f99125bd0196709aa
+llama.cpp/ggml/include/opencoti-devtype.h 94dea946173449bb76bffd7f733ab822fe658fe7198992de750ac012d29308ec
+llama.cpp/ggml/src/ggml-backend-impl.h f9fdfe6bae61d1387593b40fc0f90ceee490c6a339ec6d7388c51e4fca318d8f
+llamafile/gpu_backend.h 4929c36c034b6589e88653bcb7a5c5a225743a07320f64c0857d92bdaf048208
+```
+
+CUDA 13 for Linux aarch64 (SBSA: DGX Spark / GB10 = sm_121, Jetson Thor). Chain 0662, from the c12 object cache.
+**Never run: there is no such machine in the test fleet.**

components/sbsa/2610141455003/ggml-cuda-sbsa-aarch64.so

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:da4b31254b54458f78e94a63c0d4cd54a3388664b468f8cb3609af06d77354a3
+size 345350128

components/vulkan/2610141455004/BUILD_INFO.vulkan.md

@@ -0,0 +1,31 @@
+# vulkan 2610141455004
+
+Component of the opencoti pin format v2 (docs/protocols/PIN_FORMAT.md).
+
+Built from the engine source of build 2610141255001; gated floor (engine-min) 2610141212001; gated with 2610141212001, 2610141255001.
+
+| file | platform | kind | sha256 | bytes |
+|---|---|---|---|---|
+| `ggml-vulkan-x86_64.so` | x86_64 | vulkan | `0b045791d249ebef909c517324d465d38f704c8d7f8e3fc5afda8a5443d76732` | 61336208 |
+| `ggml-vulkan-win-x86_64.dll` | win-x86_64 | vulkan | `c81d6f307eb5029fd76b1861bf6d64e61df4598494623b0c6d46635f3c5f804b` | 62378928 |
+
+ABI `ggml` = `bac5050ddc8c58c51f0b059a7f9753b33af0ed79d355090e7c9a17ffaaaf340a` (exported) — sha256 of this list:
+
+```
+llama.cpp/ggml/include/ggml-alloc.h 94e4cd069b9313b2ceb35dacec901981e0bb478d8bb31035b7126be091998c23
+llama.cpp/ggml/include/ggml-backend.h 764d1640ad62579513c4837539708f89e8603c9dbeae053a13d1cbd84a82c865
+llama.cpp/ggml/include/ggml-cuda.h 674ea064021cf5b9970e78f1209e513dd23350b94a44a705a27eec1423fe4b0a
+llama.cpp/ggml/include/ggml-metal.h 322f36cd30f3e9e7aad7b5b9bc63078012fd0b7706ac4e24f5721b604f3d8980
+llama.cpp/ggml/include/ggml-neo-pipeline.h 190f39e183542d16588a3104d8cb4359bf52fc7a62afe8d8cc4c822f4085086d
+llama.cpp/ggml/include/ggml-oc-kvarn-range.h 76635fdd8978632f670640ddf5125a666441e3d47c680bb8b5c0a099e4861202
+llama.cpp/ggml/include/ggml-oc-ledger.h c1be11d69a57a1e721f1ae8720b87887d239ac0183d4277fbaa07f1be257d20c
+llama.cpp/ggml/include/ggml-vulkan.h f9df4f7161a84ce7600d53c3fac8a62837d5716676c9f38cd340699ef1c070d9
+llama.cpp/ggml/include/ggml.h ecfbf114f9e9151d22e30ef6d5139570c5c41b595843ca4f99125bd0196709aa
+llama.cpp/ggml/include/opencoti-devtype.h 94dea946173449bb76bffd7f733ab822fe658fe7198992de750ac012d29308ec
+llama.cpp/ggml/src/ggml-backend-impl.h f9fdfe6bae61d1387593b40fc0f90ceee490c6a339ec6d7388c51e4fca318d8f
+llamafile/gpu_backend.h 4929c36c034b6589e88653bcb7a5c5a225743a07320f64c0857d92bdaf048208
+```
+
+Built from the clean-room tree of chain 0662. Carries the reserved (sparse) buffers of 0659: the default slots cost nothing
+until used.
+Gated with the 2610141255001 engine: the Vulkan cases of the bs2 56 / 56 gate (KLD 0.000311 against the CUDA base, needle 3 / 3 at 112k; two non-Bonsai models BIT-IDENTICAL between the c12 library and this one); the slot gate on Vulkan 6 / 6 (GPU 0 only); the glibc 2.28 image; the Windows DLL on eleven2go (RTX 3090): 64 tokens at 48.4 tok/s, the memory on the adapter the counter names, no debug-level line without -lv.

components/vulkan/2610141455004/ggml-vulkan-win-x86_64.dll

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c81d6f307eb5029fd76b1861bf6d64e61df4598494623b0c6d46635f3c5f804b
+size 62378928

components/vulkan/2610141455004/ggml-vulkan-x86_64.so

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0b045791d249ebef909c517324d465d38f704c8d7f8e3fc5afda8a5443d76732
+size 61336208