chokevin · chokevin · Apr 20, 2026 · Apr 20, 2026
diff --git a/docs/profiling/INDEX.md b/docs/profiling/INDEX.md
@@ -0,0 +1,8 @@
+# Autoresearch run index
+
+One row per profiling run produced by the swordfish-autoresearch chart.
+Newest first. PR column links to the draft PR carrying the artifacts.
+
+| timestamp (UTC) | source SHA | shapes | impls | GPU | 8b-b1 marlin TFLOPS | run dir | PR |
+|---|---|---|---|---|---|---|---|
+| 20260420T020830Z | `eb0f6e3` | voice | fp16,marlin | NVIDIA A100-SXM4-80GB | 0.7 | [`20260420T020830Z/`](./marlin/20260420T020830Z/) | [link](https://github.com/chokevin/swordfish/pull/4) |
diff --git a/docs/profiling/marlin/20260420T020830Z/70b-tp2-b1.ncu-csv.log b/docs/profiling/marlin/20260420T020830Z/70b-tp2-b1.ncu-csv.log
diff --git a/docs/profiling/marlin/20260420T020830Z/70b-tp2-b1.ncu.csv b/docs/profiling/marlin/20260420T020830Z/70b-tp2-b1.ncu.csv
@@ -0,0 +1,10 @@
+==PROF== Connected to process 1198 (/usr/bin/python3.10)
+
+==ERROR== An error was reported by the driver
+
+==ERROR== Profiling failed because a driver resource was unavailable. Ensure that no other tool (like DCGM) is concurrently collecting profiling data. See https://docs.nvidia.com/nsight-compute/ProfilingGuide/index.html#faq for more details.
+==ERROR== Failed to profile "distribution_elementwise_grid..." in process 1198
+==PROF== Trying to shutdown target application
+==ERROR== The application returned an error code (9).
+==ERROR== An error occurred while trying to profile.
+==WARNING== No kernels were profiled.
diff --git a/docs/profiling/marlin/20260420T020830Z/70b-tp2-b1.ncu.log b/docs/profiling/marlin/20260420T020830Z/70b-tp2-b1.ncu.log
@@ -0,0 +1,10 @@
+==PROF== Connected to process 1117 (/usr/bin/python3.10)
+
+==ERROR== An error was reported by the driver
+
+==ERROR== Profiling failed because a driver resource was unavailable. Ensure that no other tool (like DCGM) is concurrently collecting profiling data. See https://docs.nvidia.com/nsight-compute/ProfilingGuide/index.html#faq for more details.
+==ERROR== Failed to profile "distribution_elementwise_grid..." in process 1117
+==PROF== Trying to shutdown target application
+==ERROR== The application returned an error code (9).
+==ERROR== An error occurred while trying to profile.
+==WARNING== No kernels were profiled.
diff --git a/docs/profiling/marlin/20260420T020830Z/70b-tp2-b4.ncu-csv.log b/docs/profiling/marlin/20260420T020830Z/70b-tp2-b4.ncu-csv.log
diff --git a/docs/profiling/marlin/20260420T020830Z/70b-tp2-b4.ncu.csv b/docs/profiling/marlin/20260420T020830Z/70b-tp2-b4.ncu.csv
@@ -0,0 +1,10 @@
+==PROF== Connected to process 1360 (/usr/bin/python3.10)
+
+==ERROR== An error was reported by the driver
+
+==ERROR== Profiling failed because a driver resource was unavailable. Ensure that no other tool (like DCGM) is concurrently collecting profiling data. See https://docs.nvidia.com/nsight-compute/ProfilingGuide/index.html#faq for more details.
+==ERROR== Failed to profile "distribution_elementwise_grid..." in process 1360
+==PROF== Trying to shutdown target application
+==ERROR== The application returned an error code (9).
+==ERROR== An error occurred while trying to profile.
+==WARNING== No kernels were profiled.
diff --git a/docs/profiling/marlin/20260420T020830Z/70b-tp2-b4.ncu.log b/docs/profiling/marlin/20260420T020830Z/70b-tp2-b4.ncu.log
@@ -0,0 +1,10 @@
+==PROF== Connected to process 1279 (/usr/bin/python3.10)
+
+==ERROR== An error was reported by the driver
+
+==ERROR== Profiling failed because a driver resource was unavailable. Ensure that no other tool (like DCGM) is concurrently collecting profiling data. See https://docs.nvidia.com/nsight-compute/ProfilingGuide/index.html#faq for more details.
+==ERROR== Failed to profile "distribution_elementwise_grid..." in process 1279
+==PROF== Trying to shutdown target application
+==ERROR== The application returned an error code (9).
+==ERROR== An error occurred while trying to profile.
+==WARNING== No kernels were profiled.
diff --git a/docs/profiling/marlin/20260420T020830Z/70b-tp2-b8.ncu-csv.log b/docs/profiling/marlin/20260420T020830Z/70b-tp2-b8.ncu-csv.log
diff --git a/docs/profiling/marlin/20260420T020830Z/70b-tp2-b8.ncu.csv b/docs/profiling/marlin/20260420T020830Z/70b-tp2-b8.ncu.csv
@@ -0,0 +1,10 @@
+==PROF== Connected to process 1522 (/usr/bin/python3.10)
+
+==ERROR== An error was reported by the driver
+
+==ERROR== Profiling failed because a driver resource was unavailable. Ensure that no other tool (like DCGM) is concurrently collecting profiling data. See https://docs.nvidia.com/nsight-compute/ProfilingGuide/index.html#faq for more details.
+==ERROR== Failed to profile "distribution_elementwise_grid..." in process 1522
+==PROF== Trying to shutdown target application
+==ERROR== The application returned an error code (9).
+==ERROR== An error occurred while trying to profile.
+==WARNING== No kernels were profiled.
diff --git a/docs/profiling/marlin/20260420T020830Z/70b-tp2-b8.ncu.log b/docs/profiling/marlin/20260420T020830Z/70b-tp2-b8.ncu.log
@@ -0,0 +1,10 @@
+==PROF== Connected to process 1441 (/usr/bin/python3.10)
+
+==ERROR== An error was reported by the driver
+
+==ERROR== Profiling failed because a driver resource was unavailable. Ensure that no other tool (like DCGM) is concurrently collecting profiling data. See https://docs.nvidia.com/nsight-compute/ProfilingGuide/index.html#faq for more details.
+==ERROR== Failed to profile "distribution_elementwise_grid..." in process 1441
+==PROF== Trying to shutdown target application
+==ERROR== The application returned an error code (9).
+==ERROR== An error occurred while trying to profile.
+==WARNING== No kernels were profiled.
diff --git a/docs/profiling/marlin/20260420T020830Z/8b-b1.ncu-csv.log b/docs/profiling/marlin/20260420T020830Z/8b-b1.ncu-csv.log
diff --git a/docs/profiling/marlin/20260420T020830Z/8b-b1.ncu.csv b/docs/profiling/marlin/20260420T020830Z/8b-b1.ncu.csv
@@ -0,0 +1,10 @@
+==PROF== Connected to process 712 (/usr/bin/python3.10)
+
+==ERROR== An error was reported by the driver
+
+==ERROR== Profiling failed because a driver resource was unavailable. Ensure that no other tool (like DCGM) is concurrently collecting profiling data. See https://docs.nvidia.com/nsight-compute/ProfilingGuide/index.html#faq for more details.
+==ERROR== Failed to profile "distribution_elementwise_grid..." in process 712
+==PROF== Trying to shutdown target application
+==ERROR== The application returned an error code (9).
+==ERROR== An error occurred while trying to profile.
+==WARNING== No kernels were profiled.
diff --git a/docs/profiling/marlin/20260420T020830Z/8b-b1.ncu.log b/docs/profiling/marlin/20260420T020830Z/8b-b1.ncu.log
@@ -0,0 +1,10 @@
+==PROF== Connected to process 631 (/usr/bin/python3.10)
+
+==ERROR== An error was reported by the driver
+
+==ERROR== Profiling failed because a driver resource was unavailable. Ensure that no other tool (like DCGM) is concurrently collecting profiling data. See https://docs.nvidia.com/nsight-compute/ProfilingGuide/index.html#faq for more details.
+==ERROR== Failed to profile "distribution_elementwise_grid..." in process 631
+==PROF== Trying to shutdown target application
+==ERROR== The application returned an error code (9).
+==ERROR== An error occurred while trying to profile.
+==WARNING== No kernels were profiled.
diff --git a/docs/profiling/marlin/20260420T020830Z/8b-b4.ncu-csv.log b/docs/profiling/marlin/20260420T020830Z/8b-b4.ncu-csv.log
diff --git a/docs/profiling/marlin/20260420T020830Z/8b-b4.ncu.csv b/docs/profiling/marlin/20260420T020830Z/8b-b4.ncu.csv
@@ -0,0 +1,10 @@
+==PROF== Connected to process 874 (/usr/bin/python3.10)
+
+==ERROR== An error was reported by the driver
+
+==ERROR== Profiling failed because a driver resource was unavailable. Ensure that no other tool (like DCGM) is concurrently collecting profiling data. See https://docs.nvidia.com/nsight-compute/ProfilingGuide/index.html#faq for more details.
+==ERROR== Failed to profile "distribution_elementwise_grid..." in process 874
+==PROF== Trying to shutdown target application
+==ERROR== The application returned an error code (9).
+==ERROR== An error occurred while trying to profile.
+==WARNING== No kernels were profiled.
diff --git a/docs/profiling/marlin/20260420T020830Z/8b-b4.ncu.log b/docs/profiling/marlin/20260420T020830Z/8b-b4.ncu.log
@@ -0,0 +1,10 @@
+==PROF== Connected to process 793 (/usr/bin/python3.10)
+
+==ERROR== An error was reported by the driver
+
+==ERROR== Profiling failed because a driver resource was unavailable. Ensure that no other tool (like DCGM) is concurrently collecting profiling data. See https://docs.nvidia.com/nsight-compute/ProfilingGuide/index.html#faq for more details.
+==ERROR== Failed to profile "distribution_elementwise_grid..." in process 793
+==PROF== Trying to shutdown target application
+==ERROR== The application returned an error code (9).
+==ERROR== An error occurred while trying to profile.
+==WARNING== No kernels were profiled.
diff --git a/docs/profiling/marlin/20260420T020830Z/8b-b8.ncu-csv.log b/docs/profiling/marlin/20260420T020830Z/8b-b8.ncu-csv.log
diff --git a/docs/profiling/marlin/20260420T020830Z/8b-b8.ncu.csv b/docs/profiling/marlin/20260420T020830Z/8b-b8.ncu.csv
@@ -0,0 +1,10 @@
+==PROF== Connected to process 1036 (/usr/bin/python3.10)
+
+==ERROR== An error was reported by the driver
+
+==ERROR== Profiling failed because a driver resource was unavailable. Ensure that no other tool (like DCGM) is concurrently collecting profiling data. See https://docs.nvidia.com/nsight-compute/ProfilingGuide/index.html#faq for more details.
+==ERROR== Failed to profile "distribution_elementwise_grid..." in process 1036
+==PROF== Trying to shutdown target application
+==ERROR== The application returned an error code (9).
+==ERROR== An error occurred while trying to profile.
+==WARNING== No kernels were profiled.
diff --git a/docs/profiling/marlin/20260420T020830Z/8b-b8.ncu.log b/docs/profiling/marlin/20260420T020830Z/8b-b8.ncu.log
@@ -0,0 +1,10 @@
+==PROF== Connected to process 955 (/usr/bin/python3.10)
+
+==ERROR== An error was reported by the driver
+
+==ERROR== Profiling failed because a driver resource was unavailable. Ensure that no other tool (like DCGM) is concurrently collecting profiling data. See https://docs.nvidia.com/nsight-compute/ProfilingGuide/index.html#faq for more details.
+==ERROR== Failed to profile "distribution_elementwise_grid..." in process 955
+==PROF== Trying to shutdown target application
+==ERROR== The application returned an error code (9).
+==ERROR== An error occurred while trying to profile.
+==WARNING== No kernels were profiled.
diff --git a/docs/profiling/marlin/20260420T020830Z/SUMMARY.md b/docs/profiling/marlin/20260420T020830Z/SUMMARY.md
@@ -0,0 +1,26 @@
+# Autoresearch run `20260420T020830Z`
+
+- **source SHA:** `eb0f6e3`
+- **GPU:** NVIDIA A100-SXM4-80GB (cc 8.0, 79.3 GB)
+- **CUDA / torch / triton:** 12.4 / 2.4.0a0+07cecf4168.nv24.05 / 3.0.0
+- **shapes:** `voice`  **impls:** `fp16,marlin`  **repeats:** 5
+- **marlin SHA:** `1f25790bdd49fba53106164a24666dade68d7c90`
+
+## Results
+
+| shape | impl | ms_mean | ms_p95 | TFLOPS | speedup vs fp16 | error |
+|---|---|---|---|---|---|---|
+| 8b-b1 | fp16 | 0.031 | 0.032 | 1.1 | x1.00 | |
+| 8b-b1 | marlin | 0.050 | 0.055 | 0.7 | x0.62 | |
+| 8b-b4 | fp16 | 0.031 | 0.031 | 4.3 | x1.00 | |
+| 8b-b4 | marlin | 0.049 | 0.049 | 2.7 | x0.63 | |
+| 8b-b8 | fp16 | 0.031 | 0.032 | 8.6 | x1.00 | |
+| 8b-b8 | marlin | 0.049 | 0.050 | 5.5 | x0.63 | |
+| 70b-tp2-b1 | fp16 | 0.050 | 0.055 | 1.3 | x1.00 | |
+| 70b-tp2-b1 | marlin | 0.051 | 0.052 | 1.3 | x0.99 | |
+| 70b-tp2-b4 | fp16 | 0.049 | 0.049 | 5.5 | x1.00 | |
+| 70b-tp2-b4 | marlin | 0.049 | 0.050 | 5.4 | x1.00 | |
+| 70b-tp2-b8 | fp16 | 0.049 | 0.050 | 10.9 | x1.00 | |
+| 70b-tp2-b8 | marlin | 0.049 | 0.050 | 10.9 | x1.00 | |
+
+![roofline](./roofline.png)
diff --git a/docs/profiling/marlin/20260420T020830Z/env.txt b/docs/profiling/marlin/20260420T020830Z/env.txt
@@ -0,0 +1,43 @@
+=== profile_marlin.sh @ 20260420T020830Z ===
+--- host ---
+Linux swordfish-profile-sf-prof-260420-020113-847z9 6.6.126.1-1.azl3 #1 SMP PREEMPT_DYNAMIC Wed Mar  4 05:04:40 UTC 2026 x86_64 x86_64 x86_64 GNU/Linux
+--- nvidia-smi ---
+Mon Apr 20 02:08:30 2026       
++-----------------------------------------------------------------------------------------+
+| NVIDIA-SMI 580.105.08             Driver Version: 580.105.08     CUDA Version: 13.0     |
++-----------------------------------------+------------------------+----------------------+
+| GPU  Name                 Persistence-M | Bus-Id          Disp.A | Volatile Uncorr. ECC |
+| Fan  Temp   Perf          Pwr:Usage/Cap |           Memory-Usage | GPU-Util  Compute M. |
+|                                         |                        |               MIG M. |
+|=========================================+========================+======================|
+|   0  NVIDIA A100-SXM4-80GB          On  |   0000000B:00:00.0 Off |                    0 |
+| N/A   35C    P0             69W /  400W |       0MiB /  81920MiB |      0%      Default |
+|                                         |                        |             Disabled |
++-----------------------------------------+------------------------+----------------------+
+
++-----------------------------------------------------------------------------------------+
+| Processes:                                                                              |
+|  GPU   GI   CI              PID   Type   Process name                        GPU Memory |
+|        ID   ID                                                               Usage      |
+|=========================================================================================|
+|  No running processes found                                                             |
++-----------------------------------------------------------------------------------------+
+--- nvcc ---
+nvcc: NVIDIA (R) Cuda compiler driver
+Copyright (c) 2005-2024 NVIDIA Corporation
+Built on Thu_Mar_28_02:18:24_PDT_2024
+Cuda compilation tools, release 12.4, V12.4.131
+Build cuda_12.4.r12.4/compiler.34097967_0
+--- nsys ---
+NVIDIA Nsight Systems version 2024.2.1.106-242134037904v0
+--- ncu ---
+NVIDIA (R) Nsight Compute Command Line Profiler
+Copyright (c) 2018-2024 NVIDIA Corporation
+Version 2024.1.1.0 (build 33998838) (public-release)
+--- python / torch / triton / marlin ---
+python 3.10.12
+torch 2.4.0a0+07cecf4168.nv24.05 cuda 12.4
+triton 3.0.0
+marlin unknown
+--- repo SHA ---
+eb0f6e31ca0e2e4d75507f9e77bfba6738c3a22a