mlboydaisuke commited on
Commit
5ec4060
·
verified ·
1 Parent(s): c3b8e82

Add measured Galaxy S26 NPU/GPU section

Browse files

Numbers from the S26 classic sweep. Primary logs:
litertlm-convert/community_accel_work/s2_npu_sweep/results.jsonl (on-device
JIT) and portal_work/bench_all.txt + bench_redo.txt (the 2026-08-23 AOT
sweep); every row states which of the two produced it. Rows carry device,
SoC, runtime, N and thermal state, and join to this repo by exact .tflite
file name. Rows taken outside the sweep's thermal window are not reported.

Files changed (1) hide show
  1. README.md +15 -0
README.md CHANGED
@@ -99,6 +99,21 @@ Measured on a **Pixel 8a** (Tensor G3, Android 16) with the standard TFLite [`be
99
 
100
  XNNPACK declines these fp16 graphs — it reports `failed to delegate DEPTHWISE_CONV_2D` and then fails to allocate tensors — so there is no usable CPU number. Disabling XNNPACK falls back to reference kernels, which measured about 20× slower than the GPU on models of this size and would not represent CPU inference anyone would ship.
101
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
102
  ## License
103
 
104
  [MIT](https://github.com/megvii-research/NAFNet/blob/main/LICENSE). Upstream:
 
99
 
100
  XNNPACK declines these fp16 graphs — it reports `failed to delegate DEPTHWISE_CONV_2D` and then fails to allocate tensors — so there is no usable CPU number. Disabling XNNPACK falls back to reference kernels, which measured about 20× slower than the GPU on models of this size and would not represent CPU inference anyone would ship.
101
 
102
+ ## Snapdragon NPU (Hexagon)
103
+
104
+ The NPU is **1.66x faster** than the GPU (18.58 ms against 30.89 ms) and loads 9.01x faster (204 ms against 1840 ms).
105
+
106
+ | backend | compiled | inference (median / min) | load |
107
+ |---|---|---:|---:|
108
+ | NPU (Hexagon v81) | on-device JIT | 18.58 ms / 18.07 ms | 204 ms |
109
+ | GPU (Adreno) | — | 30.89 ms / 30.04 ms | 1840 ms |
110
+
111
+ Measured on a **Samsung Galaxy S26** (Snapdragon 8 Elite Gen 5 / SM8850, Hexagon v81, Android 16) with LiteRT `CompiledModel` 2.2.0, one accelerator per process, 5 warm-up runs then N=50 timed runs, median reported. Every run held thermal status `NONE` throughout. Headroom 0.76, where 1.0 is the throttling threshold.
112
+
113
+ **The NPU rows ran the published file unchanged.** LiteRT compiled it for the Hexagon on the device at first load. That first compile took 88 s here. The `load` column above is the cached load every later run pays. Recipe and the runtime libraries it needs: [NPU guide](https://github.com/john-rocky/hf-to-litertlm/blob/main/docs/android-npu.md).
114
+
115
+ GPU wiring: [GPU guide](https://github.com/john-rocky/hf-to-litertlm/blob/main/docs/android-gpu.md).
116
+
117
  ## License
118
 
119
  [MIT](https://github.com/megvii-research/NAFNet/blob/main/LICENSE). Upstream: