Correct GPU speedup: 14-16x (baseline was measured at the wrong thread count)

README.md

@@ -120,6 +120,20 @@ Sampling is `--temp` / `--top-k` / `--top-p` / `--seed`; the default is greedy
 and seed-reproducible. Generation stops at `<|im_end|>`, so `max_new` is a
 ceiling.
 
+### On an NVIDIA GPU
+
+The same package runs entirely on the GPU, with byte-identical output and
+roughly 14-16x the throughput. Opt-in, so the default build is unchanged:
+
+```bash
+cmake -S src/c -B build -DCASCADIA_CUDA=ON -DCMAKE_BUILD_TYPE=Release \
+    && cmake --build build -j
+./build/cascadia_generate_cuda ./pkg 512 --chat "Explain gradient descent."
+```
+
+Same arguments, same tokens. Needs CUDA and sm_75 or newer.
+
+
 ### Python
 
 ```python