claude-noether 67316774a0 ggml-rocket: make ROCKET_MIN_BATCH runtime-tunable via GGML_ROCKET_MIN_BATCH ...

The offload gate (M < 32 -> CPU) was compile-time only. Reading the env var
(default 32 unchanged) gives a clean NPU on/off toggle: GGML_ROCKET_MIN_BATCH=999999
forces CPU-only for benchmarking without rmmod or a CPU-only build. Used for the
2026-07-18 ampere NPU-vs-CPU gemma prefill benchmark.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01EWpfhDgYNA21tETDP9ueBE

2026-07-18 15:21:55 +02:00

..

cmake

ggml : Parallelize quant LUT init (#23595)

2026-05-25 10:15:46 +03:00

include

TP: quantized KV cache support (#23792)

2026-06-01 12:30:10 +02:00

src

ggml-rocket: make ROCKET_MIN_BATCH runtime-tunable via GGML_ROCKET_MIN_BATCH

2026-07-18 15:21:55 +02:00

.gitignore

…

CMakeLists.txt

ggml: add rocket NPU backend (RK3588 mainline accel driver)

2026-07-14 10:52:06 +02:00