Speed bench compile vs execution
Speed bench report v5.0.0 · distro 5.0.0 · full manual →
Belt triad comparison
bench-triad
Field vs host compare
bench-compare

Grok16 5.0.0 — Version One

5.0 is a clean start (v1.0 operator framing). Self-hosted g16 @ 16.2.0. One belt (belt_2_0). One 2D platform. No field files.

DO NOT CREATE FIELD FILES — they heat neighboring fields. Placement on the field platform *is* field at depth 0.

Benchmarks (Speed Bench 5.0.0)

ResultValue
Fastest execution102.8M ops/s — C++ g16 belt_2_0
Belt triadperformance.html#triad
Full tablespeed-bench.html

Start

git clone https://github.com/ZacharyGeurts/Grok16.git
cd Grok16 && git checkout v5.0.0
export G16_PREFIX="$(pwd)" G16_BELT_PROFILE=belt_2_0
./scripts/grok16-toolchain.sh rebuild && ./scripts/grok16-toolchain.sh verify

Binary: grok16-5.0.0-linux-x86_64.tar.gzsource grok16-env.sh./share/ammocode/ammocode

Docs