#!/usr/bin/env bash
# Build llama.cpp with CUDA for the 4090 (sm_89).
#
# Pinned to a tag rather than master: qwen35 arch support and the current
# --spec-type names are recent, and master moves fast.
set -euo pipefail

TAG="${TAG:-b10448}"
LLAMA_DIR="${LLAMA_DIR:-$HOME/code/llama.cpp}"

export PATH=/usr/local/cuda-13.2/bin:$PATH
export CUDACXX=/usr/local/cuda-13.2/bin/nvcc

command -v nvcc >/dev/null || { echo "nvcc missing — run 00-cuda.sh first" >&2; exit 1; }

if [[ ! -d "$LLAMA_DIR" ]]; then
    git clone --depth 1 --branch "$TAG" https://github.com/ggml-org/llama.cpp.git "$LLAMA_DIR"
fi

cd "$LLAMA_DIR"

cmake -B build \
    -DGGML_CUDA=ON \
    -DCMAKE_CUDA_ARCHITECTURES=89 \
    -DLLAMA_CURL=ON \
    -DCMAKE_BUILD_TYPE=Release

cmake --build build --config Release -j "$(nproc)"

./build/bin/llama-server --version 2>&1 | head -3