#!/usr/bin/env bash
# Builds matmul_blis.c to public/matmul/matmul_blis_{st,mt}.{js,wasm}, linking
# the WASM libflame + BLIS static libraries from a sibling libflame2wasm
# checkout. Override its location with LIBFLAME=/path/to/libflame2wasm.
#
# Produce those .a files first with libflame2wasm's build-all-wasm.sh, which
# yields install/lib/{libflame.a,libflame-mt.a,libblis-st.a} and
# blis/lib/generic/libblis.a. The matmul_blis_* outputs here are NOT committed
# (see .gitignore); CI rebuilds them (see .github/workflows/deploy.yml).
#
# Requires: emsdk (expected at ~/emsdk, or already on PATH).
set -euo pipefail

cd "$(dirname "$0")"
LIBFLAME=${LIBFLAME:-../../../libflame2wasm}

if [ ! -f "$LIBFLAME/install/lib/libflame.a" ]; then
  echo "error: prebuilt libraries not found under $LIBFLAME (set LIBFLAME=...)" >&2
  exit 1
fi

if ! command -v emcc >/dev/null 2>&1; then
  source ~/emsdk/emsdk_env.sh
fi

OUT=../public/matmul
mkdir -p "$OUT"

COMMON="-O2 -sMODULARIZE -sENVIRONMENT=web,worker,node \
  -sEXPORTED_FUNCTIONS=_matmul_blis,_malloc,_free \
  -sEXPORTED_RUNTIME_METHODS=HEAPF64"

# Single-threaded: BLIS(st) linked first so its BLAS symbols win over
# libflame's f2c reference BLAS. No threads -> no SharedArrayBuffer needed.
emcc $COMMON matmul_blis.c \
  -sEXPORT_NAME=createMatmulBlisST \
  -sALLOW_MEMORY_GROWTH \
  "$LIBFLAME/install/lib/libblis-st.a" \
  "$LIBFLAME/install/lib/libflame.a" \
  -o "$OUT/matmul_blis_st.js"

# Threaded: everything compiled with -pthread; fixed shared memory (growth is
# costly with shared memory). Pool must cover max threads + margin.
emcc $COMMON -pthread -DBLIS_MT matmul_blis.c \
  -sEXPORT_NAME=createMatmulBlisMT \
  -sPTHREAD_POOL_SIZE=20 \
  -sINITIAL_MEMORY=768MB \
  "$LIBFLAME/blis/lib/generic/libblis.a" \
  "$LIBFLAME/install/lib/libflame-mt.a" \
  -o "$OUT/matmul_blis_mt.js"

echo "Built public/matmul/matmul_blis_{st,mt}.{js,wasm}"