#!/bin/bash
# Build BLIS (generic config) as a WebAssembly static library to serve as a
# fast BLAS backend for the WASM libflame build.
#
# Link order matters: put libblis.a BEFORE libflame.a so the linker resolves
# BLAS symbols from BLIS; libflame's built-in f2c BLAS members are then never
# pulled in (they remain only as fallback for routines BLIS doesn't provide,
# e.g. banded/packed level-2).
#
# Usage: ./build-blis-wasm.sh              (single-threaded)
#        THREADING=pthreads ./build-blis-wasm.sh   (wasm-threads build;
#            link apps with -pthread -sPTHREAD_POOL_SIZE=N, and all other
#            linked objects must also be compiled with -pthread)
set -euo pipefail

THREADING=${THREADING:-single}
EXTRA_CFLAGS=""
if [ "$THREADING" = "pthreads" ]; then
  EXTRA_CFLAGS="-pthread"
fi

cd "$(dirname "$0")"

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

if [ ! -d blis ]; then
  git clone --depth 1 https://github.com/flame/blis
fi
cd blis

# --- Patches to make BLIS's BLAS interface WASM-compatible with libflame ---
#
# WASM requires call sites and definitions to agree on exact signatures.
# libflame's f2c-derived code declares BLAS subroutines returning int and
# calls lsame_/xerbla_ with 2 args (no hidden Fortran string lengths).
# Stock BLIS uses void returns and 4-arg lsame_ / 3-arg xerbla_.

# 1. Drop BLIS's f2c-derived compat sources (banded/packed level-2 routines,
#    lsame_, xerbla_). libflame.a already provides all of them in the
#    2-arg/int f2c convention. Keep cabs1 (no conflicts) and all headers.
ls frame/compat/f2c/*.c | grep -v cabs1 | xargs rm -f
rm -f frame/compat/f2c/bla_xerbla_array.h
sed -i '/#include "bla_xerbla_array.h"/d' frame/compat/bli_blas_defs.h

# 2. Prototype lsame_/xerbla_ with 2 args to match libflame's definitions.
sed -i 's/, long ca_len, long cb_len//; s/, int ca_len, int cb_len//' frame/compat/f2c/bla_lsame.h
sed -i 's/, ftnlen srname_len//' frame/compat/f2c/bla_xerbla.h

# 3. Strip hidden string-length args from BLIS's internal lsame_/xerbla_ calls.
sed -i 's/, *(ftnlen)[0-9]*//g' frame/compat/check/*.h

# 4. Give every BLAS interface function an int return (f2c convention), so
#    signatures match libflame's declarations. Bare `return;` must become
#    `return 0;` (a hard error in C23 otherwise).
sed -i 's/void PASTEF77/int PASTEF77/g' \
    frame/compat/*.c frame/compat/*.h frame/compat/f2c/*.h \
    frame/compat/extra/*.c frame/compat/extra/*.h frame/compat/amd/*.c
sed -i 's/\breturn; \\/return 0; \\/g; s/^\( *\)return;$/\1return 0;/' \
    frame/compat/*.c frame/compat/check/*.h frame/compat/extra/*.c frame/compat/amd/*.c

# 5. BLIS tests the obsolete EMSCRIPTEN macro; modern emcc defines __EMSCRIPTEN__.
sed -i 's/defined(EMSCRIPTEN)/defined(__EMSCRIPTEN__)/' frame/include/bli_system.h

# --- Configure and build ---
# -msimd128 lets the generic C kernels autovectorize to wasm SIMD.
CC=emcc AR=emar RANLIB=emranlib CFLAGS="-msimd128 $EXTRA_CFLAGS" ./configure \
  --disable-shared --enable-static \
  --enable-blas --disable-cblas \
  --int-size=32 --blas-int-size=32 \
  --complex-return=gnu \
  --enable-threading="$THREADING" \
  generic

# The generic make_defs.mk rejects the "emcc" vendor string; emcc is clang.
make -j"$(nproc)" CC_VENDOR=clang

echo "Done. Library: $(pwd)/lib/generic/libblis.a"