So slower than tcc on O0 and slower than qbe on O1, with less backends, but tight llvm integration usable as baseline jit compiler, where it already outperforms

So slower than tcc on O0 and slower than qbe on O1, with less backends, but tight llvm integration usable as baseline jit compiler, where it already outperforms Cranelift.