Efficient Sparse Matrix-Vector Multiplication on x86-Based Many-Core Processors