A Quantitative Performance Analysis Model for GPU Architectures