There are several factors that motivated us to develop a coarse-grained mechanism over a fine-grained approach when implementing pointer swizzling at page fault time in Texas. The primary motivation is the fact that we wanted to exploit existing hardware to avoid expensive residency checks in software. However, we believe that there are also other factors against using a fine-grained approach as the primary address translation mechanism. In this section, we discuss fine-grained address translation techniques and why we believe that they are not practical for high-performance implementations in terms of efficiency and complexity.
Overall, fine-grained address translation techniques are likely to incur various hidden costs that have not been measured and quantified in previous research. In general, we have found most current fine-grained schemes appear to be slower than pointer swizzling at page fault time in terms of the basic address translation performance.