We have prepared tool to gather statistics during execution about TT performance. The main goal is to prepare some tests to evaluate the overall performance of the TT and based on them derive most efficient implementation of replacing scheme to maximise successful reads from the table. Currently TT is one item - one bucket type with simple depth preferred replacing scheme. We should try some bucket type table with mixed strategy for example depth preferred / recent one. At the moment of the issue creation the average is about 35/40% successful reads.
We have prepared tool to gather statistics during execution about TT performance. The main goal is to prepare some tests to evaluate the overall performance of the TT and based on them derive most efficient implementation of replacing scheme to maximise successful reads from the table. Currently TT is one item - one bucket type with simple depth preferred replacing scheme. We should try some bucket type table with mixed strategy for example depth preferred / recent one. At the moment of the issue creation the average is about 35/40% successful reads.