Performance Improvements

Following the last release, we continue to improve the performance of the DoMulti, DoMultiCache and the MGet, MGetCache helpers.

This time, by reusing intermediate allocations and batched cache lookups:

• DoMulti and MGet get about 13% throughput improvement and 50% allocation reduction in cluster mode.
• DoMultiCache gets about 60% throughput improvement and 65% allocation reduction in both single mode and cluster mode.
• MGetCache gets about 35% throughput improvement and 50% allocation reduction in both single mode and cluster mode.

Changes

• perf: use batched lru.Flights to reduce cache contentions (#288)
• perf: reuse intermediate allocations (#287)
• perf: reduce contentions caused by the usage of runtime.GOMAXPROCS (#286)

🚀 New Features

• feat: ForceSingleClientoption to force using the single client implementation (#284)

🐛 Bug Fixes

• fix: handle cluster slots with unknown IP correctly (#285)

Contributors

We'd like to thank all the contributors who worked on this release!

@fabienjuif and @rueian