unknown

The DoltHub team tested Go 1.25's experimental Green Tea garbage collector against their version-controlled SQL database to measure performance improvements. Using sysbench benchmarks with varying concurrency levels, they found no meaningful difference in throughput, median latency, or latency distribution between the standard and Green Tea collectors. Analysis of garbage collection traces revealed Green Tea actually spent slightly more CPU time during mark phases, though this didn't impact observable performance. The team concluded Green Tea offers no benefit for Dolt's workload and won't enable it in production builds.

We tried Go's experimental Green Tea garbage collector and it didn't help performance

A supportive squad for Go developers who want to dive deeper into what makes Go special. We focus on the language's clean simplicity, powerful concurrency, and reliable performance.