How Swiss tables work in Go built-in map

summary

Go 1.24 replaces its old map implementation with Swiss Tables to reduce pointer chasing and improve lookup speeds. This new design utilizes control bytes and SIMD instructions to scan multiple slots simultaneously, significantly decreasing the time spent on key equality checks. By organizing storage into groups and tables that grow independently, the runtime achieves better CPU cache locality and more efficient memory redistribution during growth.

key points

1. Swiss Tables use a 7-bit hash (H2) stored in control bytes, allowing the runtime to use SIMD instructions to identify potential key matches across eight slots at once without reading the full keys.
2. The map structure evolves from a single group to a table-based system that uses a triangular probe sequence to resolve collisions, avoiding the long overflow bucket chains of previous versions.
3. To prevent expensive global redistributions, Go employs a directory system with global and local depths, allowing individual tables to split and grow without requiring all other tables to be rebuilt.

who should read it

Software engineers and Go developers who want to understand the internal memory layout and performance optimizations of the Go runtime.

discussion positive