Why Are Local Go Slices Faster Than Local Arrays for Element Access?

Array vs Slice: Speed of Element Access

Question:

Despite expectations of arrays being faster than slices due to their contiguous nature, empirical testing reveals that local slices are significantly quicker than local arrays when accessing their elements. Why is this the case?

Background:

A benchmark test was conducted to compare the performance of accessing array and slice elements, including global and local variants. Surprisingly, the local slice emerged as the fastest option.

Answer:

An examination of the amd64 assembly of the local array and slice benchmarks reveals a potential culprit. While both methods employ similar operations, the array version repeatedly loads the array address from memory during each access:

In contrast, the slice version primarily operates on registers, performing only one initial load from memory:

Additionally, the array version invokes the runtime.duffcopy assembly routine, while the slice version does not. This additional runtime call may contribute to the slower performance of the array version.

Additional Notes:

The observed performance discrepancy persists regardless of variations in array/slice size, element type, and the order of benchmark execution.

The above is the detailed content of Why Are Local Go Slices Faster Than Local Arrays for Element Access?. For more information, please follow other related articles on the PHP Chinese website!