Go good reads

Index

Performance
Concurrency
Generics
Reiaibility
Others

Performance

If you want to know how high performance systems written in Go were built, read VictoriaLogs: https://github.com/VictoriaMetrics/VictoriaLogs by Phuong Le

Custom bloom filters to reduce disk I/O and CPU cycles for redundant logs.

Column-oriented block storage for better compression ratio and cache locality.

Memory-mapped files with automatic pread fallback for zero-copy reads.

Lock-free object pools and arena allocators to reduce heap allocations.

Reflection-free JSON parsers for streaming at hundreds of MB/s per core.

Compile-time templates replace text/html parsing with near-printf speed.

Dictionary-aware Zstd compression balancing CPU cost against bandwidth savings.

Multi-core parallelism everywhere with adaptive concurrency limits.

Scatter-gather fan-out writes with adaptive concurrency for network saturation.

Fast hashing and lock-free randomization for minimal contention.

How Go 1.24's Swiss Tables saved us hundreds of gigabytes by Nayef Ghattas

Finding performance problems by diffing two Go profiles by Zach Musgrave

Optimising and Visualising Go Tests Parallelism: Why more cores don't speed up your Go tests

Leveraging benchstat Projections in Go Benchmark Analysis!

Benchmarks and performance testing

New unique package by Michael Knyszek

This work also led us to reexamine finalizers, resulting in another proposal for an easier-to-use and more efficient replacement for finalizers. With a hash function for comparable values on the way as well, the future of building memory-efficient caches in Go is bright!

It is wonderful to see that the Go team is working on improving the memory management of Go.

Inline Heuristics Review

On Thursday 3/28 we held a design review looking at the new inlining heuristics framework being developed as part of the inlining overhaul effort (this was a separate session from the usual C&R meeting)

From Go compiler and runtime meeting notes

Go 1.22 inlining overhaul

pprof documentation

A Deep Look Into Golang Profile-Guided Optimization (PGO)

Profiling and Execution Tracing in Go

What I learned
- pprof can diff with go tool pprof -http=:8080 -diff_base <file2> <file1>

How to Write Accurate Benchmarks in Go

Not resetting or pausing the timer
Making wrong assumptions about micro-benchmarks
Not being careful about compiler optimizations
Being fooled by the observer effect

Obscure Go Optimisations - Bryan Boreham

Take-aways
- Slice-to-interface cast will allocate.
- Heap allocation is costly.
- Generic methods are not fast.
- Heap Ballast is dead.
- Long live GOMEMLIMIT!