Pgo - GoKawiil

Technical series explains second phase of profile-guided optimization in compiler builds

A blog post details how profile-guided optimization (PGO) works in its second installment, describing a two-pass compilation process where an instrumented first build records per-edge counters across branches, loops, and call sites. The author explains how these fine-grained counts feed into the compiler's second-pass decisions, starting with inlining choices that can override static size and cost heuristics based on actual execution frequency.

percona.com · 2026-10-14

Rust compiler sees 4.57% mean speedup in latest two-month performance report

A new Rust compiler performance report covering July 29 to September 28, 2026 shows a mean wall-time reduction of 4.57% across 629 benchmark measurements, with 555 improving and only 74 regressing. Key contributors included PGO enablement for Clippy (up to 18% faster in some cases), an upgrade to LLVM 23 that cut mean wall-time by 1.2%, and optimizations to the new Polonius Alpha borrow checker that reduced instruction counts for crates like serde by 3-5%.

nnethercote.github.io · 2026-10-01